Artificial intelligence (AI) has recently become an operational reality for the legal profession, and it has arrived with the character of a double-edged sword. For some firms, it looks like a force-multiplier. For others, a new category of risk with unclear boundaries. Most are somewhere in the middle, unsure whether to lean in, hold back, or do both at once.
That uncertainty is now being pushed into the open by clients. More clients are asking for clarity on how AI is used, whether it affects confidentiality, and what it means for the quality of advice. Some want reassurance that AI is being leveraged without compromise to strategic insight. Others want a more direct explanation of what ‘using AI’ actually means for their matter and for the profession more broadly.
Trevor Withane has been willing to talk about the practical reality, not the headline version. Not the simple story of subscribing to a big-name platform and declaring the firm ‘AI-enabled’, but the harder question: can the technology be embedded into legal work in a way that protects standards, and avoids turning a precision tool into a blunt instrument.
“AI is not something you bolt on and then announce,” Mr Withane says. “If you are not disciplined about how it is used, it will create new problems faster than it solves old ones.”
Not a trend to follow, a capability to build
At Ironbridge Legal, the decision to adopt AI was not about novelty, marketing, or chasing a news cycle. It was treated as a strategic investment in capability, made with the assumption that complexity and volume are not going down, and client expectations around responsiveness and clarity are not easing.
That matters because the conversation about AI often focuses on theory rather than what changes in practice. It is easy to debate principles. It is harder to build a system. The point is not whether AI exists, or whether it is ‘the future’. It is whether a firm can integrate it in a way that raises the floor on quality and consistency, without lowering the ceiling on judgement.
Ironbridge’s approach started with a simple premise: a tool is only as safe as the framework around it. The value is not in the software alone. It is in governance, training, supervision, and the discipline to define what AI can be used for, what it cannot be used for, and what must always remain lawyer-led.
“We did not want speed at the expense of control,” Mr Withane says. “The goal was to expand capacity without diluting standards, so more clients can access our service with the same level of care and accountability.”
Competing at the top end means investing at the top end
There is also a commercial reality that many firms avoid saying. Used properly, AI has the potential to narrow the gap between elite specialist boutiques and larger national or global firms, particularly in document-heavy disputes, investigations, and restructuring work where time and pattern-recognition matter.
But that benefit does not come from half-measures. The most capable tools are expensive and resource-intensive, and meaningful adoption requires investment not only in licences, but also in secure environments, usage controls, and internal capability-building.
“With AI, doing it halfway is often worse than not doing it at all,” Mr Withane says. “If you want the upside, you have to commit to the safeguards, the training, and the discipline that make it reliable.”
Importantly, the investment was not framed internally as ‘automation’. It was framed as uplift: a way to reduce friction in parts of legal delivery that absorb disproportionate time, while strengthening consistency and analytical depth across matters.
What changes once it moves from testing to daily work
Once AI moved from controlled experimentation into everyday use, the biggest change was not that lawyers suddenly worked faster. The bigger shift was where lawyers spent their attention.
In most practices, there is a long tail of work that is necessary but expensive in time: summarising large records, building chronologies, extracting themes from correspondence, preparing structured issue lists, and reworking material into client-ready form. Done well, these tasks support judgement. Done repeatedly from scratch, they can drain it.
AI reduced that friction. It helped teams start from a structured base, freeing time for what clients actually pay for: strategy, risk assessment, and decisions under uncertainty. Implemented well, it should mean fewer people are spending billable hours on the mundane, repetitive work that neither juniors nor clients want to carry.
Special Counsel, Patrick Gallego put it this way: “It does not replace thinking. It changes where you begin. No junior wants to manually arrange 1,000 documents into chronological order, and no client wants to pay an hourly rate for that when AI can do it in seconds or minutes. But that does not mean you outsource the chronology to AI and move on. The ‘old-school’ process has a collateral benefit: it forces you to become deeply familiar with the record. The discipline is using AI to remove the grind without removing the benefits of the process.”
What did not change: standards and accountability
For all the hype around innovation, the core principles at Ironbridge were intentionally unchanged.
Responsibility for advice remains lawyer-led. Review processes remain intact. Confidentiality obligations remain absolute. Professional standards continue to apply in full. AI is treated as an assistant, not a decision-maker.
“Technology can assist, but it cannot be accountable,” Mr Withane says. “That line is non-negotiable. The lawyer remains responsible for the advice, the judgement, and the outcome.”
That principle has practical consequences. It informs governance settings, shapes usage rules and determines how the technology is introduced and supervised. It also affects how lawyers are trained. While AI can support research and analysis, learning remains human-led. Early-career lawyers still need to understand how advice is formed, how judgement is exercised and how conclusions are tested and challenged. That includes developing the instinct that comes from living in the documents, not only reading an output. In practice, AI can accelerate the first pass, but the lawyer still needs to verify, curate, and interrogate the source material, so the work product is not just faster, but better. Relying solely on AI outputs risks narrowing that development, particularly where tools are designed to be helpful rather than sceptical.
The same applies to culture. Legal culture is created through the behaviour and expectations of people, from partners through to junior lawyers and support teams. It is reinforced through supervision, discussion and accountability, not through technology. AI may accelerate aspects of legal work, but it cannot set standards, model judgement or instil professional discipline.
For Ironbridge, the point is simple. AI can make good lawyers more effective. It cannot replace the human responsibility that makes legal advice reliable.
What clients notice (and what they do not)
Clients rarely care about the tool itself. They care about service quality: responsiveness, clarity, confidence, and outcomes. When AI is used responsibly, it is often invisible to the client, but its benefits can be felt through tighter turnaround, clearer drafting, and lawyers who are more focused on the issues that matter.
Hyugo Hayashi, Ironbridge Legal’s Head of Growth and Marketing, describes it simply: “Clients do not ask about the technology. They notice that advice is sharper, communication is quicker, and lawyers are less distracted by process.”
This reflects the broader market reality. Clients do not want novelty, they want assurance that any new capability improves judgement and service, without introducing uncontrolled risk. Some clients also do seek clarity that the repetitive, time-consuming (and therefore expensive) tasks are being handled with appropriate AI support, and that they are not paying for manual work that technology can now complete faster. That expectation is likely to become more prevalent, and it reinforces the same point: transparency and discipline matter as much as capability.
The lesson for the profession: the blade is not the risk, the grip is
Ironbridge’s experience suggests that the real challenge in adopting AI is not choosing a product. It is deciding how seriously to commit to the discipline around it.
Governance matters more than features. Training matters more than enthusiasm. And accountability matters more than speed.
AI will keep evolving. The firms that benefit most will be those that treat it as a capability to build, not a shortcut to take. The objective is not to hand decisions to a machine. It is to support better human decisions, delivered with greater consistency, under real-world pressure.
Or, put differently: the question is not whether the samurai has a sharper blade. The question is whether the blade can be wielded without cutting the hand that holds it.
