A narrow firm, on purpose
When I founded Ironbridge Legal, I made a choice that puzzled some advisers. The firm would act in commercial disputes, insolvency and restructuring, and regulatory enforcement. Nothing else. No deals, leases, or front-end drafting of any kind. If a client asks us to paper a transaction, we send the work to another firm, often one that sends contentious work back to us.
People assumed the narrowness was a constraint we would outgrow. It is the opposite. It is the strategy. A firm that only fights is never a competitive threat to the firms that transact. That makes us a safe pair of hands for conflict referrals from the largest firms in the country, for accountants whose clients hit trouble, and for foreign lawyers who need Australian counsel without a turf war. The narrow position built our referral economy. It also built something harder to copy: a bench that does nothing but contested work, every day, across Australia.
Maister still holds
David Maister’s Managing the Professional Service Firm is now more than thirty years old. I’m getting my leadership team to read it now.
Maister sorted professional work into three kinds. First, expertise work, which he called brains: new, high-stakes problems where the client buys creative thinking. Second, experience work, or grey hair: problems the firm has seen many times before. Third, efficiency work, which he called procedure: well-understood problems where the client mainly buys reliable output at a sensible price.
Most firms slide down that ladder without ever deciding to. Procedure work is easy to leverage. You can put rows of junior lawyers on it. The economics seduce, and the brand then follows the work. Maister’s warning was that a firm cannot be all three things at once, because each demands a different shape: different leverage, different pricing, different people, and different marketing.
We took the warning at face value. Disputes, insolvency and regulatory defence sit at the brains and grey hair end of the ladder, and we stay there. That choice fixes the shape of the firm. Senior-heavy teams. Pricing that reflects outcomes and judgement rather than hours of process. It also fixes who we hire: lawyers who want to run hard matters, not manage volume.
Then the machines arrived
Here is the part Maister could not have foreseen. Generative AI is collapsing the procedure layer of legal work. The tasks firms have long used to feed the leverage model, first drafts, document review, summaries, research memoranda, chronologies, are exactly the tasks the new tools now do quickly and cheaply.
We use these tools every day in our workflow alongside established research platforms. They build a working chronology from a large document set in hours rather than weeks.
What they cannot do is the product. They cannot take responsibility for advice. They cannot weigh a witness, read a bench, judge when a regulator is signalling a way out, or decide that a commercial settlement today beats a forensic win in two years. They cannot hold a director steady the night before a freezing order hearing. Judgement, accountability and trust are the product. The documents were only ever the packaging.
For a procedure firm, AI poses a hard question about margin, because the thing being sold is the thing being automated. For a judgement firm, AI is leverage of a new kind. Maister’s old profit equation, built on rates, utilisation and leverage, still holds. Only the meaning of leverage has changed. It no longer refers to administrative and base-level work for junior lawyers. It means a smaller team focused on the work only elite lawyers can do, while the machines absorb the rest. It is a true win-win case of efficiency, enjoyable work, and client satisfaction.
Guard rails before glory
Australian courts have now issued practice notes and guidance on the use of generative AI in litigation, and recent decisions across several jurisdictions show what happens when lawyers file unverified machine output. Responsibility lands on principals, where it belongs.
So our rules are strict and dull. Every citation is checked against the judgment itself before it leaves the building, pinned where possible to paragraph numbers. No AI output reaches a client or a court until a named partner has read it, tested it and owned it. Client material goes only through enterprise deployments with proper confidentiality terms, never through open consumer tools. We treat the tools as brilliant, tireless juniors with no practising certificate, no insurance and no shame. Useful, and never left alone with the client.
Trust is the asset that never appears on the balance sheet
Maister’s later work with Green and Galford, The Trusted Advisor, says the plain thing many firms forget. Clients in trouble do not buy documents. They buy confidence in their adviser’s judgement under uncertainty. A director facing a public examination, a fund whose portfolio company is burning cash, a board served with a regulator’s notice: none of them wants a memorandum. They want to know what to do by Friday, and they want to believe the person telling them.
Our business development is therefore built on teaching rather than selling. The firm reads the court lists every morning and publishes what matters, written each time for one reader: the liquidator, the general counsel, the foreign lawyer briefing Australian counsel for the first time. A piece written for everyone persuades no one. There is a useful twist here. The habits that make writing valuable to a busy human reader, plain English, a direct answer in the first paragraph, a named author willing to take a view, are the same habits that make it valuable to the AI assistants those buyers now consult first. Good writing has become good distribution.
The apprenticeship problem
The hardest management question AI has put to us is not about clients. It is about juniors. If the machine writes the first draft, where does a young lawyer learn judgement? Judgement was always learned by doing the lower-order work badly, then less badly, under close supervision.
Our answer is unfashionable. Juniors still do the thinking. They may use the tools, but they must reach their own view before they see the machine’s, and they must be able to defend any difference between the two. They come to court, and sit in client meetings. Supervision is not a compliance afterthought; it is the production line for the only asset we sell. A firm that lets AI hollow out its training programme has quietly agreed to stop making partners.
What I would tell another founder
Decide what you will not do, and write it down. The narrowness is the moat. Reread Maister before you buy another piece of software; the strategy questions have not changed, only the tools. Adopt AI early and govern it hard, because the firms that wait will inherit both the disruption and none of the fluency. Price judgement rather than time wherever the matter allows. And measure trust, not just lock-up and utilisation: referrals received, repeat instructions, the call that comes at eleven at night.
The firms that do well over the next decade will not be the ones with the most software. They will be the ones whose names come to mind when a sophisticated buyer is in real trouble and wants a human being to carry the weight with them.
