AI for legal, built around the matter file.
Document review, drafting, disclosure, bundling. We embed senior AI engineers inside UK law firms and in-house teams — with citations back to source, audit logs and partner sign-off where they belong. The fee earner keeps the judgement; the AI takes the volume work.
AI for legal in 2026 is either a glossy product demo or a frontier-lab proof-of-concept that never reaches a fee earner. Neither helps the firm running an actual matter. We build inside the practice — document review, drafting, disclosure, bundling, conflict checks, time-recording narratives, NDA triage — in the systems your fee earners already use.
We have spent the last three years rebuilding our own businesses around AI — Homemove, home.co.uk and homedata.co.uk. Confidentiality, document control, audit trails and matter-level segregation are part of how we already build, not a bolt-on at the end.
Single-office practices through to mid-market firms. In-house legal teams inside operating businesses. Monthly retainer plus transparent hourly delivery; most engagements pay back inside the first quarter.
Document review and drafting
First-pass review with confidence scoring. Drafting against your firm’s precedent bank. Comparison and redline at speed. Every output cites the source clauses it relied on so the fee earner can verify in seconds.
Partner sign-off intact
Nothing leaves the firm without the human review you have today. The AI shortens the cycle; it does not replace the gate. Audit logs make a partner’s sign-off defensible to a client, an insurer or an SRA review.
First-quarter pay-back
We start with the workflow consuming the most associate hours — usually first-pass document review or bundling. Most clients save more in the first quarter than the entire engagement costs.
Common questions.
How is this different from a legal-tech product like Harvey, Spellbook or Luminance?
Those are products with a fixed surface. We are an engineering team that builds inside your stack. If a product already solves your problem, use it. If your workflow needs custom integration with your DMS, your CRM and your matter management, that is where we come in — we build with the same models the products use, sized to your firm.
Will our client data train external models?
No. We default to deployments inside your own cloud and your own controls, with private model endpoints. We never train external models on client data without an explicit written agreement. NDAs and DPAs are signed as standard.
How do you handle privilege and confidentiality?
Per-matter access controls, structured audit logs on every prompt and output, no retention by external model providers under enterprise agreements, and full segregation of duties. We sit with your information-security officer during the diagnostic and answer the questionnaires honestly.
Will the AI hallucinate citations?
Models can and do hallucinate. We treat that as a design constraint. Every output cites the source clause or document it relied on. The fee earner reviews the cite, not just the text. Workflows where hallucination is unacceptable run with strict retrieval-only modes and confidence thresholds, with human-in-the-loop on anything below them.
What does AI for a law firm cost?
A monthly retainer reserves capacity and covers planning and oversight; delivery hours are billed transparently on top. Rate quoted on the first call once we have seen the practice areas and the volume.
AI for professional services
The wider professional-services pitch — accountancy, consultancy, advisory firms — with the same audit-log and confidentiality defaults.
About James, Louis, Phil & Shivam
The operators behind the work. £5m+ raised for their own companies, ~£10m across their broader career, AI-native since the week ChatGPT shipped.
More Academy entry points.
The Academy is one course with many doors. Each of these pages is a different entry point into the same twelve weeks.
AI for the matter, not the marketing.
Tell us where the associate hours are draining away. If we’re a fit, we’ll come back inside the week with a build plan, not a vendor pitch.
© 2026 Moonlabs Incubator. All rights reserved.