Definely Launches Agentic AI Capabilities + AL Interview – Artificial Lawyer

Definely, which focuses on the review of complex contracts, has launched ‘Enhance’, a new proprietary agentic AI system – joining a growing movement of agentic roll-outs across the legal tech market. Plus – see the In-Depth AL Interview below with Sigurjón Ísaksson, Head of Artificial Intelligence.

First, it’s great to see how rapidly agentic systems are spreading now. Why? Because single prompts only get you so far in terms of efficiency. An agent that can be customised to your needs and then take on a series of tasks, or one task again and again, adds extra efficiency to your workflow.

So, what does it do? The UK-based company, which now has a global reach, explained that with ‘Enhance, each model — or agent — specialises in specific tasks, such as summarisation or drafting support. Definely’s agentic system analyses the task, creates a plan, and assigns agents to handle different parts of it’.

Plus – and AL has to say, this site always likes it when companies say this bit….the ‘agents collaborate’. There’s something very pleasing about being able to wield multiple agents that work together to complete a complex task. It’s a modern version of Adam Smith’s treatise on the specialisation of labour. Although in this case it’s the specialisation of legal labour performed by a co-ordinated group of AI agents.

Plus, this connects with the ‘power-sharing’ idea that AL has been exploring recently, i.e. that if we’re really going to get the most out of legal AI tools, then we have to actually give them ‘responsibility’, as it were, for doing a task and completing that task – or most of it, or in other words, give them ‘agency’ 🙂 ..!

—

Definely Webinar: Success Tactics – Change Management + AI – June 17

—

Of course it’s not all a question of pressing a key and off they go. Agents’ performance needs to be refined with human expert input. (And as mentioned yesterday re. Microsoft’s Contract Builder agent – this is actually an essential part of the process when doing legal work.)

As Definely noted: ‘The performance of the AI agents [is] evaluated and improved using Definely’s proprietary legal dataset and in-house expertise.’

And, ‘lawyers [can] interact with the agentic system within Microsoft Word’ – which will be essential, AL would add, to make sure you’re getting the outputs you want.

All in all what you get is ‘a 40 – 70% speed improvement in workflows versus a traditional manual approach’.  

Feargus MacDaeid, Co-founder and Chief Strategy Officer of Definely, commented: ‘Lawyers work in high-pressure, high-stakes environments where a minor error in the wording of a contract could cost a law firm millions. At Definely, we are committed to building AI-focused products that will genuinely help our customers de-risk their day-to-day work. As a company, we’ve not rushed head-first into building generic AI tools; rather, we’ve spent time identifying segments of the market that will derive the most value for lawyers and working alongside them to solve the problems that matter most.’

–

In-depth AL Interview with Sigurjón Ísaksson, Head of Artificial Intelligence at Definely. 

You state that ‘Agents collaborate and gather information, passing it along until completion’ and then performance of the AI agents is evaluated and improved using Definely’s proprietary legal dataset and in-house expertise’ – this sounds very interesting – can you please explain how this happens?

Each agent within our ecosystem is designed to handle a distinct task essential for comprehensive legal work. By adding new agents, we continuously expand the system’s capabilities, enabling it to manage increasingly complex legal workflows and queries.

Importantly, these agents have access to a robust set of tools, which include our proprietary Definely products. The communication with these products is orchestrated using MCP (Model Context Protocol), which ensures seamless integration. Definely’s suite of tools, developed specifically to address complex legal workflows, equips our agents to perform end-to-end legal tasks efficiently and accurately. This unique integration between our agents and Definely’s advanced tools is critical – it enables the completion of comprehensive workflows that significantly streamline a lawyer’s workload, something achievable only because of the groundwork laid through the prior development of our Definely product suite.

How would you define ‘agent’ in this context?

An agent is a modular, autonomous component within our system that is designed to perform a specific function or set of tasks as part of a larger, orchestrated workflow. Each agent is built for a clearly defined purpose (e.g., document understanding, precedent retrieval, legal research) and is equipped with its own logic, tools, and access to context, allowing it to make decisions, process information, or interact with data in a goal-oriented way.

How much can users customise and control the agents?

We’ve integrated a human-in-the-loop process, which means users can review and adjust the generated plan and validate or refine individual agent outputs. This keeps the workflow transparent and adaptable while leveraging the strengths of automation.

Our agents integrate directly with our Vault product, which contains all the precedents and gold standards accumulated by the law firm, along with precedents individually stored by lawyers in their personal vaults. As a result, outputs such as drafted legal texts or specific task solutions draw directly from these tailored precedents, reflecting each user’s own historical practices rather than relying on generic responses from a general-purpose language model.

Why do this and why now?

Recent advancements in large language models, sophisticated agent-to-agent communication, and seamless integration of external tools via the Model Context Protocol (MCP), all critical elements, are now aligning. These technological breakthroughs enable the creation of powerful, collaborative multi-agent systems that can comprehensively solve complex workflows end-to-end, precisely at a moment when businesses are eager for AI-driven solutions that surpass traditional chatbots and offer real operational transformation.

Which LLMs are you using for this?

We use a range of leading LLMs, selecting the best model for each task based on its strengths. Some models are better at reasoning, others at summarisation, or handling long documents. Our system is designed to route subtasks – like legal research, precedent matching, or context retrieval – to the most suitable model, ensuring high accuracy, efficiency, and control.

How do you ensure accuracy?

To ensure our system performs reliably and accurately, we’ve established rigorous evaluation and benchmarking processes. Using automated evaluation pipelines and our proprietary legal dataset, developed through extensive collaboration with in-house legal experts, we continuously test and refine our platform. Our evaluation process checks include:

Plan correctness: Assessing whether the ‘orchestrator’ agents generate a sensible and effective breakdown of subtasks.

Execution path: Evaluating if the task trajectory through the system follows an optimal path.

Output accuracy: Conducting detailed evaluations to verify that outputs align closely with our curated expert references.

Thanks!

—

Definely Webinar: Success Tactics – Change Management + AI

On June 17, Definely and Artificial Lawyer are partnering to bring you a free webinar all about achieving success in AI and legal tech deployment, with change management at the centre of the discussion.

Speakers will include:

Jayanth Poorna – Partner, Culture and Change practice, Korn Ferry – As an engineer by qualification, a software developer by training and a specialist in culture, leadership and change, Jay brings the unique ability to see technology, operating models and transformation through the lens of human experience.

Molly Taylor – Head of Customer Success, Definely – who leads the Customer Success team at Definely, where she is dedicated to ensuring clients achieve measurable value from Definely’s legal products. Before joining Definely, Molly held roles at DXC Technology, Workshare, Litera and Legl.

Panel chair – Richard Tromans, Founder, Artificial Lawyer.

Plus – a mystery guest….and yep, I don’t know who it is either 

To register for the live webinar on June 17, at 3PM UK time (BST) and 11AM ET, please RSVP here.

–

In the webinar we will cover:

1 – Why Legal Tech Initiatives Fail — and How to Set the Stage for Success. Covering misaligned expectations, lack of engagement from senior partners or GCs, and fears around job displacement.

2 – Understanding and Demystifying AI for Legal Teams. Deep diving into what AI is (and isn’t) in the legal context, as well as concerns around reliability, ethics, hallucinations, and explainability.

3 – Creating a Change Management Playbook for Legal Tech. Talking about how to map stakeholders, run pilots, create champions, and structure a phased rollout to capture hearts and minds.

4 – Driving Adoption and Sustaining AI Momentum. Showcasing practical steps to build excitement, confidence, and credibility over time.

And more….

It’s free to attend, but please RSVP here. 

—