Artificial Intelligence (AI) continues to reshape industries, none more cautiously than the legal sector. Drawing on our experience from AI implementations across professional services and our learnings from clients, this article explores organisational approaches, key challenges, and practical learnings that law firms and legal departments can act on now.
Decoding AI: Five key terms you should know
Understanding the different categories of AI is foundational to grasping where, how, and why to apply it. Terminology generally centres around five areas of AI technology. While they aren’t a sequential evolution of one to the other, they can be thought of as varying branches of the same family tree, inheriting some traits.
- Artificial Intelligence (AI): The broad umbrella term describing systems capable of learning. These systems combine data repositories, analytics platforms, and applications. A feedback loop that allows them to learn and update from experience distinguishes AI from traditional predictive models.
- Generative AI (GenAI): Think ChatGPT, Midjourney, Copilot, Gemini. GenAI generates content (text, music, video) based on patterns in its training data. It’s a powerful, creative assistant, but fundamentally reactive, waiting for user prompts.
- Retrieval Augmented Generation (RAG): RAG enhances GenAI by grounding responses in internal company data. It allows AI to generate answers with specific organisational context, like explaining your firm’s sick leave policy accurately.
- Agent: An agent performs tasks autonomously, like transferring data between systems based on predefined rules. It works through multi-step reasoning but within boundaries.
- Agentic AI: A step beyond agents, Agentic AI involves multiple agents that can set sub-goals and act independently. While powerful, this form raises valid concerns about decision-making boundaries and the need for strong oversight.
Each of these plays a distinct role. It’s less about one being better than another and more about choosing the right tool for the job.
Measuring the value of AI
The value of AI is realised, broadly, through three different metrics.
- Cost reduction: For instance, through automating tasks previously outsourced to third parties.
- Hiring deferral: Do more with current staff, delaying new hires.
- Productivity gains: Free time for creative, client-focused, or high-value tasks, or simply enable a better work-life balance.
What is interesting is that in addition to traditional return on investment (ROI) and return on experience (ROE), many organisations are starting to discuss return on future (ROF), which is an investment in readiness, even when the immediate payoff isn’t clear.
Though ROF lacks rigorous measurement, it reflects a willingness to experiment and evolve. That said, don’t let it become a vague excuse. My view is that ROF is important… but more as a strategic direction, not as a metric to track.
Applying AI in Legal: Case studies and lessons
As legal and consulting firms explore the transformative potential of AI, it is clear that success depends as much on people and processes as on the technology itself. These case studies highlight how organisations navigated early missteps, cultural resistance, and tooling confusion to unlock real value, often by learning from others' experiences. These stories offer actionable lessons for anyone looking to adopt AI in complex, knowledge-driven environments.
Case study 1: Enterprise ChatGPT rollout
Initially banned due to confidentiality concerns, this organisation pivoted to adopt an enterprise version. For a company billing via fixed-price models, GenAI freed up an average of 6.5 hours per week per consultant.
Successes:
- The firm created an internal AI Blackbelt community for peer support.
- Gamified adoption worked well with a competitive workforce culture.
Lessons learned:
- Licensing costs add IP: Only pay for what gets used. Evaluate existing tools like Microsoft Copilot before investing in standalone licenses.
- Executive training is crucial: When leaders lacked time to engage, training through Executive Assistants proved effective.
- Keep it simple: Multiple tools created confusion. Clarify use cases and streamline tooling choices.
- Avoid over-involvement: Too many stakeholders caused friction. Choose your project team wisely and maintain focus.
Case study 2: Knowledge finder tool
Designed to improve the reuse of consulting knowledge across SharePoint and Teams, this tool made finding existing content easier and more effective.
Successes:
- Clear strategy and business case from the start.
- Cross-functional collaboration (engineers, product, and change teams) led to better execution.
Lessons learned:
- Redundant tooling: The solution mirrored other available tools. Broader knowledge capability planning would have saved time and money.
- Tool overlap confusion: Employees weren’t sure when to use this vs. ChatGPT. Clearly articulate each tool’s value proposition.
- Cultural hurdles: IP ownership caused resistance -people were reluctant to share work if others could repurpose it. Knowledge-sharing culture must be addressed early.
- Trust must be earned: MVPs in AI need to go beyond the bare minimum. A poorly received tool damages trust quickly.
Three recommendations for measuring AI value
While these case studies focus on AI in professional services, the underlying lessons apply broadly across industries and technology initiatives. The same principles apply whether you're rolling out AI, cloud platforms, or other digital tools: success hinges on clarity, alignment, and adaptability. Applying best practices, such as defining success upfront, building a business case linked to strategy, and setting clear, actionable metrics, can help any organisation make smarter investment decisions, avoid common pitfalls, and scale what works.
- Clear metric indicators will help you understand when to scale if it works, when to pivot if it’s failing, and when to shut it down.
- Build a business case. Tie it back to strategic priorities to avoid project drift, leaking costs, and confused employees and customers.
- Know what you’re measuring. Be transparent about whether it’s ROI, ROE, or ROF, and avoid paralysis through measuring too much.
Choosing the right use cases for AI in professional services
Most organisations begin by crowdsourcing AI use case ideas. Once gathered, assess them against your AI strategy and create structured business cases to filter priorities.
Show your early wins by choosing small, impactful projects to build credibility. Foster trust by starting where AI scepticism is low and value can be demonstrated clearly. Use early projects to develop internal expertise and build capability.
Your organisation’s readiness matters. Some can handle a full-scale rollout while others should take a gradual, iterative approach.
Lookout for resistance to change
Now that you know the value you are aiming for and the use cases that will help you achieve it, you can begin building them out and deploying them, but watch out for resistance. At this point, organisations often discover their employees are not on board with change.
AI-related change management is often harder than traditional digital transformation because it affects everyone differently and broadly. Change fatigue is real, especially when tools and processes shift frequently. Employees are unsure what tools are safe, how to use them, and what the expectations are. Time and skill development are also barriers to adoption.
According to Gartner, only 36% of employees say they know how to use AI effectively. And 60% of AI initiatives fail due to unmanaged resistance.
Mitigate this by communicating your AI vision early and clearly, even if it evolves over time. Leaders must be actively onboard and teach people how to use it in their daily working activities. Consider a ‘day in the life of’ methodology to build awareness and understand the needs of stakeholders. Set company-wide OKRs and individual learning targets for AI that are tracked through management meetings and appraisals.
Overcoming legal sector-specific challenges
In the legal sector, trust and confidentiality are everything. Tools must be secure not only from external threats but also from internal misuse. If employees don’t trust the tool, they simply won’t use it.
Key strategies for the legal sector
Build trust over time with activities such as live demos to showcase confidentiality capabilities. Ensure early and continuous testing, celebrate wins loudly, and enable a champion network amongst other initiatives.
A big value add from GenAI and RAG is the ability to search and create knowledge from your own systems. We know there are problems with GenAI hallucinations, and in the legal sector, accuracy is, of course, vital. Until that accuracy can be assured, the human-in-the-loop concept by design is of paramount importance. Equip your employees with skills to critique responses in these early years of GenAI.
AI tools that work for your organisation need to be trained on your data and on legal-specific terminology. The data you need for this is often confidential, which represents a challenge in the fine-tuning stages and may require a heavy investment to get it right. Consider what use cases you tackle here, and also where off-the-shelf products have already done a lot of the hard legal knowledge work for you. When you do decide to build your own AI systems, the quality of your data and processes will become exceptionally important.
When it comes to ROI, define your vision and strategy, shortlist your use cases, understand the value they will bring you, and understand how you will track it. Remember, the data you need for this may not always exist and can be subjective. Is cost, employee experience, or future proofing the most important thing to your organisation, or perhaps it’s a mixture of all three? Your business case should tell you the answer to this.
These lists can become long, so select the use cases that will give you the value you need. This doesn’t necessarily mean the biggest return or tackling the biggest inefficiencies. It could be a small use case proof of concept to get leadership onboard, or an end-to-end process, automated to release capacity from a burnt-out team.
Resistance to change is a significant challenge. It can happen for a variety of reasons, so you must invest in change management to get it right, otherwise, your efforts will not generate the results you are aiming for. The legal sector has a reputation for slower change than some sectors, though innovation certainly exists, so resistance needs to be considered and carefully managed.
Conclusion: Strategic steps for the legal sector
The legal profession is understandably cautious, but the opportunity for AI-driven transformation is growing. Whether you’re at the beginning of your AI journey or building on existing initiatives, focus on three core areas:
- Start with purpose: Choose use cases that align with your strategy, culture, and risk appetite.
- Manage change relentlessly: Change fatigue, tool confusion, and trust issues can derail projects. Invest heavily in people, not just technology.
- Measure what matters: ROI, ROE, and even ROF can guide you, but clarity and consistency in your metrics are critical.
Above all, don’t wait for a perfect moment. Start small, learn fast, build trust, and scale with intent. For the legal sector, the road ahead is one of transformation, not disruption, if approached wisely.
