Implementation
Not a tool you buy or a licence you renew. A short, focused engagement that leaves four things behind — custom skills running on your workflows, a trained team, an AI Use Protocol, and a clear picture of what to do next.
A methodology that
happens to use AI.
Not the other way around.
The frontier labs ship the kit. The methodology is what makes it work in a regulated firm — specialist legal, compliance and technical expertise, applied as one team. What sets one firm's output apart from another is not the model it uses. It is the methodology, the governance, and the senior person putting their name to the work.
82%
of organisations do not measure whether their AI investment is actually working. They are buying tools, training staff, signing contracts — and hoping.
Thomson Reuters · AI in Professional Services 2026
What You Get
An implementation engagement is not a set of deliverables handed over and filed. It is an installation — four outputs the firm owns and operates after we leave.
Output 01
Custom-built AI workflows for the work your firm actually does — DD, red flag review, drafting, instructions to counsel, whatever the engagement scopes. Built to your house style, deployed inside your environment. Owned by you. Not a licence that renews.
Output 02
Foundational AI literacy for the firm, workflow training built around your matters, and a champions programme for the people who will carry the methodology forward after we leave. Not a one-day course. A working programme with a practice gap built in.
Output 03
The governance document, signed before any AI-assisted work goes out. The thing that lets the firm stand behind its output to a regulator, a client, or a court. Written for your matters, your risk appetite, your standards.
Output 04
A graded, costed view of the work AI changes in your firm — prioritised by impact, complexity and risk. If you came via Discovery Day, this is already in your hands. If you are coming direct to Implementation, we scope it at intake.
The Methodology
Every implementation runs the same gated process. Specialist legal, compliance and technical expertise, working as one team. Frontier AI models doing the heavy lifting. A qualified lawyer accountable at the end.
The methodology is what makes the output defensible — and what separates implementation from a generic AI tool pointed at legal work.
The frontier labs ship the kit. The methodology is what makes it work in a regulated firm.
Checks before action
Every workflow starts with a structured check of scope, source material and intent. No skill runs until the basis is clear and documented.
Model selection per task
The right model for the right work. Top tier for judgment and drafting, cheaper tiers for bulk processing. Explicit flagging when we downgrade.
Evidence discipline
Disclosure versus response. Filing versus content. Versioned change logs. Citations and provenance throughout — the evidential standards a regulator would expect.
Verification before delivery
AI re-reads its own work. A second pass through a different platform where available. Human lead-adviser sign-off on every output before it ships.
Who Works on It
The person who pitches is the person who designs the methodology and the person who delivers the work. The firm meets all three on the first call and works with all three through to handover.
Project lead
Designs the engagement, owns the methodology, signs off the work. UK lawyer with a consultancy practice. The accountable professional on every matter — the person you call if something needs resolving.
Build engineer
Builds the skills. Configures the frontier model layer, tunes the workflows to your house style, and deploys into your environment. Stays through go-live and the 30-day support window that follows.
Project manager
Holds the engagement together. Diary, deliverables, deadlines. The person making sure the work lands when it should and the firm always knows where things stand.
What Gets Built
Every build is scoped to the firm's specific workflows. These are the most common starting points — what firms reach for first once the methodology is in place.
Build 01
A gated skill that processes a data room and produces the full DD package — risk report, file index, and outstanding items tracker — to a consistent house standard.
Build 02
Upload a new matter bundle and receive a structured first view — summary, chronology, issues list, and the missing documents list — ready for the first client call.
Build 03
AI-assisted preparation of briefs to counsel, pulling from matter documents and structuring instructions in the format the firm's counsel expect.
Build 04
Staged drafting from a firm template with variation instructions — produces a populated, matter-specific draft in the firm's house style, ready to review.
Build 05
Combines disclosure tool output with Claude analysis — a seamless pipeline from financial data room or disclosure output to structured financial findings.
Custom scope
The five above are the most common starting points. Most firms have a repeatable workflow that is not on this list. If you know what it is, we scope it. If you are not sure, a Discovery Day finds it.
Proof
The skills are running in live client matters across the UK, US, Dubai and Australia. The methodology has been benchmarked against the leading legal AI platform on real documents, and outperformed it.
Benchmarked matter
A company acquisition. 500 documents. A full risk review and indexed report. Quoted at £35,000 elsewhere over two weeks. Delivered through Silva AI in under a working week, refined across four reviews since.
The AI pass took one hour. The skill outperformed the leading legal AI platform on the same documents.
500
documents reviewed
1hr
AI pass
4×
refined since delivery
Market signal
In April 2026 a magic circle firm went wall-to-wall with a frontier model across 5,700 staff in 33 offices — co-developing workflows with the provider. Weeks later the same lab released practice-area plugins for everyone else.
The architecture they spent millions building is the architecture we install in firms of five to fifty fee-earners.
Get Started
Every implementation is scoped before it is priced. The Discovery Day produces the scoping document — the costed Menu of Opportunities. If you are coming directly to implementation without a Discovery Day, we scope on a short call and follow with a fixed-fee quote.
Includes documentation, internal champion training, and 30 days of post-go-live support on every build.
Most firms run Discovery Day into Implementation into a Retainer. Each tier is standalone — take one, combine, or run them in sequence.
Where this sits