For years, legal technology has sat beside practice rather than within it. A firm might have a case management system, an electronic library, and a billing platform, but the real work still felt stubbornly human. Pleadings were drafted in Word. Advice was shaped in a barrister’s head. Costs budgets were built by hand, or something close to it. Artificial intelligence has so far lived in much the same territory: a separate login, a deliberate choice to “ask the bot”, and then a return to proper work.
That separation is unlikely to survive the next three to five years.
The most important point is not that AI will arrive with a fanfare. It is that it will disappear into the tools lawyers already use. Word, Outlook, Teams, bundle platforms, billing packages, case management systems and costs software will all acquire generative features as part of their ordinary furniture. You will open a bundle and find a draft index already prepared. You will dictate a note of evidence and be handed a chronology cross-referenced to the documents. You will begin drafting particulars of claim and be offered authorities in the margin. This will not feel like a special act of innovation. It will feel like a software update.
That matters, because lawyers are not natural early adopters. Most will not abandon familiar workflows for the thrill of novelty. The software industry has grasped that, and so the route in is obvious: do not ask lawyers to go to AI, put AI into the places where lawyers already live. The result is that many practitioners will find themselves working in an AI-augmented environment without ever consciously deciding to do so.
The consequences will be immediate. Tasks that once absorbed hours of drafting, formatting, checking and tidying will shrink to a fraction of their former length. A piece of work that previously took three hours may take one. That is an obvious gain for clients, who will receive quicker turnarounds, cleaner documents and fewer avoidable errors. It is also an obvious problem for a profession still built, to a remarkable extent, on the billable hour. If the same work takes half the time, then the old arithmetic begins to look fragile. Clients, especially sophisticated clients, will not willingly pay for hours the machine has erased.
That is why the billable hour, though it will not vanish overnight, is likely to come under increasing pressure. The likely direction of travel is towards fixed fees, capped fees, subscriptions and value-based pricing. In truth, this is not a matter of fashion. It is a matter of economics. Once automation strips away a large amount of routine labour, fees can no longer sensibly be anchored to the time once spent on mechanics. The parts of the job that retain value are judgment, strategy, persuasion and client management. Those are the things for which lawyers will increasingly be paid.
A second change will come from litigation analytics. At present, such tools are useful in a limited way, but they remain patchy. One can obtain rough statistics about courts, judges, settlement ranges or the duration of claims, but the scope is narrow and the quality uneven. Over the next few years that should improve. The technology will become better at drawing together what is presently fragmented: filings, judgments, costs awards, budgets, interlocutory rulings, and perhaps even anonymised ADR outcomes. The result will not be prophecy in the mystical sense, but it may well become prediction in a commercially useful sense.
That will alter the economics of disputes. If a solicitor or funder can obtain a quantified forecast of likely duration, likely damages, likely recoverability of costs, and the probable range of settlement, that changes how cases are priced and run. It affects whether a claim is worth bringing at all. It affects how Part 36 offers are pitched. It affects how litigation funders assess risk and return. It affects how costs budgets are set, and how far it is sensible to spend on particular phases of a case.
Costs lawyers will feel this change keenly. At present, much of costs practice still operates in a world of partial visibility. Published authority is incomplete, assessments are unevenly reported, and reliable benchmarking is often difficult. Better analytics should narrow that uncertainty. A system able to say, with reasonable confidence, that in a given court before a given judge one can expect recovery at a certain level, or that certain categories of dispute attract a predictable degree of reduction, would alter negotiation from the outset. It would not abolish argument, but it would narrow the field of honest dispute.
A third development is the likely emergence of more domain-specific legal AI. General-purpose large language models are impressive, but they remain generalists. Their weakness is that they are trained on almost everything, and therefore have no natural respect for legal hierarchy, procedural nuance, or regulatory boundaries. The next phase will be narrower and more serious: systems built for law, trained on closed and curated bodies of material, and designed to work within professional constraints.
That matters because the real problem with current public tools is not merely that they sometimes err, but that they err with confidence. A legal system trained on statute, case law, procedural rules and approved materials should reduce that tendency. It will not eliminate mistakes, and no sensible lawyer will surrender professional judgment to a machine, but it will make it more realistic to delegate first-pass tasks: initial research, early issue spotting, draft structures for argument, and the first shaping of documents.
The economic effect is plain enough. Fewer junior hours will be spent on first-pass work. Human effort will move upwards, towards supervision, selection, judgment and strategy. That is efficient, but it also raises awkward questions about training. For generations, junior lawyers have learnt their craft in part through the very drudgery that AI now threatens to consume. If the routine work disappears, the profession will have to think harder about how experience is acquired.
A fourth change will come from the client’s side of the table. Clients will not remain passive while lawyers automate. Corporates, and in time private individuals too, will begin using legally trained assistants of their own. They will use them to prepare instructions, assemble documents, summarise problems, and run preliminary checks before approaching a solicitor or barrister. The first meeting will therefore change. Instructions will often be cleaner. Fact-gathering will be more complete. Some of the billable triage that once occurred at the beginning of a matter will diminish.
That is, plainly, a threat to one revenue stream. It is also an opportunity. Firms and chambers can develop their own branded intake systems, designed around their own practice areas and their own methods. Instead of letting outside platforms shape the first contact with the client, they can keep that process in-house. The practical attraction is obvious: lower administrative overhead, better compliance, more predictable pricing, and an easier route for clients to engage directly.
Lastly, cost management itself is likely to become more live, continuous and unforgiving. At present, budgets are often treated as static artefacts: produced at one stage of the case, argued over, and then too often left behind while the litigation moves on. AI systems should change that. They will reconcile work in progress against budget phases in real time, flag likely overspend before it occurs, and suggest corrective steps. In costs disputes themselves, they will analyse bills, points of dispute, replies and prior outcomes, and give an indication of probable results.
That will make inefficiency harder to defend. It will also make value for money easier to demonstrate where work has been lean, targeted and proportionate. In that sense, AI will not merely speed up legal work. It will expose it. That may be uncomfortable, but it is not necessarily a bad thing.
The broad direction is clear enough. Artificial intelligence will not abolish the lawyer’s role. It will strip away much of the routine that has long surrounded it. What remains at the centre will be the things that machines still cannot truly do: judgment, persuasion, responsibility, tact, and the management of human affairs. The change will not arrive with trumpets. It will come quietly, through routine updates and altered habits, until one day the old way of working simply looks old.