Getting started with AI – Your New Year’s Resolution?

Piece of paper with New Year Resolutions written on it

Social and mainstream media is overflowing with stories about developments in artificial intelligence. Who is buying which tools, which companies are being bought up by which companies, claims as to how the world as we know it will change beyond recognition and an underlying sense of antsy anxiety that everyone else is further ahead in their journey than you and you’ll be irreparably left behind. Feel familiar? With this frantic backdrop, there’s a danger that AI goes into the “Too Hard” bucket and you’re paralysed into inactivity, not knowing where to begin. This article is intended to give you a moment to take a breath and formulate a plan. There were some 2,000 AI tools launched in a 30-day period during last summer alone so being methodical about how you’ll cut through the noise is imperative. And with the EU AI Act having come into force in August 2024 (with key provisions starting to bite during 2025/2026) which imposes new duties on those involved in developing, selling and using AI tools, this is a hotter topic than ever. These are some of the steps our law firm clients are taking to get their efforts off the ground.

Educate

Part of the anxiety around AI is that most lawyers aren’t entirely sure how it works or what it really means for their day-to-day practice or their firm. There is post-pandemic change fatigue where many lawyers can’t contemplate getting their heads round another new system, however whizzy its capabilities or how it could purportedly save them heaps of time. There are over 8,000 AI tools out there and keeping track of what is both available (reading beyond the hype as to whether a tool actually is “AI-assisted” and what that means in practice) and commercially viable is tricky. Part one of your plan should involve skilling up your people on what you’re dealing with, what potential systems are out there, what potential use cases exist that seem to resonate with a problem you are trying to address in your business and what your existing vendors are doing. This could mean teaching sessions across the business, or it could mean focussing the training on your core people charged with exploring how AI can work for your business. Both will probably be necessary in the longer term. There is also an “AI literacy” duty in the EU AI Act (Article 4) which firms using AI tools who are caught by the scope of the legislation will be under to train up their people as to how the tools they are using work.

Strategise

If you haven’t already done so, set up an “AI Working Group/Task Force” comprised of different voices from around your business. This must include lawyers from your transactional, advisory and contentious practices, AI enthusiasts and sceptics, experienced and junior practitioners, IT, Risk, Business Development/Marketing (for the client voice to be heard) and other business services professionals and, if you have them, representatives from your Knowledge team. However informal your knowledge management activities, it’s vital to involve those people who wrangle your templates, data, current awareness/research tools and access to your firm’s collective wisdom. Having actual decision-makers on this committee is also important so that you can be agile to action decisions when you reach conclusions on how your firm should address the challenge of implementing AI.

This group needs to have a clear understanding of the key strategic goals of your business. Whether these be growth, revenue/profitability targets, people retention and development or reducing the burden of non-chargeable time on client-facing lawyers; all the typical strategic missions we see reflected in firms’ strategies could be influenced by the use of AI tools. Without a clear understanding of what the firm is trying to achieve, it is impossible to know where to focus in the crowded AI marketplace.

Policy

One of the first tasks of the AI working group will be to consider issuing a policy statement to the firm (and potentially a more public statement about the firm’s approach to AI if you feel your clients and competitors should be aware of your commitment to exploring how the biggest tech news story of our time may affect your services). Even if it feels like a “holding statement” which you later develop, those concerned with managing risk issues at the firm will wish to set out in clear terms for all staff where they may and may not use AI-assisted tools at work and on firm devices. These range from policies which prohibit the use of publicly available AI-assisted tools (such as ChatGPT) on firm devices and with client/matter data (including prompts which may contain traceable client information). Others encourage experimentation with certain sanctioned tools which have been developed and/or tested by the IT team for use inside the firewall but with restrictions on use (e.g. only after receiving training on the risks and limitations of the tools, flagging up copyright and data protection concerns, warnings about not using the output without a human critically reviewing it before it is circulated, etc.).

Plan

When your AI working group has a handle on the art of the possible and what the priority strategic drivers of the firm are, you can start to formulate a plan for how to move forward. Your aim is to develop a roadmap of actions your firm will take over the forthcoming 6 to 48 months to take forward your ideas about how AI could be used. This may include gathering ideas from staff as to things which currently make them want to throw their computer out of a window, things which feel basic and people feel should just work better than they do or things which they have tried in their non-work life which they feel could help make a task or process easier to manage in the office. Lawyers aren’t typically very good at completely “blue sky” thinking so this is not about reimagining the entire practice of law or your workplace. It could be about small incremental changes which, when multiplied across your staff each day, add up to real time savings or incrementally better client service. Of course, if thinking outside the box feels hard, you can engage with an expert “Design Thinker” to kick-start your ideas mill and try to get your people imagining what your practice and clients will need from you in five years’ time. We have run workshops with firms across their legal and business services teams to consider where opportunities might exist to improve processes and gather ideas so we can help with this if required.

Many firms have found that, when gathering ideas from staff about what potential uses AI-assisted tools could be put to, the vast majority of them are things which are already available in tools which are not-AI assisted and which are often already available in the firm’s existing tech stack! There is a big issue in many businesses with gaining “business as usual” adoption of tools by time-poor lawyers. It’s imperative when triaging use case ideas to examine the functionality of existing tools and consider whether refresher training is a first step to gaining efficiencies without distracting yourselves with a deep dive into AI-assisted tools which may not provide satisfying answers, nor a return on investment.

At 3Kites our motto is “It all comes down to people in the end” and technology (let alone AI-assisted technology) is not always the answer to a business problem – we advocate starting with the question, not the (AI) answer. Solutions can often be found through tweaks to how your people carry out a process or by having a different person carry out a task so not all initial suggestions from staff or your AI working group will prove ripe for the application of generative AI tools.

Avoid demo fatigue

When you have isolated some ideas for things you would like your firm to be better at doing which you feel a generative AI tool may be able to help with, the daunting prospect of finding the right tool in the crowded marketplace begins. Avoiding “demo fatigue” from sitting through canned presentations from vendors whose tools are not yet mature enough for reliable use or are not on general release yet can feel like a drain on the time of your AI working group participants or IT/Knowledge Management (KM)/Innovation teams. Our top tips for managing the stream of vendor approaches are:

Speak to your existing vendors to ask about features they may introduce which could assist with the issues you are trying to solve. Your people may already be familiar with the systems and so the change management piece could prove easier when introducing users to new features.

Whether it is a new vendor or an existing one who is coming to demo their tool to you, ask in advance for them to show you how their tool would perform in executing your specific use case or improvement you are seeking to make rather than just give you their canned demo with their shiny demo-ready data set. This can help you cut through the sales pitch to hone in on whether their tool would really address the issue you are trying to solve and whether that is available right now or further down their roadmap.

Ask the vendor to speak candidly about what you need to do in practice to your data to ensure that the tool works as demo’ed. Many firms do not have up to date, clean, consistent or tagged/categorised data or existing standard templates/agreed clauses for legal documents in their systems and some generative AI tools can only work at their best and give best results when some of these data issues are ironed out. That could be a big piece of work to undertake so you need to be aware of what would be required so you can plan your roadmap of activities.

KM fundamentals

Many businesses are waking up to the limitations of their knowledge landscape and how they may hamper efforts to deploy generative AI effectively at their firm. In practice, this could mean not having robust processes for creating firm standard templates or clauses and keeping those reliably up to date, or not having a curated collection of up-to-date example documents (rather than the whole of a firm’s document management system) over which a tool could be deployed to find wording or current market positions on particular legal points. It may mean that information architecture challenges have not been met so systems do not integrate effectively to allow data to flow from different systems (like the client relationship management and practice management systems). Cries of needing robust “IA (information architecture) before AI” are being heard. Those with responsibility for knowledge management at your firm (however informal their roles) may become curators of libraries of examples of prompts used to ask generative AI tools to perform tasks. This service is designed to try to help improve the consistency of outputs from the tools and to save staff time when seeking to experiment with tools by preserving and sharing the fruits of the testing which colleagues have performed.

Monitor

Your AI working group will wish to consider how it will tackle staying up to date with the market and keeping a watching brief on tools which may not yet perform well enough for the firm to invest in but which may well mature helpfully in the coming months. Enlisting library and information staff in helping to keep a watching brief on developments and engaging regularly with existing vendors on how their roadmap plans are evolving are good places to start – much of this is the art of the (commercially viable) possible. It may also be important to encourage staff members to raise ideas they see outside the office for discussion for business applications. This effort takes a village.

Change and expectation management

Any projects you roll out in relation to generative AI tools are not just technology projects, they are change projects. Skimping on the change management planning and budget will result in patchy adoption, frustration and a lack of return on your investment in the tool. Managing sceptical lawyer expectations about the use of generative AI tools is also fundamental to success. Even when you are asking lawyers to test/pilot tools or functionality, you need to message very carefully whether the tool is intended to simply assist them with a part of a process or whether the tool is intended to remove the need to perform a task entirely. A mismatch in these expectations can be very damaging to the future roll-out of that and other tools if your staff over-expect and the tool under-delivers. Lawyers are not known for their infinite patience and can jump to unhelpful conclusions when tools do not produce consistent or 100% accurate results. Framing experiments as such and as part of a “pilot” programme or broader journey, can be helpful.

Ultimately, generative AI is still maturing as a technology and firms often worry that they are not doing enough to implement it or to plan for its use in their business because they don’t feel like they have “solved” how it will work. The truth is that no one has this insight or can say how legal practice will look in five years’ time. Even for those who are experts steeped in developing generative AI tools and know their potential inside out, there is still the issue of trying to deploy change in a law firm to contend with! Everyone is still getting to grips with what generative AI should be deployed on, what these tools are actually good at (or better than humans at) and which are reliable and accurate enough to gain acceptance by sceptical and risk averse lawyers and their clients to be genuinely helpful and provide long-term benefits/RoI. We are not there yet by a long chalk so do not despair if you feel your firm has not yet embraced generative AI in practice. It’s a great time to get started, to take your project to the next level or to put plans into action… we are here to help!

Jenni Tellyn is a recovering debt capital markets and structured finance lawyer, former Knowledge Lawyer and Head of KM. She leads the Knowledge Management practice at 3Kites, is a regular speaker on KM and AI topics and has helped several firms with getting started on their AI journies. Check out her bio on our website here: https://www.3kites.com/jennitellyn.