Mr G Bouchard v Facebook UK Ltd: 2202201/2020
JUDGMENT
The judgment of the Tribunal is that there was no transfer of an undertaking within the meaning of TUPE when the Claimant commenced employment with the Respondent so that the Claimant does not have continuity of service of more than 2 years and accordingly his claim for unfair dismissal is struck out.REASONS
[1]The preliminary issue directed to be heard by Employment Judge Palca was as follows: whether there was a transfer of undertaking within the meaning of regulation 3(1)a of the Transfer of Undertakings (Protection of Employment) Regulations 2006 (TUPE), between Bloomsbury AI and the Respondent in 2018 with the result that the Claimant has continuity of service at the Respondent pursuant to Regulation 4 of TUPE, of more than two years.[2]There was a second preliminary issue which the Tribunal postponed to be heard at a later date which related to whether the Claimant was at the material time disabled within the meaning of section 6 of the Equality Act 2010. Preliminary matter[3]At the outset of the hearing, the Respondent applied for an order striking out parts of a witness statement which had been submitted by the Claimant in support of his application that he was disabled within the meaning of the Equality Act. I heard submissions from both parties on the question. The relevant facts put briefly are that the Claimant has brought a claim against the Respondent which includes a claim for disability discrimination. At a closed Preliminary Hearing for case management purposes on 22 September 2020, Employment Judge Palca ordered that there be an open Preliminary Hearing starting today to answer two questions, one of which was whether or not the Claimant was disabled for the purposes of the Equality Act. The Order that Judge Palca made, which is recorded as made by consent, provided on the question of disability for disclosure of the medical evidence by 12 November 2020 and also for the Claimant to provide the Respondent with an impact statement by the same date. Having reviewed that evidence, the Respondent was then to inform the Tribunal and the Claimant of the extent to which it conceded disability and, if it did not, the reasons why in full. The Respondent did not concede disability and on 30 November wrote to set out the reasons why.[4]I have been referred to the Respondent’s letter which explained the reasons for their not conceding disability. The letter referred to the Equality Act guidance and then said; “From the evidence provided to date, it is not clear over what time period the Claimant’s impairment had an impact on his day to day activities. It is also of note and the Respondent would wish to explore this matter further, that to the extent the Claimant was formally diagnosed with an impairment in February 2018, this was done over the telephone by his ex-wife in France. In addition, the second episode referred to by the Claimant in October and November 2019 is referred to by his GP as “work related stress and anxiety” and not depression.”[5]The directions given by Judge Palca provided for the exchange of witness statements by 25 February 2021. The Claimant submitted his own witness statement and also a further witness statement from his ex-wife, Doctor Basson, who I understand had treated him in February 2018. Doctor Basson is a medical professional.[6]The Respondent’s application is to strike out parts of that witness statement, saying that in effect it amounts to an expert's report. The Respondent says expert evidence is inadmissible unless permission is given, and the Respondent has referred me to the CPR and the case of Darby Properties Limited and Darby Investments Limited v Lloyds Bank Plc [2016] EWHC 2494. Master Matthews decision recited quite significantly from Phipson on Evidence and from other legal authorities and set out the legal principles relating to expert evidence.[7]Both parties are aware that the Tribunal is not bound by the CPR. The Claimant effectively says it is not relevant because the Employment Tribunal is not bound by it. The Respondent says that it is helpful. 10.5 Reserved judgment with reasons – rule 62 2 March 2017[8]The Claimant says that the burden is on the Claimant to prove disability and that in response to the Respondent’s explanation of the reasons why it would not concede that issue, the Claimant obtained the evidence from Doctor Basson. I have been referred by the Claimant to the case of J v DLA Piper LLP UKEAT/0263/109/RN where the Tribunal failed to give weight to the GP’s evidence despite the fact that he was treating the individual concerned. In that case, it was noted by the EAT that a GP would have a good opportunity to see the individual and his explanation should have been given weight but was not mentioned in the judgment of the Tribunal. The Claimant says Doctor Basson’s evidence should be considered because Doctor Basson treated the Claimant and had a good opportunity to see him and so her evidence should be taken into account as the GP’s evidence in J v DLA should have been taken into account. The Claimant also made comments about this being the most cost effective route to providing the Tribunal with the medical information it needs.[9]It is my view that the most important guideline for the Tribunal is the overriding objective. One aspect of that is to ensure the parties are on an equal footing. The overriding objective also addresses the question of saving expense which the Claimant mentioned.[10]If at the outset, the Claimant wanted to rely on expert evidence, the Tribunal would have given directions to ensure that the parties were on an equal footing in that process. It is the Tribunal’s experience that individual experts are not helpful because they can leave the Tribunal with contradictory views. The Tribunal, which is not a medical expert, is then required to find some way to decide between them. It is therefore our practice, if there is to be a medical report, to have a joint report. When a joint report is prepared, the expert has to be appointed by a letter which is agreed by both parties and which sets out the relevant law in terms of what factors the Tribunal would want to hear about when determining disability, and also importantly explains that the expert is advising the Tribunal and not individual parties. The duty of the expert is clearly set out as a duty to the Tribunal.[11]The Claimant in this case chose to rely on an impact statement and on contemporaneous medical records. The Claimant elected that route. The Tribunal will give directions which have an option for the Claimant to proceed to obtain a joint expert report if the Respondent does not concede when supplied with the impact statement and the contemporaneous medical records, but the Claimant did not choose that way forward. After the Claimant got the Respondent’s objections, the Claimant sought to address those objections. Their explanation is that in order to address the matters raised by the Respondent, they went to Doctor Basson. However, it is clear that it was not necessary for the Claimant to produce a witness statement which addresses all the matters which I understand are covered by Doctor Basson’s witness statement, order to deal with the points which are highlighted in the Respondent’s letter.[12]I should say that I have not read Doctor Basson’s witness statement because of my concern that it would be impractical for me to do so and then to try to deal with excised parts of it, if that was the outcome of this application. I have to rely on the analysis of it prepared by the Respondent who tells me that it addresses a possible new diagnosis for the Claimant, the prevalence of certain conditions, the Claimant’s mental health at an earlier period, which I understand was not relied upon by the Claimant himself, the effects of his medication, the risk of recurrence, whether he should have been signed off work and the cause of an episode of 10.5 Reserved judgment with reasons – rule 62 3 March 2017 depression in 2019. It is not clear to me that Doctor Basson was treating the Claimant at that stage.[13]It is clear is that Doctor Basson’s witness statement goes considerably further than dealing with the issues raised by the Respondent in their letter. As such, where there was no permission given for an expert’s report, I cannot treat the witness evidence from Doctor Basson simply as if she was a witness of fact whose evidence is provided under the general order for witnesses of fact. I cannot ignore the fact that her witness statement goes into a number of matters which are matters of her professional opinion.[14]In order for the overriding objective to be addressed properly I have to put the parties on an equal footing. To that extent it is my intention to explore with the parties the options to address that. I am not prepared to review Doctor Basson’s witness statement in its entirety. Therefore, the options appear to me, either to excise parts of that witness statement and have a shorter witness statement dealing solely with what is necessary to deal with the issues raised by the Respondent or otherwise, if the Claimant wants expert evidence in relation to his claim, to consider how that can best be achieved in compliance with the overriding objective.[15]Having heard my decision on this point, the parties took some time to consider the situation and the Claimant elected to have directions given which enabled him to have medical evidence by way of a joint report and the question of whether he was a disabled person for the purposes of the Equality Act was adjourned to a later date.The Evidence
[16]I heard evidence from Lilia Sirota, a director of corporate development at the Respondent and from Tim Rocktaschel, a research scientist at the Respondent. I also heard evidence from the Claimant, and I was supplied with an agreed bundle of documents.[17]The hearing took place by CVP. Given the problems of the current pandemic this was a reasonable manner in which to conduct this hearing. The parties consented and the witnesses all gave evidence clearly and could be seen and heard by all parties. I was satisfied that each witness was giving evidence on his own account and there is no question that their evidence was interfered with in anyway in the course of giving evidence. Facts BloomsburyFacts
[18]The Claimant is a very highly qualified and experienced expert in artificial intelligence. He has a PhD in statistical machine learning and computer vision. He has worked as a research scientist and senior research scientist. He has specialised in machine learning and statistical natural language processing as well as optimisation tools. In 2015 the Claimant together with two other individuals, Sebastian and Luis, founded a company called Bloomsbury AI Limited (Bloomsbury”), which was based in London. A third individual joined them, Mr Rocktaschel, who, after an initial period as an employee with Bloomsbury, ceased to be an employee in about April 2017, having chosen to focus his career on 10.5 Reserved judgment with reasons – rule 62 4 March 2017 academia. Mr Rocktaschel became a consultant with Bloomsbury, supervising research projects for a short period each week.[19]Mr Rocktaschel describes how Bloomsbury’s business initially focused on finding potential applications for their research at UCL for Bloomsbury and involved developing research focused prototypes. He says that in or about early 2016, Bloomsbury sought to become a more commercially viable business. The Claimant says that during the first two years of the company, they had explored several different avenues to pitch and commercialise their particular expertise, before settling on the direction of research and development that Bloomsbury ultimately took in the summer of 2017. He explained that they had explored the possibility of identifying fake information online automatically. They realised this was mainly a problem facing large tech companies, so other online platforms would not be ready to pay for a service that specifically detects misinformation. For this reason, in 2017, they decided not to follow this route and used their skills and experience in the domain of regulation and compliance. Around 2017, at the time when Mr Rocktaschel decided to become less involved and pursue his post doctorate studies at the University of Oxford, the Claimant says the main business focus which he and his colleagues decided to work on in 2017, was developing code that could read text and answer questions. The likely clients were envisaged to be professional companies.[20]The Claimant explained that the algorithms are the same for each application. The type of data collected (“the data set”) changes for each domain of application. The domain data set is trained to automatically generate the software.[21]Bloomsbury developed an artificial intelligence software called Cape which applied natural language processing to read text documents and answer questions about their contents. The Claimant says that Bloomsbury was not set up specifically to commercialise the Cape product; Cape was one potential, commercial application that came from their wider research and development work at Bloomsbury related to open-domain question answering and machine reading. He says that the technology could easily be adapted and applied to a different domain of data so that it could in fact be used for fact checking, as the question/answer process is very similar. While Bloomsbury was not set up to commercialise the Cape product, by mid 2017 that was in fact its main activity.[22]An article in the bundle titled “Introducing Cape” written in September 2017 by J Godwin, with input from Sebastian, explained what Bloomsbury were doing. It said that much of your organisation’s knowledge is communicated or stored in natural language, through documents on your shared drive or instant messaging. It went on to describe how natural language was hard for computers to understand. Bloomsbury’s mission was to solve this problem. The article said, “everybody should be able to ask any questions they like to whichever set of knowledge and get the answer instantaneously”. The article said “We've started off built by building an artificial intelligence that reads text documents and answers questions about their contents (called Cape). You can use Cape on your own documents or website, so that users of that knowledge can get immediate answers to their questions”. The article concluded with future plans for Cape explaining that it could already answer a lot of questions and was ready to use right now. They were impressed by how well it performs on documents like internal gifts and expenses policies . It then said for the future they were going to use the data from this version to improve the AI, so that it could answer more questions - even ones that require elements of reasoning and synthesis. Eventually they wanted to be able to answer 10.5 Reserved judgment with reasons – rule 62 5 March 2017 any question that requires reading better than a human. Ms Sirota referred to Cape as an enterprise tool. As she said, it was designed so that a user could ask questions of the text, even if those questions used different phrases to those in the text itself.[23]The Bloomsbury staff had developed into two teams, one of which I shall refer to as the Claimant’s team, which was focused around the Claimant and another founder, Luis. The Claimant’s team focused on developing and applying the research into a commercial project. The other team was focused around Sebastian and their emphasis was on research. Mr Rocktaschel was not part of Sebastian’s team as he only spent about 2 hours a week at Bloomsbury as an external consultant during which he supervised a couple of interns/students in their research work which was part of their Master’s degrees.[24]The Claimant says the Bloomsbury research team spent their time on literature review, data set development, running and analysing experiments and collaborating with the other staff at Bloomsbury. They had some ongoing involvement with universities. Sebastian was contractually required to spend 20 per cent of his time at UCL.[25]The Claimant says the staff working on the applied side, (that is to say his team) largely spent their time coding and DevOps, collaborating with the research team and interacting with end users or potential customers, which I take to include pitching to potential customers.[26]Together the teams collaborated on the specific single project called Cape. Their focus was research and development of a data set called ShARC and the Cape software and commercial application of Cape.[27]The Claimant said they used to sit round two different tables within the same office space. They socialised together. They did not have any specific management staff, so the Claimant, as CEO, had to spend some part of his time on management work. Decision to sell Bloomsbury[28]The Claimant described the background to the decision to try to sell Bloomsbury, explaining that in early 2018, although they had had a lot of success in developing Cape, it had not turned into commercial deals that were developing income. The funding they had secured from various investors was running out and there were limited options. One of those was to sell the business and that is the route the Claimant and his colleagues chose to follow. Mr Rocktaschel provided some more detail about this explaining that another factor was that Sebastian had taken a one year sabbatical from UCL which allowed him to spend 80% of his time on Bloomsbury and 20% at UCL. This sabbatical was coming to an end and he would have to switch his time focus so that he spent 80% of his time with UCL and 20% on Bloomsbury.[29]Bloomsbury was in touch with several large technology entities, all of whom were interested in developing solutions for misinformation. There was interest in how his team and their research and development work could do this. There was particular interest in acquiring this group of individuals who were highly skilled and experienced in dealing with natural language processing and artificial intelligence. 10.5 Reserved judgment with reasons – rule 62 6 March 2017 Negotiations with the Respondent[30]It is not necessary for me to go into great detail about the negotiations or the transaction which eventually took place. I refer to it as a transaction even though it was comprised of a series of steps including a Release and Waiver agreement and various employment contracts and a restricted stock agreements. It is however relevant to examine the various documents which described what Bloomsbury was doing and those that described the part of the Respondent’s operation that the Claimant and his colleagues joined.[31]The negotiations started around 20 April 2018. An email dated Monday 23 April 2018 sent by Sebastian to Jerome of the Respondent group, referred to a meeting they had on the previous Friday and said: “Great talking to you on Friday. I really think there are a lot of synergies here! As promised, here is an overview of our team. I have divided it into people I see at FAIR (research scientists), and those I see in Applied ML (Applied Scientists and Research/ML engineers). You asked me to be honest, which I was. We hired this team with great care, partially through the UCL pipeline (which is amazing). I really love the team and want to continue working with them. That's obvious for the research scientist but holds true for the engineers as well. For example, I'd love feeding our cutting edge reading & reasoning research into a fact checking engine our engineers would help to develop.”[32]In May 2018 the Claimant, Sebastian and Tim went to Menlo Park which I understand is the Californian headquarters for the Respondent group. In preparation for that a note was prepared by Sebastian which discussed who from the Respondent group would be involved and their agenda. It was amended and commented on by others including the Claimant. Sebastian set out his impression of the aspects that made the possibility of acquiring Bloomsbury interesting for FB, stating there was a strong desire to set up a FAIR lab in London and strengthen their profile here, a strong desire to tackle misinformation(etc)] across the organisation. They had a strong interest in NLP, they were setting up SAIL and hiring Bloomsbury in this context would be a strong signal. He referred to Bloomsbury having several strong researchers, but also drive and vision to go beyond research. One notable comment made by Sebastian, under the heading “what could make this deal still difficult” was “Bloomsbury isn't really addressing misinformation”.[33]The note detailed what they should be pitching to the Respondent group as well as what they would like to do, what they were building and developing and where they were successful. As this was a pitch document, it aimed to present the most attractive picture and one they thought most relevant to the Respondent group, saying: “We should have a strong and coherent story to tell about [Bloomsbury], its successes, vision and possible role with FB.” in relation to that Bloomsbury story, he referred to Bloomsbury's vision of reading/reasoning/conversation and how it was part of the project and part of the research. He referred to an interest in real world impact, not only papers and to strong teamwork and believing in the team. He also listed strong research background and strong software engineering background, as well as very strong 10.5 Reserved judgment with reasons – rule 62 7 March 2017 overlap with the goals of SAIL. What he said they wanted to do was teach machines to read, reason and converse to enable everyone to share expertise at scale. What they said they were building and developing was:• “Cape, a product/engine that understands shared expertise to answer questions and test claims• a world class team of leading researchers and engineers (and a culture…)• Knowledge/research to make fundamental progress in this direction”.[34]When they went to meet the Respondent group at Menlo Park, they delivered a presentation which described Bloomsbury and the Cape project. Under the heading “Progress” a description of Bloomsbury explained that it had a seed round of £1.4 million in April 2017. A series of bullet points briefly explained that Bloomsbury had built a world-leading question answering system for unstructured text, across many domains (law, travel, websites, regulations) and tested it extensively. They had developed a scalable version of the algorithms by distributing the heavy computational needs. It went on in further bullet points to explain they had built a team of AI/NLP experts, world leading in machine comprehension, filed 2 patents on conversational machine reading, engaged the conversation with up to 20 commercial partners and designed a framework that enables the system to understand regulatory questions on demand. The presentation concluded with a summary describing the vision of the Bloomsbury which was software that can be taught to read, reason and communicate stating they were a world leader in this domain. Then under the heading “Applications” it said Cape is up and running and solves real problems. Cape can help Facebook in some of its key challenges. It concluded “We are ready for the next phase!”. Around the same time the Claimant and his colleagues at Bloomsbury also pitched their skills to Twitter. A similar presentation summarised their operation stating at the end “production ready, state of the art software Cape1.0 is uniquely positioned in NLP, with enhancements for Cape 2.0 already being tested and proving to outsmart its peers”. In fact, Cape 2.00 was largely at a conceptual stage.[35]The negotiation process included meetings between the various individuals and key figures within the Respondent group. The notes of those meetings in the bundle are headed “Interview Feedback” and include assessments of their technical competence which suggest the interviewers had greater know how than some of the interviewees. They also assessed their attitude to a number of different situations. None of the interviews suggest the Respondent was looking to maintain any part of the Bloomsbury operation. Rather they are entirely consistent with an interview for individual employees. Transaction[36]All of the evidence confirms the Respondent’s assertion that it did not wish to acquire Bloomsbury. It did wish to acquire the Claimant, Sebastian and Tim Rocktaschel as well as the significant expertise they had. The Respondent did not know that Tim Rocktaschel was not an employee of Bloomsbury. To achieve this, the Respondent paid a significant sum of money to Bloomsbury and entered into the Release and Waiver agreement which was signed on 26 June 2018. This agreement was between a Respondent group company and Bloomsbury and provided for the payment of the sum of money (“the Consideration”) which was to be paid within 5 days of the three key employees identified above, joining the Respondent. The money enabled the Claimant and his colleagues to repay their 10.5 Reserved judgment with reasons – rule 62 8 March 2017 investors. The Claimant and his colleagues entered into employment contracts with the Respondent and were also granted restricted stock. They were released from Bloomsbury, so that they could join the Respondent as employees. After that, the Respondent group interviewed the remaining staff from Bloomsbury of whom two did not meet standards and one chose to work elsewhere. The others were offered employment and there was a general aim that they should all start on the same date which was eventually agreed as 20 August 2018. Additionally, Bloomsbury had offered an internship at a future date to an individual who was offered an internship with the Respondent in its place.[37]The Respondent made it clear to the Claimant that it did not wish to have any of the Bloomsbury software. The Claimant and his colleagues arranged for Bloomsbury to be wound up and for the software they had developed to be made open source. The effect of it becoming open source was that anybody could use it without violating any intellectual property rights and this enabled the Claimant and his colleagues to access it if they chose. The Respondent told the Claimant that it was quicker and simpler to develop from scratch any software that might be required. I understand that this was because any software would have to work within the Respondent group’s own systems and modifying existing software was more complicated then starting from the beginning. It also avoided the risk of later claims for intellectual property infringement.[38]There was no instruction from the Respondent to make the software open source, but it is clear that the Respondent was aware from its previous experience that this route had some advantages. The Claimant and his colleagues were clearly made aware of that option somehow. However, they elected to do the windup and open source process of their own volition and in a time scale they chose.[39]The only other tangible assets which Bloomsbury had was a rented office and some computer equipment. None of those assets were passed across to the Respondent as a result of its hiring the Bloomsbury people. In practice Bloomsbury was wound up, the office space given up and the computer equipment most likely sold. FAIR and SAIL[40]At the time the Claimant began work for the Respondent he says there were around 1000 employees working for the Respondent group in London, including around 100 working on artificial intelligence although there may have been some other engineering teams also working on the AI systems. Within the AI branch at the Respondent’s group, there were two particular groups, one of which was called FAIR and the other called SAIL.[41]The Claimant said the FAIR group was primarily focused on academia and pure scientific research into artificial intelligence while SAIL was focused on developing practical applications for artificial intelligence. The Claimant also said that the FAIR group did exist internationally but had no presence in London before the Claimant joined the Respondent. Ms Sirota explained that FAIR is an acronym for Fundamental AI Research Lab. It had been established for some time. It’s role is to conduct open AI research to create the technology that powers the Respondent apps, but also looks to promote the pioneering usage and development of AI. The documents show its mission is to advance the state of the art in artificial intelligence through open research for the benefit of all. 10.5 Reserved judgment with reasons – rule 62 9 March 2017[42]The SAIL group was formed in April 2018, which was shortly before the discussions with Bloomsbury started. An email dated 13 April 2018, i.e., before the negotiations between the Bloomsbury founders and the Respondent group, from Jerome Pesenti to a group of staff within the Respondent group announced the formation of SAIL. SAIL stood for Society and AI Lab. It was a new group formed within the AI organisation.[43]When SAIL was formed the announcement issued by the Respondent group said: “Society and AI Lab (SAIL). I am excited to announce the formation of a new group within our AI organisation … focused on leveraging the power of AI for the good of society, and on making sure we are doing this responsibly and ethically. Joaquin will lead SAIL and take on the challenge of building up the technical and research capabilities we need in his space. SAIL will start by addressing three main areas, algorithmic fairness and transparency, collective behaviours with a focus on misinformation and polarisation, and well-being with an initial focus on vertical social mobility”.[44]SAIL started in April 2018 with four people. Joaquin Quinonero, (“JQC”) was to lead SAIL. That announcement was followed by a further, more detailed explanation by JQC about the work that SAIL was to do. His announcement said that SAIL will by design work in tight collaboration with AML and with FAIR. A number of Fair researchers had been doing pioneering research in the ethics of AI, and AML had built an integrity solutions team that was tackling issues such as misinformation. Another team he identified that they were going to collaborate tightly with was Core Data Science. Additionally, the email explained they would continue to team up with product teams across the company. They would continue partnerships with people in news feed integrity, privacy policy, product support operations, jobs, civic engagement and community integrity. The announcement finished by stating how to engage and pointed out they were hiring. It seems this triggered the Respondent’s negotiations with Bloomsbury. Joining the Respondent[45]On 3 July 2018, Jerome made an announcement internally about the Bloomsbury team. It said: “I'm excited to announce that the team behind Bloomsbury AI has agreed to join Facebook in London. They've built a leading expertise in machine reading and understanding unstructured documents in natural language in order to answer any question. They have a strong fundamental research culture illustrated by world-class publications. The team including founders [the Claimant] and [Sebastian] will join SAIL and FAIR respectively, and be based in London. The FAIR part of the team will report into [AB]. The SAIL part of the team will report into [JQC]. They will continue working together and be initially dedicated to News Feed integrity issues. Their expertise will help us improve automatic content understanding as we continue our work reducing misinformation and fighting hate speech.” 10.5 Reserved judgment with reasons – rule 62 10 March 2017[46]On 20 August 2018, the Claimant and Sebastian as well as most of the other employees who were going to work with the Respondent started work.[47]The Claimant argues that by acquiring a team of people with expertise who were used to working together, the Respondent acquired a part of the business for the purposes of TUPE. The Respondent did acquire most of the staff at from Bloomsbury. In addition, it acquired Tim Rocktaschel who was not an employee of Bloomsbury, having already reduced his shareholding to a nominal amount and ceased to be an employee of Bloomsbury in 2017. As noted previously, Mr Rocktaschel had been a self-employed consultant with Bloomsbury. He had been doing up to 2 hours a week supervising some interns/students for Bloomsbury.[48]Ms Sirota, who was a director, corporate development, for the Respondent group and had led the deal from the Respondent’s point of view explained that the interest in Bloomsbury was driven by the high quality of the expertise that some of the Bloomsbury people had in artificial intelligence. The advantage of acquiring several people together was that they would “seed” the London office. In practice, it seems that by having a few very highly expert employees, the Respondent group was able to attract other highly skilled people to join their London operation.[49]The July announcement about the Claimant and his colleagues joining the Respondent, which I have referred to previously, said about the work they were joining to do: “They will continue working together and be initially dedicated to News Feed integrity issues. Their expertise will help us improve automatic content understanding as we continue our work reducing misinformation and fighting hate speech”.[50]The documents confirm the Respondent’s assertion that the purpose of acquiring the Claimant and his colleagues as a team was to “seed both FAIR and SAIL in London”. There were a couple of people in London called Ben and Andrea who were working in the wider FAIR team, but they were working with the Paris team and reported into another manager. They were not seen as acting as a local team. Working at the Respondent[51]When the Claimant joined the Respondent, he says that he worked as the London head of the SAIL team. He was joined by one of the other founders, Luis, who had worked with him previously on the more applied matters, and another employee, MS, who had primarily worked with Sebastian on research but worked with the Claimant from time to time.[52]The other former Bloomsbury people who joined the FAIR group were: Sebastian who went into the research team in FAIR, Tim Rocktaschel who worked on research within FAIR, and PL who went to the FAIR team as a PhD student. As I have noted, three other employees did not join the Respondent, one choosing to take another offer and two being deemed not up to the Respondent group’s hiring standards at the applicable time.[53]Four of the five employees joined the Respondent on the same day, 20 August 2018. Tim Rocktaschel, who I have noted was not an employee of Bloomsbury before the alleged transfer was on paternity leave for a few months at the beginning and PL had to wait until he started his PhD at UCL before he could 10.5 Reserved judgment with reasons – rule 62 11 March 2017 start with the Respondent so, in practice, they both joined slightly later than the rest.[54]Immediately on joining the Respondent group, the Claimant and his colleagues had a period which was referred to as “Bootcamp” at which they were introduced to the Respondent group and to the computer code and products they would be working with. The Bootcamp took about five weeks for the SAIL team who had to go to the Respondent group headquarters in California for most of this induction, and then stayed for another week or two for the AI Offsite, but about two weeks for the FAIR team as they were not required to be as familiar with the internal code of the Respondent and did not need to go to the States, but possibly did go to Paris where other FAIR team staff were based. Immediately after the Bootcamp process, the Bloomsbury people were joined by two internal Respondent employees in each team after the Bootcamp process.[55]An announcement issued on Monday 20 August from AB (who was the line manager for the FAIR team) to other staff within the Respondent group and asked them to help on board the team from Bloomsbury AI. It said, “The team is joining with deep NLP expertise and will help accelerate work related to misinformation, polarization, and harm in conversations”.[56]After the Bootcamp induction, the Claimant started working in London, which by now would have been about 8 October 2018. The Claimant says that his work arrangements continued largely as before. He describes the similarities between the position before and after the move to the Respondent. I have noted the Claimant’s comments carefully. In terms of the work, the Claimant says that the Bloomsbury research team who joined FAIR carried on doing literature review, data set development, running and analysing experiments and collaborating with the SAIL team on specific projects.[57]The Claimant says in SAIL he and his colleagues continued coding and DevOps and collaborating with FAIR on specific projects and other teams and interacting with and users.[58]The Claimant says his team and the FAIR research team continued collaborating together on specific projects whereas previously they collaborated on the specific single project called Cape. He says that the work they carried out previously had been the research and development of a data set called ShARC and the Cape software and commercial application of Cape whereas now they were working on research and development of two different data sets which were required for the Respondent, but they did so, based on the know how used in developing the ShARC data set and using their expertise of developing the Cape software and applications. After moving to join the Respondent, the Claimant says they were developing software on two different data sets called LAMA and MLQA, based on know how they had used in developing the ShARC dataset at Bloomsbury. He also says they were using their expertise of developing the Cape software and applications to develop software to solve issues with Facebook products including fake account detection, corroboration and claim similarity as initial projects, which were all based on machine learning expertise gained during their time at Bloomsbury.[59]The Claimant also argues that at Bloomsbury some of them had had part time roles at universities and Tim Rocktaschel had a fixed term contract at Oxford. An 10.5 Reserved judgment with reasons – rule 62 12 March 2017 involvement with academic institutions continued, as Sebastian and Tim carried on with part time roles at UCL and another colleague continued doing a PhD at UCL.[60]In the past, Bloomsbury had had made a number of pitches to potential clients to promote Cape, and had some interest, but had no sales. Within the Respondent the London SAIL team, the Claimant says they anticipated that identifying problems to solve for other teams was key. In essence, they tried to identify problems they could solve for other teams in the Respondent group, and they were effectively pitching ideas to product teams within the Respondent group who were working on particular problems, such as the misinformation team.[61]The Claimant argues that they continued to operate in teams on a similar basis to the way they had at Bloomsbury. They used to sit round two different tables within the same office space. Once they moved to the Respondent the desks were organised into two groups and then into two adjacent tables or groups of tables. They continued to socialise as teams incorporating other members as the teams grew. The Claimant on questioning said that he collaborated with the product teams which he accepts were international and he collaborated with Sebastian over research issues, having one meeting a week with Sebastian.[62]The Claimant summarised the position saying that Bloomsbury was essentially a research and development company looking to solve problems using artificial intelligence and machine learning. After the Claimant joined the Respondent with his colleagues, he said: “we were still a London based team of researchers and developers working to solve problems using artificial intelligence and machine learning, but we pivoted away from looking for potential customers to focusing on and pitching to internal customers in respect of Facebook products. The teams have developed and grown over time, but immediately after we joined (and had completed the onboarding bootcamp) there was essentially very little difference between the usual day at Bloomsbury and a usual day at Facebook in terms of our work tasks and output, it is just that we were doing this for Facebook rather than for potential external clients.”[63]The Claimant’s description of how he and his colleagues operated within the Respondent is focused on London, but in practice it is clear that while Respondent is a UK company, it is part of what I have referred to as the Respondent group. The Respondent group is a multinational operation and as a digital technology company, people work with others across borders in teams which are multinational in themselves.[64]The initial weeks were spent trying to find a role. The position was slightly different for those in FAIR and those in SAIL as reflected in the different bootcamp experiences they had. The London SAIL team worked to find opportunities within other groups where they could collaborate on the key issues that were within the scope of the SAIL mandate described above. Importantly, other than working with in collaboration with other groups on matters within the SAIL mandate, there is no evidence that Respondent made any specific determination of what it wanted the former Bloomsbury staff to do. It is clear that they were not given any specific tasks but were expected to liaise with other teams and work to find synergies. The Bootcamp was clearly important as a way to introduce the SAIL team to the various areas the Respondent group was working on and enabling them to make contacts as well as understanding the coding and systems. 10.5 Reserved judgment with reasons – rule 62 13 March 2017[65]The Respondent group operates in a manner under which the pure research scientists in FAIR were free to choose the research area they wanted to follow. They were not directed as to what project they had to work on. In short, they were able to join the Respondent, be highly paid and research their own specific interests. Clearly, they were only hired because their interests were consistent with the range of work that the Respondent was interested in developing within FAIR, but they had enormous freedom in what they did. In consequence, Mr Rocktaschel, who had previously been in academia and expected his career to continue that way, joined the Respondent because of that freedom in his relation to his research and although he is a natural language processing expert, he decided to pursue the study of reinforcement learning which had been the focus of his previous role at the University of Oxford. Sebastian did continue to cooperate with the Claimant, and he did continue his natural language processing research, but none of the FAIR team were obliged to do specific work.[66]Just a week or so after their SAIL team returned to London, on 16 October there is an exchange of messages between various team members about a presentation they were giving which I understand to be the document entitled “Machine Reading, FAIR/SAIL London” which was presented by the Claimant, Sebastian and Luis with collaboration from MS, Mr Rocktaschel, PL and two others. It was partially based on old Cape presentations and included some Cape slides.[67]The presentation started by outlining AI at the Respondent group’s operation in London referencing three teams being the FAIR team, AML, and SAIL. According to the presentation, FAIR consisted in London of Sebastian, Tim Rocktaschel, Andrea and Ben and their focus was on academic excellence and open source. AML London focused on language, vision/video/speech, personalization and integrity. SAIL in London consisted of the Claimant, Luis, MS and two others and focused on Trust in AI (Bias, Fairness), Wellbeing and Collective Behaviour. It concluded with a slide addressing ongoing projects which included Fake Account Detection, Critical Events Detection such as child abuse, terrorist threats and low resource languages and Provenance Graph referencing integrity which included more efficient reviews/fact checking, chain of trust and polarisation indicators. That aspect also included better experience which involved Video Chaining, Exploration Tools and Improved Search. SAIL London move to AML[68]A few days later, on 19 October 2018, an announcement was made about SAIL London transitioning to join AML in the Integrity Solutions team. The announcement made it clear that the only change in practice was a reporting line from the Claimant to another line manager. The explanation for the change was that they had landed on high impact integrity projects for SAIL London. Specifically, it said the Claimant “and team are partnering closely with FAIR and the Harmful Behavior/Actor teams within Community Integrity in London, to increase the quality and speed of their integrity classifiers via structured representations of shared content, model robustness, behavioral and conversational analysis, and natural language explanations. This work is very well aligned with the integrity work taking place in AML integrity solutions. As a Facebook AI organisation, this is going to help us be more efficient in our work on integrity.” I include this because it describes what the SAIL London team were 10.5 Reserved judgment with reasons – rule 62 14 March 2017 doing. The reporting line change is not material as there was no change in their work. Contemporaneous Respondent comments on how the Bloomsbury staff integrated[69]There are various other documents which describe the work which the Bloomsbury staff were doing for the Respondent group. On 17 May 2019 an email between various people within the Respondent group made a comment in an email headed “Bloomsbury acquisition - 6 months check in” and said, “In general, this deal is exceeding expectations as it achieved the strategic rationale of seeding the London AI/FAIR efforts (talent performing well, accelerating growth efforts in London, impactful to misinformation integrity problems).”[70]The email carried on with a summary which reported: the “strategic thesis is still true for FB. The only change is that [the Claimant] and M are now under Patrick in AML (moved from SAIL). The SAIL team doubled down on fairness and the AML vision/mission aligned better with the work this team was doing. Overall the deal is exceeding expectations. By acquiring this team, we seeded the first Facebook AI presence in London and have managed to grow both AI and Applied Research there. By bringing in talent like Sebastian, we now are getting more inbound requests from companies like DeepMind and are launching an AI Resident program. Team is already hitting 12 month headcount/hiring plans for London”.[71]The review summary referred to the team’s contributions to meaningful improvements in protecting children from groomers, new models on improving URL hoaxes, and improvements in identifying content used out of context. A question in the summary asked what value did the Respondent get from the deal and the answer was “Strong talent. Ability to seed the London AI presence and grow the site. The acquisition also brought much needed growth in integrity talent at FB”.[72]The Claimant’s performance review in February 2019 referred back to the expectations when the team joined and explained “I expected [the Claimant] to lead his team through this transition, to immerse himself and his team in the FaceBook’s infrastructure/culture/integrity problems, to grow a critical mass for his team, to establish London as a new site for AML, to develop an initial team vision, an investment plan for Q4 and 2019, and to develop partnerships with FAIR and CI Harmful Behaviour”.[73]The list of Team Impact items included: “Setup AML London team ..by hiring a strong team of highly motivated senior people.., Established concrete 2019H1 projects with partners CI team in Harmful Behaviour [The Claimant] has worked with various Integrity teams to come up with a working project plan for AML London. With the Harmful Behaviour team, he has come up with two concrete projects that are now being scheduled for 10.5 Reserved judgment with reasons – rule 62 15 March 2017 2019 H1 goaling: age detection (crucial to IIC/grooming work), and sequential models (as an extension of existing behavioural classifiers once E2EE comes into force)” -partner team”.[74]There were a number of 360 degree appraisal forms which have been collected from staff working with the Claimant which commented on his performance and some of these gave indications of the work that had been done. One of those submitted in January 2019 was from Luis, who said the Claimant was the principal driver in the formation of the new team, by which I understand he is referring to SAIL, London. He described how the Claimant talked to other FB teams to find the possible objectives and direction for the new team. Luis said the Claimant also organised and ran the meetings to discuss and plan the milestones and objectives of the team. Another one written in late December 2018 described the Claimant’s work saying he ”has worked with various Integrity teams to come up with a working project plan for AML London. With the Harmful Behaviour team, he has come up with two concrete projects that are now being scheduled for 2019 H1 goaling: age detection (crucial to IIC/Grooming work), and sequential models (as an extension of existing behavioural classifiers once E2EE comes into force).[75]Bearing in mind expectations identified by the Claimant’s manager Patrick Pantel in the performance review cited above, (which were to immerse himself and his team in the FaceBook’s infrastructure/culture/integrity problems, to grow a critical mass for his team, to establish London as a new site for AML, to develop an initial team vision, an investment plan for Q4 and 2019, and to develop partnerships with FAIR and CI Harmful Behaviour), as well as the freedom allowed to FAIR researchers, it is clear that the Respondent had no specific objectives for either team. The expectation was only that the former Bloomsbury staff would find for themselves, a way to collaborate with other project groups to enhance the overall objectives of the Respondent’s international operation, within the mission of either the FAIR or the SAIL teams to which they had been assigned. As far as the SAIL team were concerned, the Claimant led that process. The individual research scientists, Sebastian and Mr Rocktaschel led their own process. Submissions Respondent’s submissions[76]The Respondent submitted that the onus of establishing that there was a Transfer is on the Claimant - Lom Management Limited v Sweeney UK EATS /0058/11/B1 at para 17.[77]I was reminded that the Claimant relies on Regulation 3(1)a, and therefore had to establish there was: a transfer of an undertaking, business or part of an undertaking or business situated immediately before the transfer in United Kingdom to another person where there was a transfer of an economic identity which retains its identity.[78]The Respondent pointed out that the task for the Tribunal is to consider whether there was an economic entity which retains its identity. I was referred to the definition of this in Suzen v Zehnacker Gebaudereinigung Gmbh Krankenhausservice [1997] ICR 662 ECJ at para 13 and the case of Spijkers v Gebroeders Benedik Abattoir [1986] 2 CMLR 296. I was also referred to the case of CLECE SA v Marin Valor ECJ [2011] ICR at para 34. The case law set out the 10.5 Reserved judgment with reasons – rule 62 16 March 2017 kinds of matters to be considered being whether or not its tangible assets, such as buildings and movable property, are transferred, the value of its intangible assets at the time of the transfer, whether or not the majority of its employees are taken over by the new employer, whether or not its customers are transferred, the degree of similarity between the activities carried on before and after the transfer, and the period if any, for which those activities were suspended.[79]The Respondent conceded that in certain labour intensive sectors, a group of workers engaged in a joint activity on a permanent basis may constitute an economic entity. However, the Respondent submitted that while in some fields employees may be one of the major assets that may be transferred, for there to be a transfer it is necessary for the transferee to continue the transferor's business. The Respondent submitted that merely hiring employees that are not tied to a continuing organised business activity is not enough. The Respondent emphasised that if the transferee continues with similar, or even identical activities, this does not lead to the conclusion that an economic entity has been retained. The Respondent argued that the identity emerges from several indissociable factors, such as its workforce, its management staff, the way in which the work is organised, its operating methods or indeed, where appropriate, the operational resources available to it.[80]The Respondent submitted that no business activity transferred, that no clients or contracts transferred, no intellectual property transferred. The Respondent did not take over activities that were the same or similar to those at Bloomsbury in that no work was done on Cape. Additionally, the Respondent submitted that some of the staff were hired but only after a selective interview process as the Respondent was interested in hiring those with particular skills but not in continuing Bloomsbury's business activities. The Claimant’s submissions[81]The Claimant submitted that the agreement with the Respondent was structured to enable the Respondent to get a high functioning team that worked well together. The Claimant submitted that the label or mechanism was unimportant. What counted was the objective or outcome. The outcome of this arrangement was that the majority of the Bloomsbury AI team, which was the defining asset of the of Bloomsbury, moved to the Respondent and continued as research and development teams making up FAIR London and SAIL London. They continued to work in much the same way as they had done at Bloomsbury and while the SAIL London team later moved into the AML team, the only practical difference was a change to the Claimant’s line manager. It was therefore the Claimant’s submission that this talent acquisition or what was sometimes called “acqui hire” achieves the same objectives as a traditional acquisition and was essentially a TUPE transfer by another name.[82]The Claimant submitted that there were four questions for the Tribunal to determine to establish whether there was a transfer being:(a) Was there an economic entity?(b) Was situated in the UK?(c) Did it transfer to another person?(d) Did it retain its identity after the transfer? 10.5 Reserved judgment with reasons – rule 62 17 March 2017[83]In relation to the first question of whether there was an economic activity, I was referred to the case of Cheesman and Others v R Brewer Contracts Limited [2001] IRLR 144 and the guidelines for determining that. The Claimant said it was self-evident that there was an economic entity before the transfer, and it was also clearly situated in the UK.[84]On the question of whether the economic entity transferred to another person, the Claimant referred to the case of Landsorganisationen I Denmark v Ny Molle Kro 1989 IRLR 37 and said that a TUPE transfer may take place whether or not any property is transferred between the transferee and transfer at all.[85]On the question of whether the economic entity retained its identity after the transfer, I was again referred to the case of Cheesman and to the principles which the EAT distilled from the authorities. I do not need to recite them all but some of them are worth particular attention. For example, it was pointed out that in a labour intensive sector it is to be recognised that an entity is capable of maintaining its identity after it has been transferred where the new employer does not merely pursue the activity in question but also takes over a major part, in terms of their numbers and skills, of the employees specifically assigned by his predecessors to that task.[86]While Cheesman recorded the various matters for consideration which the Respondent has referred to, it also pointed out that account has to be taken, inter alia, of the type of undertaking or business in issue and the degree of importance to be attached to the several criteria will necessarily vary according to the activity carried on. It also pointed out that where an economic entity is able to function without any significant tangible or intangible assets, the maintenance of its identity following the transaction being examined cannot logically depend on the transfer of such assets. Furthermore, it also said that even where assets are owned and are required to run the undertaking, the fact that they do not pass does not preclude a transfer.[87]Additionally, I was referred to the case of Farmer v Danzas (UK) Limited [1994] UKEAT/858/93 in which it was pointed out that it was sufficient that the transferred undertaking retained its identity immediately after the moment of transfer, whether or not the transferee intended to integrate the acquired entity into its own operation. What happens later does not alter the position. It was pointed out that every business was likely to wish to integrate where possible any new business which had been acquired so that there could be flexibility in work practices and economies of scale, but an economic entity did not cease to retain its identity merely because it was subsumed into the transferee’s business. In effect the snapshot is taken immediately after the transfer although in Cheesman another point made was that there was no particular point to be attached to a gap between the end of the work done by one subcontractor and its start by the successor.[88]The Claimant also submitted that an economic entity does not need to be operated as an organizationally autonomous part of the transferee’s business post transfer in order to meet the retention of identity tests. Klarenberg v Ferrotron Technologies GmbH [2009] IRLR301 ECJ at paras 40 to 48.[89]I was also referred to the case of Camden Primary Care Trust and University College London v Skittrall and others [2005] EAT OO78/05 in which it was held that alteration of the businesses’ location and alteration in the mode of provision of 10.5 Reserved judgment with reasons – rule 62 18 March 2017 the services will not necessarily prevent the undertaking retaining its identity following transfer. It was stressed that an economic entity should not be defined so narrowly as to prevent a TUPE transfer from ever occurring, especially given the purposive approach to the application of TUPE. The relevant factor is the transfer of the principal asset of the undertaking. While the mode of providing service or the model used may differ, that does not equate to the activity itself and does not, therefore, negate a relevant transfer where the nature of the activity remains the same. Additionally, in that case it was pointed out that a change of terms and conditions was not defining factor because a refusal by the putative transferee to honour the transposed terms and conditions of employment would, if it were taken into account, undermine the whole basis of the protection afforded to employees under the TUPE regulations.[90]The Claimant took each of the factors in Cheesman in turn and argued that Bloomsbury was a start up research and development tech company and its defining feature was its intellectual capital being its employees. Bloomsbury was not an asset reliant business. Its main asset was the expertise and know-how of its employees. The Bloomsbury Cape software and data sets were of no value, but Facebook still paid a very large sum to secure the deal on the team and it also secured the use of the assets without fear of reprisal. Five out of the eight employees were taken on by Facebook, as well as Mr Rocktaschel, who was a consultant and not an employee. The fact the Facebook did not employ two of the employees in breach of TUPE could not be used to undermine the application of TUPE. Bloomsbury did not have any customers at the time of the transfer and therefore this factor was irrelevant.[91]In terms of the degree of similarity between the activities carried on before and after the transfer, it was submitted that the Bloomsbury teams did not have to carry out exactly the same activities at the Respondent as they did at Bloomsbury in order for there to be relevant transfer. All they had to show as they carried out similar or analogous activities. The Claimant accepted that the projects and applications and data sets at Facebook were different but argued that the underlying work on natural language processing and machine learning was effectively the same or analogous to the work at Bloomsbury.[92]I was reminded that Facebook had recognised they would continue to work on misinformation problems at Facebook and their expertise would help Facebook’s AI team to improve automatic content and understanding. I was also reminded that the documents included a reference to Facebook saying the result of their world class research in machine reading, meant the Bloomsbury team was a critical piece of the puzzle for Facebook.[93]The Claimant argued that the work was organised in the same way as the work at Bloomsbury with Sebastian and the Claimant managing their various teams and continuing to work together. The overlap in their work could be seen from the presentation they gave the Facebook product teams in October 2018. At Facebook the Bloomsbury team continued in much the same way as they had done before.[94]I was also reminded that the FAIR and SAIL teams did not have a presence in London before the Bloomsbury team joined and therefore securing them strengthened their profile and sent a strong signal. There were references in the documents to a strong overlap between their work and that of FAIR and SALE at Facebook. Facebook was looking to build a new group in London and the 10.5 Reserved judgment with reasons – rule 62 19 March 2017 Bloomsbury team were to remain in London and seed the AI team in that office as well as other references to seeding FAIR/SAIL presence in London.The Law
[95]Regulation 3(1) a of the TUPE Regulations 2006 provides that the Regulations apply to: the transfer of an undertaking, business will part of an undertaking or business situated immediately before the transfer to the in the United Kingdom to another person where there is a transfer of an economic entity which retains its identity.[96]The case of Cheesman is the starting point for a tribunal when considering the question of whether an economic entity which retains its identity. The multifactorial approach requires the tribunal to consider a number of factors. The EAT held: “(i) As to whether there is an undertaking … an organised grouping of persons and assets enabling (or facilitating) the exercise of an economic activity which pursues a specific objective … (ii) … such an undertaking … must be sufficiently structured and autonomous but will not necessarily have significant assets, tangible or intangible; (iii) in certain sectors, such as cleaning and surveillance, the assets are often reduced to their most basic and the activities are essentially based on manpower; (iv) an organised grouping of wage-earners who are specifically and permanently assigned to a common task may, in the absence of other factors of production, amount to an economic entity; (v) an activity of itself is not an entity; the identity of an entity emerges from other factors, such as its workforce, management style, the way in which its work is organised, its operating methods and, where appropriate, the operational resources available to it.''[97]The following factors were highlighted by the EAT in Cheesman: ''(i)… the decisive criteria for establishing the existence of a transfer is whether the entity in question retains its identity, as indicated … by the fact that its operation is actually continued or resumed; … (iii) in considering whether the conditions for … a transfer are met, it is necessary to consider all the factors characterising the transaction in question, but each as a single factor and none is to be considered in isolation; (iv) amongst the matters … for consideration, are the type of undertaking, whether or not its tangible assets are transferred, the value of its intangible assets at the time of transfer, whether or not the majority of its employees are taken over by the new company, whether or not its customers are transferred, 10.5 Reserved judgment with reasons – rule 62 20 March 2017 the degree of similarity between the activities carried on before and after the transfer, and the period, if any, in which they are suspended; (v) account has to be taken … of the type of undertaking or business in issue, and the degree of importance to be attached to the several criteria will necessarily vary according to the activity carried on; (vi) where an economic entity is able to function without any significant tangible or intangible assets, the maintenance of its identity following the transaction … cannot logically depend on the transfer of such assets; (vii) even where the assets are owned and are required to run the undertaking, the fact that they do not pass does not preclude a transfer; … (x) the absence of any contractual link between the transferor and transferee may be evidence that there has been no relevant transfer, but it is certainly not conclusive as there is no need for any direct contractual relationship; (xi) when no employees are transferred, the reasons why that is the case can be relevant as to whether or not there was a transfer. 72 All the cases I was referred to by the parties were relevant and I have considered them all. The case of Camden Primary Care Trust and University College London v Skittrall and others EAT 0078/05 made several points. I draw from that two key points. Importantly, the scope of the economic entity must not be narrowly drawn. A purposive approach to the application of TUPE, which is designed to protect employees faced with a change of employer, is necessary. Secondly, it is important to take into account all the potentially relevant factors.Conclusions
[99]In order to determine this matter I have to follow the guidance in Cheesman. I first had to consider whether Bloomsbury was an economic entity. I do not think there is any dispute that Bloomsbury was an economic entity based in the UK.[100]I have to consider what sort of entity Bloomsbury was. Bloomsbury was a labour intensive operation. The Claimant’s Counsel argued that Bloomsbury was a start up research and development tech company and its defining feature was its intellectual capital being its employees. I do not agree with that description of Bloomsbury. It had been a start-up research and development company and it is correct to say a major feature was the employees’ high level of technical knowhow and experience. However, Bloomsbury had focussed on a product line and approach to machine reading and learning, with the enterprise tool, called Cape, and data set called ShARC. The type of the undertaking was one dependent on the skill and technical knowledge of and expertise of the Bloomsbury employees. They had a very high degree of expertise which they applied to producing their product, Cape, which they actively tried to market.[101]The Respondent took over a major part, in terms of numbers and skills of the employees assigned by Bloomsbury to both the work of developing their product and researching. How that happened is immaterial. I have to consider whether this was a TUPE transfer. 10.5 Reserved judgment with reasons – rule 62 21 March 2017[102]As Cheesman noted, the decisive criteria for establishing the existence of a TUPE transfer is whether the entity in question retains its identity, as indicated by the fact that its operation is actually continued or resumed.[103]In order to assess whether the entity retained its identity, it is important to look at a series of elements of the transaction identified by Cheesman. I looked at whether tangible assets were transferred. The only tangible assets that Bloomsbury had, were rented offices and some computer equipment. The rented offices were given up and the computer equipment was probably sold and certainly did not transfer.[104]I had to consider the value of its intangible assets at the time of transfer. The intellectual property assets of the Bloomsbury were the product Cape and the ShARC data set. I have little information about the value of the assets. I would imagine that at some point Cape had some potential value, but in the light of the opportunity at the Respondent, the founders of Bloomsbury were happy to negotiate with their investors to close it down and make that open source, so that it had no value.[105]There was no customer goodwill as such. Bloomsbury had no actual customers, so not only was there no goodwill transferred and no customers to transfer.[106]As I have noted, the majority of Bloomsbury’s employees were taken over by the Respondent. Of those that did not one chose another opportunity which, if TUPE were to have applied, is tantamount to objecting. The other two were rejected by the Respondent and had that had this been a TUPE transfer, that would have been an unfair dismissal, so that for these purposes the fact they did not transfer is immaterial.[107]The employees were potentially a significant asset in two ways. First, they already functioned well together as a team and secondly, the senior staff were highly respected in their field of AI. This created a sort of goodwill in that it enabled them to attract other very experienced and highly qualified individuals who would want to work with them. However, the consideration paid by the Respondent group under the Release and Waiver agreement was payable once the key founders being the Claimant, Sebastian and Tim Rocktaschel joined them. The Respondent group was not aware that Mr Rocktaschel was not an employee of Bloomsbury. The consideration would have been paid regardless of whether other Bloomsbury employees joined or not. Therefore, the value of the full team was negligible. The value was attributable to the key founders and that had a sort of intangible value represented by the consideration which was several million dollars.[108]The next Cheesman question is the fundamental question. That is what was the degree of similarity between the activities carried on before and after the transfer.[109]In this case, the only real asset which transferred was the employees. The transfer of employees alone in this situation could be enough to amount to a TUPE transfer if the activity they carried out was sufficiently similar before and after the transaction. The key question is whether I should take a broad view of the activity carried on before and after the transaction, as I was urged to do by the Claimant, 10.5 Reserved judgment with reasons – rule 62 22 March 2017 or whether I should take a more specific view of what its activity was. Essentially that is the dispute between the parties.[110]The Claimant urges me to view Bloomsbury as a research and development tech company which provided special skills in terms of AI; almost an AI consultancy. The Respondent urged me to look at this as a group of people who had researched specific AI and had developed a product (Cape) which they were aiming to market.[111]I was reminded of the pitfalls identified in the Camden case by the Claimant’s Counsel. I recognise the situation that the Camden case seeks to avoid. There is clearly a risk that if you focus too specifically on the detail, you can miss the broader picture. However, there is a point where you cannot describe the activity too loosely either. It is a little like the story of the blind men describing an elephant. If you get too close, you see only part of the picture. On the other hand, if you look from too far away, it becomes simply a four legged animal. The case law, particularly the Cheesman and Camden cases, attempt to sign post how to identify the entity in a purposive, sensible and pragmatic way.[112]One of the Cheesman factors is the question of any gap between the previous activities and the new activities. For the purpose of this decision, I consider that to look at the Bootcamp time would distort the picture. It seems to me that the question of the gap is precisely intended to ensure that in considering an issue of this nature you look at the activities when the entity was functioning as it normally does before the transaction and the activities when it is functioning after the transaction. In this case any gap has no impact on the overall position. Similarly, the fact that the employees in SAIL transferred to a different reporting line a short time later is of no practical significance. The usefulness of the contemporaneous documents that describe this move is their description of FAIR and SAIL’s activities.[113]I do not think the activities of Bloomsbury can be described as broadly as the Claimant seeks to do. Additionally, I cannot attach the same degree of weight as the Claimant suggests, to the fact that his daily routine was not much different once he moved to the Respondent.[114]At Bloomsbury, the team of employees who were all highly skilled in artificial intelligence and machine learning had applied their knowledge into developing a product called Cape. The Claimant was managing a team whose main function prior to the transaction was to market that product and continue to develop code. As the Claimant said, “our main business focus was on developing code that could read text and answer questions”. The team decided to develop artificial intelligence software which they called Cape which applied natural language processing to read text documents and answered questions about their contents. Specifically, the commercial application that they had worked on was question/answer machine reading for professional companies in the regulatory and compliance area. As part of that they developed a specific data set called ShARC. Some staff had a focus on research, but that research was around issues which were relevant to developing Cape and ShARC.[115]When the Claimant and his colleagues at Bloomsbury joined the Respondent, there was no intention to continue to develop Cape or ShARC. The team expected to contribute towards the Respondent group’s work in developing products in the AI field and hoping to continue working together. The transaction with the 10.5 Reserved judgment with reasons – rule 62 23 March 2017 Respondent enabled them to release themselves from Bloomsbury leaving the investors in a satisfactory situation. In reaching this conclusion, I have considered the fact that the Claimant and his colleagues continued to reference Cape when they were trying to interest the Respondent group in acquiring them and they also referenced it when they were explaining their services as FAIR and SAIL London to other groups within the Respondent group’s organisation in October 2018. However, it is my finding that they did so to demonstrate their background and experience and not because they were continuing to work on it. In practice they worked on entirely new areas, particularly around Integrity and Harmful Behaviours.[116]It is critical to identify what the teams did once they got to the Respondent. The Claimant says they continued to work in two teams. It is correct to say they worked in two of the three teams in the Respondent’s London AI section. The Bloomsbury part of the FAIR group consisted only of Sebastian (and PL once he started his PhD, which the Respondent had agreed to fund in place of Bloomsbury). It seems that at least one of the two people who were rejected might have gone to FAIR if they had joined the Respondent. The third person who joined FAIR, was Tim Rocktaschel, who was technically not a former employee of Bloomsbury, his connection being that he was a part time consultant. Moreover, he worked within FAIR on a different form of research. I accept Tim Rocktaschel’s evidence that FAIR was an established international group before they joined it, although it needed a greater presence in London and that he personally did not collaborate with SAIL after joining the Respondent. Sebastian did collaborate with the SAIL team to some extent but there is no evidence this was his focus and he was free to work on the research he wanted.[117]The Bloomsbury part of the SAIL team, consisting of the Claimant, Luis and MS did constitute the core of SAIL in London and SAIL was a relatively new operation within the Respondent group. The Claimant managed that team, as he had done the more applied or engineering work within Bloomsbury. He worked with Luis who had previously been Chief Technical Officer at Bloomsbury working with him. MS joined them, but MS had primarily worked with Sebastian previously, although occasionally with the Claimant. PL had previously worked with both Sebastian and Luis, but now worked with Sebastian. As I have noted immediately they finished the Bootcamp they were joined by another of the Respondent’s employees.[118]The similarity of the teams is raised because the Claimant argues that the existence of a group which operated together before and after the transaction is relevant in identifying to what extent they were essentially the same economic entity.[119]I accept the initial SAIL team in particular, was largely similar to the Bloomsbury team which had worked with the Claimant and Luis previously. The SAIL team spent the initial period looking for projects to collaborate on with other Facebook project groups, settling on high impact integrity problems which led to them moving their reporting line to join the AML Integrity Solutions team. Their aim was to use their specialist AI knowhow to address certain problems that AML Integrity Solutions team was dealing with. The memo dated 19 October 2018 about the SAIL move to AML Integrity Solutions describes the work they had “landed” on as partnering closely with FAIR and the Harmful Behavior/Actor teams within Community Integrity in London, to increase the quality and speed of their integrity classifiers via structured representations of shared content, model 10.5 Reserved judgment with reasons – rule 62 24 March 2017 robustness, behavioral and conversational analysis, and natural language explanations.[120]I accept that the Bloomsbury team’s knowhow was adaptable and so it could be adapted to other situations such as fact checking and misinformation and that they were able to contribute to the work described above because of the expertise they had developed during their work on Cape. However, prior to the transaction, Cape was not adapted for fact checking and misinformation, nor were efforts being made to do so. Bloomsbury had spent a year concentrating on one product which was the enterprise tool called Cape. Moving to work on different products and different coding using the same knowhow and technical expertise does not amount to the same or a similar economic undertaking, even where there was a team move. The Claimant was in a similar position to other highly skilled professionals. He and his colleagues moved to the Respondent where they were able to utilise their significant skills to enhance the Respondent’s business in areas new to their team, but not new to the Respondent.[121]I am mindful that case law warns against being confused by the effect of the acquiror endeavouring to assimilate staff into its own organisation. I do not think that is what happened here. The Respondent had its own suite of software so that any work the former Bloomsbury employees did, had to be started from scratch because to do otherwise was probably more time consuming and could have had intellectual property complications which the Respondent was determined to avoid. The Claimant and his colleagues may have used their specialist knowledge, but they did so on products and systems that they had not previously used. They were not hired in order to continue the operation they had previously worked in. Rather, the Respondent acquired them because they had a high level of AI knowhow which the Respondent expected could be applied in some way or other to the benefit of some more of the projects within the scope of the new SAIL operation as well as expert researchers whom they wanted to enhance FAIR.[122]The Respondent was keen to “seedbed” its operations in AI in London, and by getting the Bloomsbury team, it gained a team with a high reputation which did act as a talent magnet and which was able to apply significant expertise to specific problems within the Respondent group’s various product range. That was a distinct benefit to the Respondent.[123]In my view, taking a wide view of the operation as a sort of AI consultancy, is not a description of the Bloomsbury business. The activities which they took on at the Respondent were not sufficiently similar to the activities of Bloomsbury for the test of an economic entity which retains its identity to be satisfied.[124]There is no doubt that during the Bootcamp, the SAIL team in London had been introduced to Facebook’s code and had learned a lot about the overall operation. They did not attempt any continuing work on Cape or on the ShARC database. They learned about the Respondent and the Respondent group’s projects and details of their coding and used their own high degree of specialist knowledge to try to suggest adaptations or additions or ways of working that would enhance operations the Respondent was already trying to do, specifically as I have noted, contributing to meaningful improvements in protecting children from groomers, new models on improving URL hoaxes and improvements in identifying content used out of context. At the same time, while Sebastian continued to liaise with SAIL London, as FAIR was expected to do, the research scientists who had joined the FAIR team were free to pursue research of their choice. 10.5 Reserved judgment with reasons – rule 62 25 March 2017[125]I do not think the fact that the Claimant’s colleagues retained University contacts and indeed spent some time working at a University is of any meaningful significance. It was part of the attraction of this group of employees. The fact was that this connection enhanced the employees’ reputations and reflected well on the Respondent group. The Respondent group wanted to attract top talent in and around the world. In London and the UK, it had a need for a high quality people who other specialists would want to work with and would regard with professional esteem. It was, as recorded by Lilia Sirota, to be a talent magnet and it achieved that.[126]I reached this conclusion having carefully considered the case of Camden Primary Care Trust and in particular the provisions in that case which pointed out that the mode of providing the service or model may differ but that does not negate a relevant transfer where the nature of the activity remains the same. It is my view that it did not remain the same or sufficiently similar. I also bear in mind the fact that not all the employees transferred, two of them because they were not offered employment by the Respondent, is not a relevant factor and I do not take it into account.[127]The consequence of this determination is that the Claimant has less than two year’s service, and thus is insufficient qualifying service for a claim for ordinary unfair dismissal. Accordingly, that claim is struck out as the Tribunal does not have jurisdiction to consider it. Employment Judge N Walker