Twenty clicks.

Somewhere in England, right now, a solicitor is finishing a client call. The advice was sound; the direction clear. Then the call ends, and the real work begins: the case management system, the attendance-note screen, the conversation reconstructed from memory, the follow-up email, the letter of advice, the task list, the time log, the file status. In the accounting of the company this transmission concerns: “Twenty clicks. Minimum. For a single phone call. And you do this thirty times a day.” Their arithmetic for what that habit costs runs to four to six hours per lawyer per week — fifty-two to seventy-eight thousand pounds of recovered capacity a year, per lawyer, if you could give the hours back. The profession’s scarcest resource, spent on transcription.

Software was supposed to have fixed this by now, and the past two years produced a wave of legal AI that promised to — most of it, underneath the demo, a general-purpose model wearing a gown: a wrapper, in the trade’s own dismissive term. The firms resisted, and the vendors read the resistance as backwardness. It rarely is. A law firm’s slowness is structural: SRA regulation, anti-money-laundering duties, professional indemnity, client confidentiality — frameworks that govern not just what a firm does but what it is allowed to buy. The buyers are not afraid of software; they are accountable for it, in a profession where one invented citation ends a career. And the people building the tools have mostly never sat in the meeting where that accountability is priced.

This series documents operators who encode hard-won judgment into systems — most recently in our last transmission, where twenty years of regulatory programmes became a platform that anchors every reported number to its rule. Today’s filing — the twelfth — carries a wrinkle this publication is built to appreciate. At the time of writing, the company’s homepage is one sentence long: “Something new is being built.” Behind that front door, we found the rooms already furnished — a finished platform, a pilot programme, a named pilot partner, a team that runs from agentic-AI infrastructure to legal partnerships. Most founders launch louder than they build. The operator in question — a UK solicitor who was writing code before most of us managed joined-up handwriting — is doing the opposite.

The operator in question

Mateusz — first names are policy here; he can introduce himself properly — describes his path as “a lifelong obsession with digital infrastructure,” and the receipts start early: by his own accounting he was hosting private servers and writing code at six years old. By his early twenties he had bootstrapped a digital agency, architecting full-stack web presences and running digital acquisition for small businesses — a founder once already, before the profession. Then came the move that makes the profile singular: he qualified as a solicitor and went into the deal rooms — complex M&A and corporate transactions, the corner of law where the documents are longest, the deadlines hardest and the tolerance for error lowest.

Inside the firm he behaved like what he says he is — “an engineer and builder at heart.” He sat on internal Tech and M&A committees, evaluating cloud and Citrix migrations; he managed post-merger software integrations, the unglamorous work of making two firms’ systems into one; and when his department’s workflows ground against their worst bottlenecks, he engineered macros to automate them — automations that, as the Asideus team page now puts it, “partners called ‘sorcery.’” Underneath it all, an education was compounding. In his words, he “learned exactly how law firms operate, why they resist change, and the strict regulatory frameworks (SRA/AML) that govern their technology procurement” — which is the sentence a hundred legal-AI pitch decks are missing.

Notice what the CV assembled, in order: how software is built, how businesses buy it, how lawyers actually work, and what a law firm is permitted to purchase. Four fluencies that almost never live in one person — the engineer who can read a share purchase agreement; the solicitor who has shipped. In January 2025 he stopped patching the machinery from inside it and co-founded Asideus.

The re-architecture

The founding claim is a sentence with its teeth showing: “the legal sector doesn’t need another generic AI ‘wrapper.’ It needs a fundamental re-architecture.” Asideus is the argument made concrete: “Infrastructure for the autonomous law firm” — “one operating system focused on the only things that matter at the architecture level — data integrity, autonomous compliance, and operational leverage.” The mechanics are deliberately unexotic. It links to the practice management systems firms already run — “Link to Clio or Leap. No data migration. No workflow disruption.” — then “indexes your precedents, documents, and working patterns within days.” Calls are captured and become attendance notes in around three minutes; an assistant named Leo reads across cases, documents and research; matters, knowledge, tasks and time recording flow through one system instead of six. The twenty clicks are the product’s whole target: “Manual admin disappears. Compliance becomes automatic.”

That last word is where the lawyer in him shows. Every automated pathway, the platform says, “is audited in real-time, ensuring strict legal compliance before any action is taken” — a system “compliant by default,” UK GDPR compliant, UK data residency, SRA-aligned. His own profile plants the boldest version of the flag: architecture developed “to solve the hallucination problem in legal AI.” Solve is a strong word in this territory; the structural approach — verification before action, audit trails by design, multi-agent frameworks built by ML engineers and legal experts together — is at least aimed at the right layer. A technical adviser quoted on the site describes the result from outside: “this isn’t a chatbot bolted onto a case management system. It’s a genuine intelligence layer that connects everything. Nobody else is building this.”

Then there is the price, which is a positioning statement wearing a number. Asideus is aimed squarely at the firms the AI wave keeps skipping — “If you’re a 10–50 lawyer UK firm, Asideus was built for you” — at £99 per user per month, in a market where the specialist tools it gets compared against start at thirty thousand pounds a year. The pilot programme underneath is live and, in the company’s phrase, “highly selective”: a named pilot partner in Aria Grace Law CIC — Law Company of the Year 2023, by the site’s accounting — ten-plus firms engaged since January 2026, seventy-plus experienced lawyers across five jurisdictions in the orbit. All of it assembled behind a homepage that still says coming soon. (The builder’s restlessness leaks out sideways, too — a part-time role at Rocket Logic Global, supplying the product vision for a browser-based operating system for autonomous AI agents. Even the side project is an operating system.)

For the twelfth time in this series, the shape — and this one is a double encoding. What goes into Asideus is not only legal judgment — how firms operate, what the SRA will ask, where the AML duties bite — but engineering judgment three decades deep, the kind that starts with a six-year-old’s server and never stops running. Most founders bridge to a domain expert; a few domain experts bridge to the engineers. The rarest case needs no bridge at all: one head where both kinds of judgment argue at native speed. That is what a re-architecture requires — and it is why the wrapper wave keeps producing wrappers. Bridges are where the judgment thins.

Notes for anyone selling AI into a regulated profession

Strip away the gowns and the rules travel to anyone whose buyer answers to a regulator:

Solve for the veto, not the demo. The user who loves your product is not the framework that approves it. Design for the person whose job is to say no — the regulator, the AML officer, the indemnity insurer — because they will be in the room, and your demo will not.

Treat accuracy as a threshold, not a score. A profession that is paid to be right cannot average you into trust. Below the line, ninety-five per cent rounds to zero; the last five points are the product.

Automate the bottleneck you have personally bled on. The macros the partners called sorcery came years before the company. A tool built from remembered pain skips the year of discovery interviews — and its founder cannot be fooled by his own roadmap.

Fluency in both rooms beats brilliance in one. A wrapper is what you get when the builder knows the model but not the firm. The moat is not the model; it is knowing precisely why the buyer resists.

Build in silence; launch with proof. In hype-allergic professions, restraint reads as competence. A one-sentence homepage over live pilots beats a launch video over a waitlist.

The finishing touches

What changes, if the architecture holds, is procurement-shaped. A firm that could never sign off a chatbot can evaluate a system: where an answer came from, what verified it, what was logged and for whom. And the before-and-after the company describes is really a before-and-after of firm memory — “institutional knowledge walks out the door” becomes “every matter compounds the firm’s knowledge”; partners deciding on gut feel become partners with real-time visibility. The profession’s resistance was never to change — the committees Mateusz sat on existed precisely to buy change. It was to unaccountable change. Meet the accountability in the architecture and the resistance becomes what it always secretly was: a queue.

So: the operating system is at asideus.com — where, at the time of writing, the front page still says “Something new is being built” while the rooms behind it fill with pilot firms. The sentence is true and it undersells. Somewhere in England, a solicitor is hanging up a phone, and the attendance note is writing itself — three minutes, audited, logged — while the twenty clicks go the way of the fax machine, unmourned. The engine he says he is building — for what he calls a “New Era of Professional Services” — will make its own noise when it launches. Until then the coming-soon page is not a promise. It is a status report, and the status is: further along than it says.

— END TRANSMISSION 06D