top of page

What a Year Studying AI Taught Me

1 day ago
9 min read

Updated: 7 hours ago

After a year immersed in learning about artificial intelligence, I arrived at an unexpected conclusion: the most important challenges were rarely technological. More often, they lay in the fundamentals of how we capture knowledge, share information, make decisions and operate as a business.


Robot teacher in red jacket points to a chalkboard reading What a Year Studying AI Taught Me, with brain and lightbulb icons.

At the start of this year, I asked my team to do something that felt increasingly out of step with the mood of the moment. Before we committed serious money, time or organisational energy to artificial intelligence, we would first take the time to understand it.

Not through headlines. Not through product demonstrations. And not through the endless stream of LinkedIn posts confidently predicting the future of legal services and the imminent demise of junior lawyers. Instead, we would approach the subject with curiosity and an open mind. We would invest the time to understand what these technologies could genuinely do, where their limitations lay, and what they might mean for our lawyers, our clients and the way we run our firm.

If that sounds unremarkable now, it certainly did not feel that way at the time.

By then, the legal profession had largely stopped debating whether AI mattered. The question was no longer if, but how quickly firms could embrace it. New products appeared almost weekly. Major firms announced partnerships with technology providers. Commentators confidently predicted a fundamental reshaping of legal practice. Clients, meanwhile, were increasingly asking what we were doing about AI and what it might mean for the way legal services would be delivered, staffed and priced in the future.

Like most business leaders, I felt the pressure to keep up.

I also felt an increasing sense of unease.

I am the first to admit that I am not a technologist. But I have enough lived experience to know that organisations rarely get into trouble because they spend too much time understanding a problem. More often, they get into trouble because they convince themselves they already understand it.

So rather than rushing to select a platform or announce an AI strategy, we spent much of 2026 back at school.

Most valuable of all was the scrutiny we applied to our own business.

What started as an effort to understand artificial intelligence gradually evolved into something else: a candid assessment of how our firm actually worked. The deeper we explored the technology, the more we found ourselves examining our processes, our data, our decision-making and some of the assumptions that had accumulated within the business over time.

A Different Starting Point

Part of the reason we approached AI cautiously lies in the way our firm, Support Legal, was built in the first place.

The firm was founded on a relatively simple belief: that experienced lawyers can do some of their best work outside the confines of the traditional law firm model. Most of our lawyers spent years building successful careers at leading international and Magic Circle firms before choosing a different way of practising. In many cases, they were seeking greater autonomy, flexibility and balance, but not at the expense of professional ambition. The objective was never to step away from high performance. It was to create an environment in which experienced professionals could sustain it for longer and on their own terms.

That philosophy continues to shape how we organise the business. We are deliberately lean, not because efficiency is an end in itself, but because we believe clients are best served when they have direct access to the lawyer they chose to instruct. Our model is built around experienced practitioners delivering advice, judgement and solutions themselves, rather than supervising layers of people beneath them. Clients come to us for expertise rather than leverage.

Viewed through that lens, AI raised a different set of questions from the ones dominating much of the legal industry.

Many larger firms were understandably focused on how artificial intelligence might affect leverage, utilisation and the traditional pyramid structure on which much of the legal services industry has historically been built. Our questions were different. Could technology allow a relatively small team of experienced lawyers to handle larger and more complex matters? Could it reduce administrative friction and free up time for higher-value work? And perhaps most interestingly of all, if artificial intelligence was making information cheaper and more accessible, would that ultimately make experienced judgement less valuable or more?

We did not know the answers to those questions at the outset, and we are still working through them today. Yet as the year progressed, I found myself increasingly persuaded that AI may ultimately reinforce some of the very qualities that attracted both our lawyers and our clients to the firm in the first place.

Start With the Problem, Not the Demo

The demonstrations were extraordinary.

These platforms could review contracts, identify risks, answer legal questions and produce competent first drafts in seconds. While the quality varied, the underlying capability was difficult to ignore. What felt experimental not long ago is increasingly becoming part of mainstream professional practice.

Yet I often left demonstrations more uncertain than when I arrived.

The problem was not that the technology seemed limited. Quite the opposite. It appeared capable of almost everything, and with every new demonstration it seemed capable of more. Every vendor offered an impressive range of use cases. Every platform came with a compelling story about how it would transform the way lawyers work. The result was a familiar problem: abundance made decision-making harder rather than easier.

One of the most useful disciplines we developed over the year was forcing ourselves to begin somewhere less exciting. Rather than starting with the technology, we started with the frustrations. Where did matters lose momentum? Where did clients experience unnecessary delays? Where were experienced lawyers spending time on work that did not require experienced judgement? Where were people compensating for weak processes through effort and institutional memory? Where was valuable knowledge trapped inside individual heads or desktops, rather than embedded in the firm itself?

None of these questions were particularly sophisticated. That was precisely why they proved so useful.

The more clearly we articulated the problems, the easier it became to assess potential solutions. Conversations with vendors changed almost immediately. The question was no longer, "What can this technology do?" It became, "Which of these problems does it solve, and how would we know?"

That distinction sounds subtle. In practice, it changes almost everything.

Technology evolves quickly. Product features change monthly. The fundamental problems facing a business tend to move much more slowly. A firm that steers by features spends its life chasing the next release. A firm that steers by problems has a much more reliable point of reference.

The longer we spent evaluating AI, the more convinced I became that technology strategy is rarely about technology alone. The real challenge is developing enough clarity about your business to recognise which opportunities genuinely matter when they arrive.

Can Doesn't Mean Should

Knowing your problems is only half the challenge. The other half is understanding whether a proposed solution actually creates meaningful value.

One of the easier mistakes to make in any AI discussion is to assume that capability and value are the same thing. They are not.

A tool that saves a lawyer twenty minutes on a task performed twice a month may be technically impressive and commercially irrelevant. The true economics of technology adoption include licence fees, implementation effort, training, governance, supervision and ongoing review. In legal services, they also include professional responsibility. Human oversight is not optional. Nor should it be.

These conversations led us into questions that many professional services firms are still grappling with. If technology genuinely allows lawyers to deliver work more efficiently, what does that mean for businesses traditionally built around billable time? How should value be measured? How should clients benefit? How should firms think about pricing?

We are still working on the answers to all of those questions.

But I am convinced that adopting technology simply because it exists is no more sensible than rejecting it because it is new.

Every AI Project Is a Data Project

One of the most useful concepts I encountered during the year was the distinction between generative AI and analytical AI.

Like most lawyers, I initially focused almost entirely on generative AI. The ability to draft documents, summarise information and generate first drafts naturally attracts attention in a profession built around language and written outputs.

What became apparent quite quickly, however, was that the quality of those outputs depends entirely on the quality of the information available to the model. AI can help draft a contract, but only if it can access the right precedents. It can analyse trends, but only if the underlying data has been captured consistently and can be retrieved easily.

Over time, I found myself increasingly fascinated by the information beneath the drafting. Which clients create the most value? Where do matters consistently exceed budget? Why do certain projects stall? How accurate are forecasts? How much of what we think we know about the business can actually be measured?

Those questions turned out to be far more challenging than asking a language model to draft an email.

Like many professional services firms, we discovered that years of potentially valuable information were scattered across systems, inboxes, spreadsheets, document folders and individual memories. The information existed. Extracting insight from it was another matter entirely.

The experience reinforced a lesson that extends well beyond artificial intelligence. Organisations often underestimate the value of good information management because talented people compensate for its absence. Technology has a habit of exposing those weaknesses.

For AI to be effective, documents and data must be organised, consistently named, clearly labelled and stored in a way that makes them accessible to both people and machines. Knowledge trapped in personal inboxes, ad hoc filing systems or individual memories is difficult for AI, and often difficult for colleagues, to use.

In hindsight, becoming more AI-enabled was as much a document and knowledge management project as a technology project. Sophisticated AI layered on top of fragmented information simply produces sophisticated mistakes. The organisations that benefit most from AI will not necessarily be those with the best tools, but those with the best information.

Bringing People With You

If data was the lesson I least expected, people were the lesson that mattered most.

Technology can be purchased. Adoption cannot.

Our lawyers are experienced, independent professionals. Many joined Support Legal precisely because they wanted greater autonomy over how they work. Any technology capable of improving their effectiveness would likely be welcomed. Any technology that sought to force them into rigid processes would need to prove its worth.

That reality pushed us towards a different perspective.

The most successful examples of AI adoption that I encountered treated technology not as a substitute for expertise, but as a complement to it. The objective was not to replace judgement but to amplify it.

That distinction feels particularly important in law.

Artificial intelligence is making information dramatically cheaper, faster and more abundant. Judgement remains scarce.

Clients rarely object to paying for expertise. What frustrates them is paying experienced professionals to perform work that does not require their experience. If AI absorbs more of that activity, the value of experienced judgement may become clearer rather than less relevant.


Trust Is Non-Negotiable

If there is one principle that remains entirely unaffected by technological progress, it is trust.

Law firms hold some of the most sensitive information their clients possess. Confidentiality is not simply one professional obligation among many. It is the foundation upon which the profession is built.

Unsurprisingly, many of these discussions quickly shifted from the capabilities of AI to the governance surrounding it. Questions of data handling, accountability and oversight came to the fore. Where is information stored? Who has access to it? How is it used? What controls and safeguards exist?

These are not peripheral concerns. They are central concerns.

The practical answers are rarely glamorous. Vendor due diligence. Governance frameworks. Policies. Training. Accountability. The more we learned, the more convinced I became that strong governance is not an obstacle to innovation but one of the conditions that makes responsible innovation possible.

Clients are entitled to ask not just whether their advisers use AI, but how, and more of them do.


Asking Better Questions

At the beginning of the year, I thought we were embarking on a technology project.

Looking back, I am no longer sure that is what happened.

The most valuable lessons had surprisingly little to do with software, language models or product demonstrations. They had far more to do with leadership, organisational learning, information management and the enduring importance of judgement.

We have spent the year learning, experimenting and testing. Yet our most important conversations today are not about what AI can do, but what it should do for our clients, our lawyers and our business. The focus has shifted from features to outcomes.

As we approach 2027, we are consolidating what we have learned, gathering feedback and incorporating those lessons into our strategic planning. That work feels every bit as important as evaluating the technology itself.

The technology will continue to evolve. New products will emerge. Some of this year's assumptions and conclusions will prove wrong. I would rather discover that through thoughtful experimentation than an expensive commitment.

After a year immersed in learning about artificial intelligence, I arrived at an unexpected conclusion: the most important challenges were rarely technological. More often, they lay in the fundamentals of how we capture knowledge, share information, make decisions and run the firm.

We began the year looking for answers about AI.

We end it with better questions about our business.

That feels like time well spent.


 This article is provided for general information purposes only. The views and opinions expressed are those of the author and do not necessarily reflect the views of Disclosure.



Author

Jamie Tredgold

Managing Partner of Support Legal.




Comments


bottom of page