抖阴成年 is working with Google Cloud to connect HighQ with Gemini Enterprise for Legal, giving legal professionals a secure way to bring authorized matter documents, structured data, and workflows into the AI environment where they are collaborating.

The HighQ MCP connection is now live alongside Google Cloud’s announcement of Gemini Enterprise for Legal.

This is an important milestone, but it is also exactly that: one milestone. HighQ is the natural place to begin because it contains the live matter context AI needs to be useful. Over time, we see opportunities to extend more of the trusted capabilities of CoCounsel Legal across the AI environments our customers choose to use.

Why matter context matters

Enterprise AI continues to improve, but legal work has always depended on context. That context is more than documents. It includes structured matter data, workflows, tasks, templates, collaboration history, deadlines, permissions, and the institutional knowledge surrounding a case or transaction. Much of that lives in HighQ.

Law firms and corporate legal departments rely on HighQ to manage matters, collaborate internally and externally with clients and outside counsel, power secure client portals, and organize work that evolves over weeks, months, or years. Connecting HighQ with Gemini Enterprise for Legal allows authorized users to securely access that context without exporting files, copying information between systems, or recreating it somewhere else.

A legal team could ask Gemini Enterprise for Legal to summarize documents in an authorized matter, identify upcoming deadlines from an iSheet, compare structured deal data with underlying agreements, or bring relevant matter information into a draft.

The value is not simply giving AI access to more information. It is giving AI access to the right information, with the governance legal organizations already depend on.

Governance isn’t optional

Legal organizations should not have to choose between adopting new AI capabilities and maintaining the controls their work requires.

The HighQ MCP connection is designed so organizations don’t have to make that tradeoff. Content remains in HighQ. Once a customer securely connects HighQ, users authenticate using their existing HighQ credentials, and existing matter walls; folder permissions, document-level security, and audit logging continue to apply. Gemini Enterprise for Legal can retrieve only the content an individual user is already authorized to access. The connection is read-only, allowing AI to retrieve authorized information without changing the underlying content.

That means organizations can confidently extend trusted matter context into Gemini Enterprise for Legal while keeping governance exactly where it belongs.

Extending CoCounsel Legal into the enterprise AI ecosystem

This announcement is about more than connecting two products.

CoCounsel Legal remains the trusted platform where legal professionals perform complex legal work. It brings together authoritative legal content, customer context, agentic workflows, verification, and human oversight to help professionals produce work they can verify and stand behind.

At the same time, legal teams increasingly collaborate with business colleagues, outside counsel, customers, and partners who may be working in different AI environments.

Open standards like MCP make it possible to extend trusted legal context and capabilities into those environments without requiring organizations to move their work out of CoCounsel Legal or compromise on governance.

Today’s HighQ integration is the first exciting step in that broader vision with Google Cloud.

Building an open ecosystem for trusted legal AI

The future of enterprise AI will not be defined by a single model or a single interface.

Organizations will continue using different AI environments for different types of work. Our role is to ensure that wherever legal professionals collaborate, they can securely access the trusted matter context, authoritative content, and professional capabilities their work requires.

That is why 抖阴成年 continues to invest in MCPs, APIs, agentic systems, and strategic partnerships across the AI ecosystem.

“Customers have told us they want our AI solutions to connect seamlessly with the trusted systems they already rely on for everyday legal work,” said Satish Thomas, Vice President, Google Cloud. “Working together with 抖阴成年, we鈥檙e making it easy to bring critical matter context directly into their workflow today, while creating a path toward even deeper integrations over time.”

The launch of Gemini Enterprise for Legal and the HighQ MCP connection demonstrates what’s possible when trusted matter context and enterprise AI are designed to work together. It’s an important step in our collaboration with Google Cloud and another step toward making trusted legal AI available wherever professionals need it, while keeping CoCounsel Legal at the center of how that work gets done.

Thomson began as an internal project, built to solve a problem we had ourselves.听

抖阴成年 holds 175 years of authoritative data across legal, news, tax and accounting: Westlaw, Practical Law, Checkpoint and Reuters. We also employ thousands of subject-matter experts whose working lives are spent deciding what is correct. For three years we watched general-purpose models improve rapidly while both assets sat outside the training loop. We also faced the questions our customers were asking us: what dependency are we accepting on someone else鈥檚 architecture and pricing, and what do we do when the capability we need most is on nobody鈥檚 roadmap?听

Our answer to that was the Thomson LLM, and it worked well enough that we now want to share it with the rest of the world wrestling with these same questions.听

Where the argument came from听

The team that built Thomson did not arrive at 抖阴成年 with just a view about legal AI, but with a view about reliability.听

Safe Sign Technologies was founded in 2022 by lawyers and researchers whose background was in model safety, robustness and reliability, several coming out of applied AI in medicine and law, from Harvard and Cambridge. Medicine and law share a property most application domains do not: being nearly right is still wrong, and the cost of a confident error is borne by someone other than the person who made it. Both have long and demanding traditions of rigorous verification as a result, and those shaped how we approached the problem.听

The argument we made from that starting point was, at the time and until recently, unfashionable. In 2022 and 2023 the field was watching capability curves. The consensus was that frontier models would absorb professional work as a by-product of getting cleverer and that any attempt to keep pace with the 鈥渟caling laws鈥� of AI was futile. On that view the sensible move for a small company was to build a layer on top and wait.听

We believed the binding constraint was different. Capability, we argued, would become abundant; it was the object of enormous and well-funded competition, and there was no reason to expect it to stay scarce. What would remain scarce was trust and reliability: being right in a way that can be checked, in a domain where someone whose career depends on it. Trust is not a by-product of capability. It is a separate research problem requiring different evidence, and nobody was going to solve it for law as a side effect of solving it for everything.听

Very few people agreed. Making that case repeatedly, to investors and to ourselves, through pivots and long stretches with nothing to point at, was most of the job.听

抖阴成年 acquired Safe Sign in August 2024, in the company鈥檚 first pre-revenue acquisition. The team became 抖阴成年鈥� Foundational Research team, and crucially, the research posture that pre-dated the acquisition survived the transition. We continued to treat the work as a research problem rather than solely a product problem, which is why so much of the effort below went into measurement.听

Starting from open weights听

I said previously that our starting hypothesis was that capability would be abundant. The rate of progress of open-source AI has continued to prove this thesis over the last several years. Thomson benefits from this directly, with a leading open-weight foundation model as its starting point. We鈥檝e changed the root model of Thomson many times over the last several years, and will continue to do so as the frontier of open-weight models evolves. 听This is a tide that Thomson moves with, not one that washes it away. 听听

At the time of writing, the base model for Thomson is the Imperial College London Snowdon model. This model was developed by the FAIR Lab at Imperial, which 抖阴成年 and Imperial founded jointly, as an academic by-product of the acquisition of Safe Sign. 听

That choice is usually framed as a trade-off, and there is something to it: open-weight models can lag the closed frontier, and published analyses generally put that lag at a few months [1]. The conventional choice is, therefore, between capability and control.听

We did not think this was an acceptable dilemma for professional work. The frontier is measured on general capability, but our customers are judged on something narrower: whether a citation holds up, whether an answer is complete, whether the reasoning survives a partner鈥檚 review. There is no rule that a model strong on the second must concede the first. As we reported at launch, Thomson performs competitively with the strongest frontier models on the market, including Claude Opus 4.8, and ahead of GPT-5.5, Claude Sonnet 5 and Gemini 3.1 Pro. It also leads them on the measure this post is concerned with: whether the citations in a research report survive being checked.听

Turning the archive into training data听

抖阴成年 content is the deepest asset in this field and the reason a model of this kind was possible at all. It is also, as any archive of this scale would be, material that has to be prepared before a model can learn from it well.听

Content has to be found, which in an organisation of this breadth and history is a substantial exercise in itself. It must be assessed for rights, selected for measurable impact on model performance rather than relevance in the abstract cleaned, structured, deduplicated, and finally deployed into a data mixture, which is where the most consequential decisions are made.听

To date we have used less than ten per cent of 抖阴成年 content in continued pre-training. Westlaw, Practical Law, Checkpoint and Reuters News have been drawn on selectively. The areas where the model is not yet best in class are not ceilings we have reached, but areas where the relevant content has not yet been brought to bear.听

The specialisation problem听

Data mixture matters so much because specialising a model can damage it. Fine-tuning on domain-specific data can cause catastrophic forgetting: the model overwrites capabilities acquired during pre-training and its general performance degrades [2]. The effect is well documented, and mitigations exist, including replay of general data and regularisation of parameter updates. None fully solve it.听

One finding matters more than the others for our purposes. Kotha, Springer and Raghunathan鈥檚 听2024 study [3] examining what degrades during domain fine-tuning identified instruction-following as the principal contributor to forgetting: what erodes first is not the model鈥檚 knowledge of the world but its ability to do as it is told. For professional work that is close to a worst case, since real legal work is never only legal reasoning but legal reasoning while adhering to a format, a jurisdiction, a house style, an exclusion, a client鈥檚 standing preference. Output that requires reworking has not saved anyone any time.听

We therefore treated general capability retention as a first-class training objective rather than an acceptable loss. Instruction following is among the capabilities specialisation is most likely to erode, and it is one of the categories in our published benchmark results where Thomson stands up best against the frontier models, scoring 0.914 ahead of Claude Opus 4.8, Gemini 3.1 Pro and GPT-5.5. That is the clearest evidence we have that the model was specialised without being narrowed.听

Where the expertise actually comes from听

Many organisations claim their AI systems are 鈥渢rained with expert input鈥�. The phrase carries little meaning without an answer to the real question: how does a lawyer鈥檚 judgement become a training signal? Experts do not produce training data, but a standard. We have had to work to ensure the collective edge in expertise held by 抖阴成年 domain experts is realised in the quality of our training data. This is how we did it.听

Rubrics at maximum complexity. Partner-level practitioners worked full-time for months constructing evaluation rubrics for the hardest legal research tasks we could specify: the kind of multi-jurisdictional question where a good answer has fifteen necessary components and a plausible-looking one has nine. Each rubric enumerates what a correct response must contain. This is slow, expensive, and cannot be crowdsourced or synthesised.听

Commercial judgement, not only legal judgement. We required lawyers fresh out of commercial practice to ground the training data in what clients actually care about, which is frequently not what a textbook would emphasise. Take an indemnity. In most commercial agreements, it is heavily negotiated and often enforced, and treating it as significant is correct. But in an NDA it is usually neither, and almost never the crux. A model trained only on doctrine cannot tell those situations apart, and one that flags an NDA indemnity as urgently as the confidentiality carve-outs has identified a legal issue and wasted a lawyer鈥檚 attention. Teaching that distinction requires people who have sat on the other side of the negotiation.听

Preference data at scale. Thousands of hours of qualified lawyer time selecting between model outputs against complex criteria. Not 鈥渨hich is better鈥� but which better serves a client with a particular posture, in a particular jurisdiction, at a particular stage of a matter.听

抖阴成年 employs around 1,500 attorney-editors whose day job is producing the analytical content lawyers rely on. The obvious move is to train on their published output. The harder and more valuable move is to capture what happens between the first draft and the published article: the judgement calls, the discarded framings, the reasons a proposition was narrowed, the authority considered and rejected. That intermediate work is where so much expertise lives, and it is almost never written down.听

This problem generalises directly to our customers. A firm鈥檚 advantage is not simply its precedent bank: precedents circulate, deals become public, documents get shared. The advantage is what years of doing the work have built in the minds of its lawyers, who eventually retire or move. Capturing the reasoning rather than the artefact is the same problem, and we have worked on it at scale on our own corpus first.听

Internal deployment as a research instrument. Thomson has been deployed widely inside 抖阴成年, with thousands of domain experts using it on their hardest problems, which gives us failure modes reported by people qualified to diagnose them. Our teams are not incentivised to use Thomson for Thomson鈥檚 sake; if they use it, it is because they have decided it can do something others can鈥檛.听

The consistency problem听

Expertise does not straightforwardly produce consistency. In some respects it produces the opposite: the more experienced the practitioner, the more nuanced their judgement, which is exactly what you want in a partner and exactly what creates noise in a training set. Two excellent lawyers can disagree on a scoring decision not because either is wrong but because each applies a refined intuition the other does not share.听

The literature bears this out. On the LEXam legal reasoning benchmark [4], three legal experts independently scoring the same answers on a ten-point scale reached a quadratic weighted kappa of 0.49, with a mean absolute deviation approaching two points. Work on implicit legal citations [5] reports similar or worse agreement. More troubling, Rehag鈥檚 survey of legal machine learning datasets found they systematically removed all traces of disagreement rather than treating conflicting expert annotations as informative [6].听

Take expert output at face value and train on it, and you teach the model an averaged version of several incompatible standards: vaguely acceptable to everyone rather than correct according to anyone.听

A large share of our effort therefore went into data quality: calibrating annotators against worked examples, measuring agreement continuously and treating drops as signals about the task specification rather than the annotator, and structuring rubrics tightly enough that disagreement surfaces as genuine ambiguity rather than noise. Where it persists, the question is usually contested, which is itself something the model should learn.听

Safety, values and red-teaming听

A dedicated team of lawyers worked on bias, political neutrality and toxic behaviour, with extensive human and automated red-teaming. We treat these as training objectives rather than output filters: a filter catches a bad answer on the way out, an objective changes what the model is disposed to produce.听

Political neutrality deserves particular mention given that Reuters sits inside this company. Realignment towards factuality and pluralism was an explicit part of the training programme rather than a compliance exercise appended to it, and it is measured rather than asserted: Thomson performs strongly against the frontier models on our internal neutrality evaluation, with detail to follow in the technical report. For a company that publishes news as well as legal analysis, that is not peripheral.听

Beyond legal data听

Not all of the training data is legal. We drew on domains rich in explicit chain-of-thought reasoning, where the reasoning must be set out rather than left implicit, on the view that a model reasoning well in structured non-legal settings reasons better in legal ones. Checkpoint and Reuters give depth in tax, accounting and world events, because legal work is rarely purely legal.听

Rigour and factuality听

Our own lawyers publish at leading AI conferences [7] the people building the evaluation apparatus treat it as research rather than quality assurance, which is what makes it rigorous enough to train against.听

We think that rigour produces the result we care most about. In our published deep research evaluation, Thomson working over Westlaw and Practical Law scored 0.83 on factuality against 0.65 and 0.68 for leading frontier models given unrestricted access to the open web. Completeness was close between all three. Factuality was not.听

Completeness is a capability measure: it asks whether the system covered the ground, and on it the three were nearly level, because frontier models with the open web and enough time will generally find the material. Factuality is a reliability measure: it asks whether the system can be checked and survive it, and on that they were not close at all.听

The metric is not a measure of whether an answer sounds authoritative or whether the conclusions are broadly sound. Every claim is extracted and matched against the source cited for it, and the score is the proportion of assertions whose own citations hold up when checked.听

That is the failure that has kept general-purpose AI in the assistant鈥檚 chair. A system reliably right about its own sources is a different category of instrument from one merely fluent about them. It is the difference between something an associate uses and something a partner signs.听

We call this Fiduciary-Grade AI: a standard for AI used where accuracy, accountability and trust are not optional, for professionals working under duties of care and regulatory oversight. Thomson demonstrates that it can be pursued at the model layer rather than bolted on above it.听

Capability and sovereignty are not mutually exclusive听

The lesson is not just about law. Any organisation holding a deep proprietary corpus and real domain expertise has been told it must choose: either rent frontier capability and accept the dependency or own an open model and accept some distance from the frontier. The choice is false, provided you are willing to do the unglamorous work of preparing the data, converting expert judgement into consistent signal, and building the evaluations before you try to move the numbers. The reward is a model you own rather than rent, pointed at the problems you choose, improving on your schedule rather than somebody else鈥檚.听

The compute is not the binding constraint. The corpus and the people who know what correct looks like within it are, and those have never been concentrated in the frontier laboratories. They sit inside institutions that spent a century accumulating them without thinking of themselves as AI companies.听

We are one of those institutions. Thomson is what happened when a research team that had spent three years arguing trust and reliability were the scarce input finally got access to the data and the experts to prove it.听

Thomson enters production this month powering CoCounsel skills including high-volume structured document review, with integration across the legal and tax portfolio to follow. A full technical report is forthcoming. Thomson was built in collaboration with DatologyAI, Lambda, Together AI, Imperial College London and the 抖阴成年鈥揑mperial Frontier AI Research Lab.听

Sources for external claims听

[1] Epoch AI, open-weight capability lag analyses (October 2025; May 2026); Stanford AI Index 2026.听

[2] Luo et al., “An Empirical Study of Catastrophic Forgetting in Large Language Models” (2023); Song et al., arXiv:2501.13669.听

[3] Kotha, Springer and Raghunathan, arXiv:2406.12227.听

[4] LEXam, arXiv:2505.12864.听

[5] “Where Experts Disagree, Models Fail”, arXiv:2603.22973 (2026).听

[6] Rehaag, “I beg to differ”, Artificial Intelligence and Law (Springer, 2023).听

[7] Yejin Bang, Kirsty Fielding, Brandan Oliver, Brian Birke, Nabeel Seedat & Andrew M. Bean, ContractScrub: A Benchmark for Final Review of Legal Contracts (抖阴成年 Foundational Research, 2026) (in Proceedings of the AI for Law Workshop at the International Conference on Machine Learning (ICML 2026), available at ); and Samuel J. Vincent, Daniel Calloway, Fangyi Yu, Andrew M. Bean & Nabeel Seedat, InsufficiencyBench: Evaluating LLM Legal Advice on Underspecified User Queries (抖阴成年 Foundational Research, 2026) (in Proceedings of the AI for Law Workshop at the International Conference on Machine Learning (ICML 2026), available at ).听

Those systems hold far more than files. They contain the history of matters, the experience of teams and the accumulated knowledge of the organization.

抖阴成年 and iManage have worked together for years to connect that knowledge with the tools legal professionals use every day. Our renewed and expanded agreement extends those integrations across products including CoCounsel Legal, HighQ and Noetica, while creating a foundation for the next generation of connections between our platforms.

Model Context Protocol (MCP) is one area where that next generation is beginning to take shape. As iManage expands access to its MCP capabilities, we plan to integrate them with CoCounsel Legal, HighQ and Noetica. MCP will add another way for these systems to work together, complementing the integrations customers already use to access, browse and sync iManage documents and bring them into 抖阴成年 workflows.

Legal organizations have established workflows, permissions and governance around their information. New AI capabilities should build on that foundation while giving professionals more ways to use the knowledge they already trust.

A more connected legal workflow

Legal work rarely happens in one application. Research, drafting, document management, collaboration and transactional work all draw on different systems and sources of information.

Our longstanding integrations with iManage already help connect those environments. The renewed and expanded partnership allows us to continue supporting those workflows while opening up new ways for the systems, context and tools our shared customers rely on to work together.

When implemented, the MCP will allow shared customers to query iManage鈥檚 AI within CoCounsel Legal. For CoCounsel Legal, that fits within a broader ambition: helping professionals bring the right legal content, documents and tools together as they work. Westlaw and Practical Law provide authoritative legal content and expertise. Customer documents provide the facts and matter-specific context. Connections with platforms such as iManage make it easier to work across those sources without reconstructing the workflow each time.

Law firms have spent years building valuable institutional knowledge. As AI becomes more embedded in legal work, that knowledge should become easier to use while preserving the controls organizations depend on.

That is where partnerships like this one matter: connecting the systems professionals already trust with the AI experiences they increasingly rely on.

Lawson Lundell, British Columbia’s largest law firm, with over 230 lawyers across four offices in Vancouver, Calgary, Kelowna, and Yellowknife, specializes in business law with deep expertise in corporate matters, real estate, labor and employment, and business litigation. As the firm continued to grow and attract top talent, leadership recognized that staying competitive meant embracing AI not just as an experiment, but as a strategic imperative.

“It’s about attracting and retaining the best and brightest lawyers, especially the next generation, and making sure we have the best tools to equip them from day one.鈥� 听said Clifford Proudfoot, KC, Managing Partner, Lawson Lundell. 鈥淏ut ultimately, it’s about our clients.”

“We’re already seeing expectations for AI use in many of our client agreements” said Derrick Li, COO and CFO, Lawson Lundell.

After piloting multiple AI solutions, Lawson Lundell selected 抖阴成年 for its unique combination of Westlaw鈥檚 authoritative legal content, enterprise-grade security, and seamless integration of the content and AI workflow capabilities. The decision was also influenced by the fact that 80% of the Am Law 100 choose 抖阴成年, as well as their strong Canadian presence and consistent relationship management.

“There’s significant resources required from an R&D perspective to stay ahead in AI,” Li explained. “In the race to dominate Legal AI, I would bet my money on 抖阴成年.”

CoCounsel’s ability to support research, drafting, and document analysis within a unified, intuitive platform allows Lawson Lundell to reduce tool sprawl while maintaining the high standards of accuracy and reliability the firm’s clients expect. With guided onboarding, best-practice templates, and enterprise governance built in, the firm is well-positioned for rapid, responsible adoption across all practice areas.

“I almost feel like a kid at a toy store,” Li said. “I’m most excited about equipping every single articling student, paralegal, and lawyer with AI solutions. This will be the first group that will have AI from day one, and it will be fascinating to see how that changes how lawyers approach client solutions.”

Lawson Lundell plans to roll out CoCounsel firmwide alongside its own internally developed AI tools, creating a comprehensive AI ecosystem designed to serve both legal professionals and support staff. The firm has established an AI committee and pilot groups to ensure thoughtful implementation and maximize value realization.

Version fran莽aise

Lawson Lundell choisi CoCounsel de 抖阴成年 pour offrir une valeur client am茅lior茅e et renforcer la nouvelle g茅n茅ration de professionnels juridiques

Alors que l’intelligence artificielle (IA) transforme le paysage juridique, Lawson Lundell LLP se positionne 脿 l’avant-garde de l’innovation en int茅grant 听CoCounsel de 抖阴成年 脿 l’茅chelle de l’entreprise. Cette d茅cision refl猫te l’engagement du cabinet 脿 茅quiper ses professionnels du droit avec des outils de premier ordre tout en r茅pondant aux attentes 茅volu茅es des clients dans une 猫re ax茅e sur l’IA.

Lawson Lundell, le plus grand cabinet d’avocats de Colombie-Britannique, avec plus de 230 avocats r茅partis dans quatre bureaux 脿 Vancouver, Calgary, Kelowna et Yellowknife, se sp茅cialise dans le droit des affaires avec une expertise approfondie en mati猫re d’entreprise, d’immobilier, de droit du travail,听 et de litiges commerciaux. Alors que le cabinet continue de cro卯tre et d’attirer les meilleurs professionnels, la direction a reconnu que rester comp茅titif signifiait adopter l’IA non seulement comme une exp茅rience, mais comme une n茅cessit茅 strat茅gique.

芦 Il s’agit d’attirer et de retenir les meilleurs et les plus brillants avocats, en particulier la prochaine g茅n茅ration, et de s’assurer que nous avons les meilleurs outils pour les 茅quiper d猫s le premier jour 禄, a d茅clar茅 Clifford Proudfoot, KC, associ茅 directeur de Lawson Lundell. 芦 Mais en fin de compte, il s’agit de nos clients. 禄

芦 Nous voyons d茅j脿 des attentes pour l’utilisation de l’IA dans bon nombre de nos accords clients 禄, a d茅clar茅 Derrick Li, COO et CFO de Lawson Lundell.

Apr猫s avoir test茅 plusieurs solutions d’IA, Lawson Lundell a choisi 抖阴成年 pour sa combinaison unique de contenu juridique autoritaire de Westlaw, de s茅curit茅 de niveau entreprise et d’int茅gration transparente des capacit茅s de contenu et de flux de travail d’IA. La d茅cision a 茅galement 茅t茅 influenc茅e par le fait que 80 % des Am Law 100 choisissent 抖阴成年, ainsi que par leur forte pr茅sence canadienne et leur gestion coh茅rente des relations apr猫s vente.

芦 Il faut des ressources importantes du point de vue de la R&D pour rester en t锚te dans l’IA 禄, a expliqu茅 Li. 芦 Dans la course pour dominer l’IA juridique, je parierais sur 抖阴成年. 禄

La capacit茅 de CoCounsel听 脿 soutenir la recherche, la r茅daction et l’analyse de documents au sein d’une plateforme unifi茅e et intuitive permet 脿 Lawson Lundell de r茅duire la prolif茅ration des outils tout en maintenant les normes 茅lev茅es de pr茅cision et de fiabilit茅 attendues par les clients du cabinet. Avec une int茅gration guid茅e, des mod猫les de meilleures pratiques et une gouvernance d’entreprise int茅gr茅e, le cabinet est bien positionn茅 pour une adoption rapide et responsable dans tous les domaines de pratique.

芦 Je me sens presque comme un enfant dans un magasin de jouets 禄, a d茅clar茅 Li. 听听听听芦 Je suis le plus enthousiaste 脿 l’id茅e d’茅quiper chaque 茅tudiant stagiaire, parajuriste et avocat avec des solutions d’IA. Ce sera le premier groupe 脿 avoir l’IA d猫s le premier jour, et il sera fascinant de voir comment cela change la fa莽on dont les avocats abordent les solutions pour les clients. 禄

Lawson Lundell pr茅voit d茅ployer CoCounsel 脿 l’茅chelle de l’entreprise aux c么t茅s de ses propres outils d’IA d茅velopp茅s 脿 l鈥檌nterne, cr茅ant un 茅cosyst猫me d’IA complet con莽u pour servir 脿 la fois les professionnels du droit et le personnel de soutien. Le cabinet a 茅tabli un comit茅 d’IA et des groupes pilotes pour garantir une mise en 艙uvre r茅fl茅chie et maximiser leur investissement.

Double-Digit Profit Growth

Law firms ended 2024 on a high note, with the average firm experiencing double-digit profit growth for the year. In the fourth quarter alone, profits grew by an impressive 11.5%. This strong financial performance underscores the industry’s ability to adapt and thrive.

Transactional Practices Lead the Way

Transactional practices were a significant driver of growth in Q4, with corporates seeing a 4.0% increase and real estate growing by 3.0%. Counter-cyclical practices slowed after two years of record growth, counterbalancing the increase in transactional work. The report suggests that demand is expected to be more static in the first half of 2025 compared to 2024.

Investment in Technology and Knowledge Management

Law firms have been investing heavily in technology and knowledge management, leading to a 6.9% increase in overhead expenses in Q4. These investments are crucial for maintaining competitive advantage and improving client service. Direct expenses also rose by 6.2% as firms paid out sizeable bonuses to their associates and staff.

Embracing AI and Upskilling

Many law firms are strategically using AI and that is expected to drive sustainable productivity growth and position firms for continued success in 2025.

鈥淟aw firms continue to successfully navigate the dynamic landscape and demonstrate their commitment to serve their clients at the highest level by investing significantly in technology, including generative AI, and the upskilling of their people,鈥� said Raghu Ramanathan, president of Legal Professionals, 抖阴成年. 鈥淟aw firms enter 2025 from a position of strength and appear ready to drive profitability. As the AI market matures, law firms can enhance job satisfaction, well-being, and work-life balance by using AI for routine tasks, freeing up time to focus on their clients鈥� complex needs and driving sustainable productivity growth.鈥�

Download the full report听for additional insights on the factors shaping the future of law firms.

The Emergence of LLMs in Automatic Evaluation

Traditional evaluation metrics in NLG primarily rely on comparing the generated text to reference texts using word overlap measures. These metrics, while useful, often fall short of capturing the nuances of language quality, coherence, and relevance. For example, suppose we have a reference summary 鈥淭he cat is on the mat.鈥� and a generated summary from a model 鈥淎 feline is resting on a rug.鈥�. If we use ROUGE as a metric for the evaluation, ROUGE only considers lexical overlap and cannot capture the semantic similarity between words or phrases. These summaries have the same meaning but would score poorly on ROUGE due to low word overlap.

LLMs have demonstrated an exceptional understanding of language, context, and semantics, making them attractive candidates for evaluating generated text. They can assess factors like fluency, coherence, and even factual accuracy, which are crucial for more sophisticated and context-aware evaluations. For instance, LLMs can be fine-tuned to understand the specific jargon and style of a particular domain, such as medical or legal texts, making them great potential evaluators.

The Promise of LLMs in Automated Evaluation

The recent technological advancements of LLMs have encouraged the development of LLM-based evaluation methods in various tasks and systems. LLMs can offer several advantages in the evaluation of NLG tasks:

• Context-Aware Evaluation: Unlike traditional metrics, LLMs can comprehend the context and generate evaluations that account for the subtleties and intricacies of human language.
• Scalability: LLMs can evaluate large volumes of text quickly and consistently, offering scalability that human evaluators cannot match.
• Reduced Subjectivity: Automated evaluation can minimize the subjective bias that human evaluators might introduce, leading to more consistent and objective assessments.

How to use LLM as an Evaluator

LLM-based evaluators are conceptually much simpler than traditional automatic evaluators for evaluation. While traditional evaluation methods rely on predefined metrics and comparisons to reference datasets, LLM-based evaluators work by directly assessing the generated text.

Figure 1 shows an overview of the LLM-based evaluation frameworks. To evaluate the quality of the text, you embed it into a prompt template that contains the evaluation criteria, then provide this prompt to an LLM. The LLM then analyzes the text based on the given criteria and provides feedback on its quality. This approach bypasses the need for extensive preprocessing and reference comparisons, making the evaluation process more straightforward and versatile.

Figure 2 illustrates the step-by-step workflow of evaluating summaries using clarity as a metric. A document and its corresponding summary are inputs. The summary is then embedded into a pre-formulated prompt template that includes detailed evaluation criteria, such as clarity, defined on a 1-to-5 scale. This prompt is then inputted to an LLM for automated analysis and evaluation. Based on the specified criteria, the LLM evaluator reviews the summary, assigns a score, and provides justification for the rating.

The Limitations of LLMs: Why Humans Are Still Indispensable

Despite the promising capabilities of LLMs, there are significant limitations that necessitate the continued involvement of human experts in the evaluation process, especially for domain-specific tasks:

• Lacking Specialized Domain Knowledge: Domain-specific tasks often involve complex knowledge. LLMs are typically trained as general-purpose assistants, and they still lack specialised domain knowledge. In contrast, subject matter experts bring in- depth domain knowledge obtained by years of dedicated training and education.
• Evolving Knowledge: Especially in fast-evolving fields like medicine and technology, staying up-to-date with the latest information is challenging for static models. Human experts, however, continuously learn and adapt to new knowledge and standards.
• Handling Ambiguities: In specialized domains, the language can be highly ambiguous and complex, and the ability to disambiguate based on deep contextual knowledge is something LLMs still struggle with.
• Ethical and Bias Concerns: LLMs can inadvertently reinforce biases present in their training data. Human oversight is crucial to identify and mitigate these biases, ensuring fair and ethical evaluations.

The Human-in-the-Loop Model: Best of Both Worlds

To harness the strengths of LLMs while addressing their limitations, a human-in-the-loop approach is essential. This combines the efficiency and scalability of LLMs with the expertise and judgment of human evaluators:

• Initial Screening: LLMs can perform initial screenings and provide preliminary evaluations, identifying clear cases of high or low quality.
• Expert Review: Human experts then review and refine these evaluations, focusing on cases that require nuanced understanding or where the LLM鈥檚 assessment is inadequate.
• Continuous Feedback Loop: Feedback from human evaluators can be used to fine-tune and improve LLMs, creating a continuous improvement cycle.

Conclusion

The integration of LLMs into the automatic evaluation of NLG tasks marks a significant step forward in the field of NLP. However, for domain-specific evaluations, the complexity and nuance of human language still necessitate human experts. By adopting a human-in- the-loop approach, we can leverage the best of both worlds: the speed and scalability of LLMs and the depth and discernment of human evaluators. This constructive interaction ensures that we maintain high standards of accuracy, fairness, and relevance in evaluating natural language generation tasks, ultimately driving the field towards more sophisticated and reliable applications.

This post was written by Grace Lee, lead applied scientist at 抖阴成年 Labs (TR Labs).

Note. This work has been done as part of the internship of Hossein A. (Seed) Rahmani at 抖阴成年 Labs (TR Labs).

Foundation of enterprise-wide innovation听

Maria Apazoglou, head of AI, BI & Data Platforms, Technology, recalled early efforts to create a GenAI platform within 抖阴成年, upon which the company鈥檚 GenAI-powered solutions 鈥� including 鈥� would be built. 听

鈥淥ur GenAI platform is designed to democratize access to AI, making it easier for non-technical users to utilize AI effectively,” Apazoglou explained. “In the creation of the platform, we focused on two critical aspects: developing a multi-cloud platform to support various AI models and LLMs, and creating an accessible interface to allow users to self-serve, enabling early experimentation and viability assessment.鈥�听

This approach facilitated rigorous testing by developers and evaluation by subject matter experts (SMEs), ensuring optimal model performance. Today, the platform has 10,000 internal users monthly 鈥� equal to 40% of the population of 抖阴成年 25,000 colleagues.听

Bridging content and customer needs听听

When it comes to responsible AI, human-in-the-loop is critical to 抖阴成年 strategy, with considerations at every stage of design, development and deployment.听听

The company has a team of hundreds of tax and accounting experts 鈥� many of whom are attorneys and CPAs themselves. These SMEs are critical to enhancing the quality of the responses produced by the GenAI models and to creating a system that meets customers鈥� needs. Evaluating results and grading answers is an ongoing and constant process. 听

Catherine Murray, director of Editorial, US Tax & Accounting Commentary, coordinated the SME grading efforts, working with 抖阴成年 developers to interpret the results. She channeled her expertise in tax law, along with her understanding of customers鈥� needs, into ensuring that the models provided valuable and accessible responses.听听

鈥淲orking closely with 抖阴成年 Labs team to conduct experiments and optimize the solution for large language models was an exciting part of the work,鈥� she said. 鈥淭hinking about how customers are going to use the solution to surface our high-quality content was fascinating to be involved in, particularly as GenAI tech is evolving all the time.鈥�听

Melissa Oaks, director of Editorial, Current Awareness, noted the excitement from customers during the alpha and beta testing phases. 鈥淐ustomers were eager to keep the beta version and asked, 鈥榃hen can I buy it?鈥� It鈥檚 thrilling to work on something that has such a significant impact on their daily lives,鈥� Oaks remarked. 听

Oaks noted that other solutions on the market rely on “just the tax code or the primary sources, and that’s really not enough to do tax research.” She said: 鈥淲hat I love about this solution is we鈥檙e providing professional insights and step-by-step guidance on how to actually go do what we鈥檙e saying is the recommended course of action, not just the primary sources. This comprehensive approach sets us apart in the market.鈥�听

Transformative time-savings听

Nancy Hawkins, vice president of Product Management, Research, underscored the product鈥檚 potential to alleviate industry pressures: 鈥淲ith current tax staffing shortages, our solution offers crucial time-savings and work augmentation, especially for non-attorney professionals conducting tax research. It鈥檚 helping customers gain more capacity, whether for work-life balance or business growth.”听

鈥淭he exacting nature of the tax profession means mistakes can have significant consequences, and our customers are hungry for solutions that can bring them time-savings and augment their work,鈥� Hawkins said. 听

Delivering accuracy, efficiency, and simplicity听

Erica Butcher, vice president of Sales & Retention – Strategic Firms, noted the industry’s cautious optimism towards GenAI.听 She said tax professionals are 鈥渂oth cautious and hopeful鈥� about GenAI, and they are looking to the tech to deliver accuracy, efficiency, and simplicity.听

鈥淥verall, we are mainly hearing excitement from customers about the potential for GenAI to support top priorities of improving efficiency and addressing the current talent shortage we鈥檙e seeing across tax and accounting,鈥� Butcher said. “We鈥檝e really listened to our customers, and we built this solution with our customers鈥� feedback and needs in mind to ensure Checkpoint Edge with CoCounsel can deliver meaningful, measurable results.鈥�听

For more on Checkpoint Edge with CoCounsel, visit鈥� or check out鈥�.

鈥淲hen I think about efficiencies in the tax industry, especially given the staffing shortages and the uptick in retirements, efficiencies are so important and a huge driver for the development of our tax tools,鈥� Hawkins said. 鈥淲e know from our research that customers are prioritizing these efficiency gains and looking to technology to really help them unlock them.鈥�

Wong cited findings from the latest 抖阴成年 Future of Professionals report that demonstrated a shift in the tax and accounting industry鈥檚 adoption of AI tools.

鈥淚n contrast to a year ago, where we had the most excitement and the most interest from the legal industry, this year in 2024, we’ve had a much clearer signal from the tax, accounting and audit industry that they are ready and eager to apply some of the AI technology to their work to seek efficiencies,鈥� Wong said.

He added that this shift can help the tax and accounting industry address a labor crunch.

鈥淭here aren鈥檛 enough CPAs or enough people who are willing to jump into the tax and accounting profession, and technology is seen as this sort of savior for this ever-growing pile of work within the industry,鈥� Wong said. 鈥淎I is one additional set of tools to help the industry get to a new frontier of automation.鈥�

Wong and Hawkins noted that AI tools must work in conjunction with the other workflow tools tax and accounting professionals already depend on.

鈥淭he exacting nature of the tax profession is extremely unforgiving when it comes to inaccuracies or mistakes,鈥� Hawkins said. 鈥淎s we put our AI-Assisted Research together, we really had that in mind 鈥� to help with making research more efficient for the new entrants into the industry.鈥�

鈥淗aving an efficient way to research is critical,鈥� Hawkins added. 鈥淭he AI-Assisted Research that we have in Checkpoint Edge really allows someone to come in and simply ask a question, like you would to a trusted colleague down the hall and get an answer back that is easily digestible.鈥�

Hawkins said she hears customers talk about how senior members of firms want to mentor junior members but don鈥檛 have the time. She said firm leaders are excited about how generative AI tools can help 鈥渦pskill more junior members of a firm quickly to have the mature and robust conversations around tax topics.鈥�

Watch the听of the TechConnect series, which brings diverse and dynamic perspectives from all corners of the technology world with thought-provoking questions and conversation.

Jake Heller, head of Product for CoCounsel, 抖阴成年, and Jake Porway, co-founder of DataKind and an NCSC AI consultant, will host the webinar as part of a new AI and the Courts series of monthly discussions. The first session will offer participants a foundational understanding of AI and its potential to enhance the efficiency and effectiveness of the judicial process, highlighting current applications of AI in the court system and the ethical implications of its use.

Launched in June, the consortium is a joint initiative designed to educate the judiciary about the opportunities and challenges of evolving AI and generative AI solutions, enabling judges and legal and court professionals to make informed decisions about adoption and use.

鈥淚鈥檓 thrilled to help bring together the legal industry鈥檚 top AI experts from the courts, law firms, academia, and technology organizations,鈥� said Heller. 鈥淭he pace of innovation in the legal industry is fast and furious, and 抖阴成年 has a long tradition of customer collaboration and leadership in applying cutting-edge technologies to legal research and legal workflows. The AI Policy Consortium for Law and Courts will be a tremendous resource for the judiciary and legal professionals seeking to keep up with the quickly evolving AI and generative AI tools available to augment the practice of law.鈥�

鈥淗aving more than 1,000 participants registered for the webinar speaks to the judiciary and legal profession鈥檚 eagerness to better understand AI solutions and their impact on how legal professionals work,鈥� Porway said. 鈥淥ur consortium is filling a gap in the industry as courts and legal professionals navigate how to best evaluate, adopt, and sanction the uses of generative AI. I鈥檓 excited to introduce the AI and the Courts series with an exploration of the current and future ways AI is used in the court system as well as its challenges and implications.鈥�

The consortium鈥檚 four workstreams 鈥� AI governance and ethics, workforce readiness for AI adoption, rules and practices pertaining to AI, and AI鈥檚 impact on access to justice 鈥� examine the opportunities and risks of AI and generative AI. The second webinar, Ethics of Generative AI: A Guide for Judges and Legal Professionals, will be held on Sept. 18. Future topics will include AI鈥檚 impact on the judicial and legal workforce, how AI can enhance court efficiency, and more.

Register for the webinar . To learn more about on the consortium, read the press release.

What truly stood out to me at ILTACON was the energy and enthusiasm of the event. I was struck by how the collaborative atmosphere among attendees 鈥� ranging from seasoned legal professionals to tech enthusiasts 鈥� created an enriching environment for knowledge exchange and networking.

The conference center was brimming with excited representatives from law firms and vendors showcasing innovative solutions. Everyone was grappling with how 鈥� and when 鈥� generative AI could transform the practice of law.

The insightful sessions on cybersecurity, data privacy, and the future of legal tech trends provided valuable takeaways that are likely to shape the industry in the coming years. Overall, the conference underscored the rapid advancements and the pivotal role of technology in transforming the legal landscape.

CIO insights

I moderated a among G100 and G200 CIOs. The CIOs discussed the evolving role of the document management system, the opportunity for generative AI to drive efficiency improvements now and in the future, the evolving expectations of corporate law departments, and data security practices.

Audience attendees seemed especially interested in the tenor of the conversation around transformative AI solutions. One of the CIOs said that his firm was using CoCounsel as its AI assistant and that lawyers at the firm were eager adopters. He mentioned wanting to buy more seats. That was a fun moment for me!

CoCounsel 2.0

We hosted a number of extremely well-attended events, including customer dinners, an appreciation event, and other celebrations. The opportunity to share perspectives and learn from law firms was invaluable.

I was thrilled to see that the announcement of CoCounsel 2.0 seemed to generate the most excitement among product news. Jake Heller, head of Product for CoCounsel, 抖阴成年, showcased a side-by-side view on the speed of the application, highlighting how CoCounsel 2.0 moves more than 3 times faster than the current version CoCounsel Core. After customers saw that, they were clamoring for access.

On a personal level, I enjoyed seeing so many former colleagues and peers. The legal community is a tight-knit group. It was amazing to walk the halls of the conference centers and catch up with so many great people.

Check out other 抖阴成年听ILTACON recaps, including the legal industry鈥檚 reaction to the听launch of CoCounsel 2.0. and highlights from the 听鈥淗ow Today鈥檚 Lawyers Are Enhancing Their Practice with AI鈥� session.