Proposal not a prophecy — When legal AI becomes default

Legal AI is not going to “arrive” one day. It may settle — quietly, by default — inside the tools we already use. And if that happens, the question won’t be which product you chose. It will be which platform chose you. In this article, Dr Corsino San Miguel proposes that this may already be the direction of travel.

When the Platform Eats the Vendor

Remember when you bought navigation? You walked into a shop, paid real money for a TomTom or a Garmin, clipped it to the windscreen, and treated it like a serious piece of kit. Then, almost overnight, navigation stopped being a product and became a feature — quietly bundled into the thing you already carried everywhere.

Smartphones plus free mapping apps did not merely “compete” with SatNav vendors; they changed the economics. Once turn-by-turn directions lived inside the default device, the specialist box on the dashboard became optional, then eccentric.

The SatNav didn’t die from lack of innovation; it died because navigation moved. Directions migrated into the pocket, and everything else became aftermarket.

Legal AI may be nearing a similar hinge-point. The question may soon stop being “which tool is cleverest?” and become “which platform becomes the place where the work happens?” The threat isn’t that your favourite legal AI vendor is worse. It’s that a foundation model becomes good enough by default — and ‘good enough, everywhere’ beats ‘best, somewhere’.

Finance has already started to run this experiment in public — and Anthropic’s moves with Claude are a useful case study in what “verticalization” actually looks like.

Pipes, Not Prose: What Claude’s Finance Push Really Shows.

In finance, Claude isn’t being sold as “a clever chatbot”. Anthropic has been doing something more strategic: turning a foundation model into a sector package — the kind of thing a regulated industry can actually adopt without blushing.

Start with the framing. In mid-2025, Anthropic launched Claude for Financial Services not as a new model, but as a “financial analysis solution”: one interface designed to sit on top of the mess of market feeds, research, and internal datasets, with an emphasis on linking back to sources so users can verify what they’re being told.

Then comes the real moat: pipes, not prose. The finance play isn’t “ask Claude anything”; it’s “let Claude see the right things”. Anthropic has pushed connectors into the stack — including integrations with enterprise data platforms and a growing list of specialist financial data providers — so the model can work inside the ecosystem firms already pay for and trust.

Recently, Anthropic made the “pipes” point even clearer by shipping Claude Cowork plug-ins — including a Legal plug-in — as packaged role modules wired into enterprise tools and data via connectors. That is verticalization in miniature: not a new model, but a distribution-and-integration wrapper that turns a general model into a profession-shaped workflow. The detail matters less than the direction: capability is being standardised into deployable bundles, so adoption becomes procurement-friendly rather than enthusiast-led.

Next, distribution. Anthropic has also moved Claude into finance’s native work surface: Excel. That shift matters more than it sounds. When the AI lives where the work lives — where assumptions are made, numbers are stressed, and models are built — it stops being a tool you “try” and starts being a feature you use.

Finally, legitimacy. Partnerships like the one announced with London Stock Exchange Group (LSEG) underline what this is really about: licensed data, governed access, and an enterprise story that can survive procurement, compliance and audit.

Why should the legal profession care? Because this is the blueprint. Not “AI for everyone”, but AI packaged for a profession: embedded into the work surface, wired into trusted knowledge, and wrapped in controls. Finance is running the experiment in daylight. Law may not be far behind.

From Add-On to Default: When Word Becomes the Desk

If Excel is finance’s native work surface, Word is ours. That is the uncomfortable symmetry. Once a model is embedded where the work actually happens, the competitive battlefield stops being “best tool” and becomes “default surface”. In finance, that surface is a spreadsheet. In law, it is the document.

And that is why a seemingly modest piece of industry news matters: Robin AI — a legal tech startup that had been on the rise — is now being dismantled in plain sight. Its capability has begun to fragment, and a meaningful slice of its talent has reportedly re-homed inside Microsoft Word. Microsoft Corp. hired at least 18 former Robin AI employees into the Word organisation, including senior figures now in Word leadership and applied science roles.

On one view, this is just a hiring story. On another, it could be a signal about direction of travel: the platform owner isn’t merely partnering with “legal AI vendors”; it is absorbing legal-workflow capability into the platform itself.

So what is the plausible scenario we should hold in mind?

Not “Word becomes a law firm overnight.” Something subtler. It is at least possible that the drafting surface becomes the AI surface. The tools lawyers already live in could start to feel more “legal-native” by default: drafting within firm templates; suggesting edits and alternative formulations; summarising and reconciling revisions; and pulling context from document stores and communication threads — and, where connectors exist, from systems like document and matter repositories — all wrapped in enterprise-grade controls that make wider adoption feasible.

In practice, that might look less like a new standalone product and more like a Copilot-style layer sitting across the everyday workflow — Word for drafting, Outlook for correspondence, Teams for collaboration — with legal-specific features designed to reduce friction rather than replace judgement. If it develops that way, the benefits are easy to see: faster turnaround on routine work, more consistent first drafts, and quicker retrieval of what the team already knows.

A Proposal, Not a Prophecy: How to Stand Up in a Moving Market

None of this is inevitable. It is a proposal about direction, not a prediction about destiny. But the direction matters, because it suggests legal AI is not “going away”. It is moving closer to the desk, becoming more ambient, more embedded — and with that comes epistemic risk —, and therefore harder to ignore.

So the practical question for firms is not whether to “buy AI” or “wait it out”. It is whether you want change to arrive as policy, or as accident.

The point is simple: the tide may be turning. The firms that do best won’t be the ones who guess the winner. They’ll be the ones who build safe footing — so they can move when the ground moves, without losing their balance — and keep the last candle lit.

Article written by Dr Corsino San Miguel PhD, LLB in Scots law and graduate in Spanish law; co-founded and led European Telecom Company before entering academia. He is now a member of the AI Research Group and the Public Sector AI Task Force at the Scottish Government Legal Directorate. The views expressed here are personal.