Ask Silva AI

Training for Professionals

AI training built
for law firms.

Two sessions. A deliberate gap between them. Foundations in Session 1 — live demos against your actual work in Session 2. Built for the people who will use AI day to day: partners, associates, and paralegals.

Book a training programme What’s covered ↓
2 × 2hrs
Two sessions of two hours each — foundations first, live demos second
Remote
Delivered online — no travel, no scheduling overhead, available to any team
3–4 weeks
Practice gap between sessions — the firm tries things and brings real questions back
Whole team
Partners, associates, paralegals — designed for everyone who will use AI on matters

Not a workshop.
A working programme.

Most AI training for law firms teaches tools. This programme teaches how to use AI on real legal work — with the judgement, habits, and governance that regulated practice requires.

Session 1 builds the foundation: how the models work, where they fail, how to structure a prompt, how to set up Claude for the firm, and what the courts and regulators currently say. Session 2 picks up where friction appeared — and runs live demos against the firm's own documents and workflows.

The gap between sessions is part of the design. It is where the firm tries things, hits the edges, and forms the questions that Session 2 answers with working examples.

AI does not mean cheaper fees. It means better service, delivered more efficiently, at a higher effective rate.

The commercial framing  ·  Session 1

Two sessions.
One practice gap.

Session 1 builds the foundations. Session 2 goes deeper against your actual work. The gap is where learning happens.

Session 1  ·  2 hours

Foundations

How the models work, where they fail, how to build a prompt that works, and how to set up Claude for legal practice. Regulatory and court guidance included.

  1. What generative AI is — and how it gets things wrong
  2. The models: Claude, ChatGPT, Gemini, CoPilot and the legal-specific tools
  3. Where AI is already useful in legal practice
  4. The commercial case — better service at a higher effective rate
  5. Six risks to keep in front of mind on every matter
  6. Regulatory and court guidance — what is permitted, what is not
  7. Setting up Claude: enterprise data, Projects, custom instructions
  8. Prompt engineering fundamentals: the five-part framework
  9. Live demo — day-to-day legal tasks run in the room
Practice gap 3–4 weeks

Session 2  ·  2 hours

Going Deeper

Advanced settings, skills, and three live demos run against documents the firm brings. Answers the real questions the practice gap surfaced. Not slideware.

  1. Recap: what to hold on to from Session 1
  2. Three layers of instruction: global, project, and prompt
  3. Setting global instructions — set once, carries jurisdiction and style
  4. Projects: one per matter, persistent, structured
  5. Chat memory and Edit Memory — what to use where
  6. Context windows in practice — and what to do when they fill
  7. The three models — Opus, Sonnet, Haiku — which for which task
  8. Artifacts: version history, ticket savings, easy export
  9. What a skill is — and why a bad one is worse than none

Weak prompt.
Strong prompt.

The five-part framework turns a generic instruction into a working brief. Here is the same task — written twice.

Before training

Write me a letter responding to the other side's offer.

What comes back

A generic letter with no understanding of position, jurisdiction, client, law, or firm tone. Plausible-looking. Not usable.

After training

You are a UK commercial solicitor acting for a technology company in an employment dispute. Role + Context

Draft a without-prejudice letter responding to the claimant's Part 36 offer of £45,000, made on [date]. Our client wishes to counter at £18,000 on commercial grounds. Task

Write in formal legal correspondence style, headed "Without Prejudice Save as to Costs", two pages maximum. Format

Do not accept liability. Do not exceed the commercial settlement figure. Flag any Part 36 procedural requirements we should address. Constraints

What comes back

A jurisdictionally accurate, properly headed, strategically calibrated first draft — ready to review and send.

The five-part framework

Role Context Task Format Constraints

Real work.
Real documents.

Session 2 is not slideware. Three demos, run live in Claude against material the firm brings to the session. This is where the training becomes practice.

Demo 01

Initial Document Review

A new matter lands. The clock is running. Here is how to get to a structured first view — fast — without cutting corners on review.

Input A new matter bundle — pleadings, correspondence, key documents
Output Structured summary, chronology, issues list, and the first questions to put to the client
The point What took half a day comes back in minutes. The review time goes to judgment, not extraction.

Demo 02

Drafting from a Template

Staged drafting from a firm template with live variation instructions. Shows the iterative workflow — how to get from template to matter-specific draft without starting from scratch.

Input Firm template plus variation instructions — parties, clauses, client-specific positions
Output A populated, varied, review-ready first draft — in the firm's house style
The point The blank page is gone. The review is where the lawyer earns the fee.

Demo 03

Finance Review

Financial disclosure is often where matters drag. This demo shows how to combine tool output with Claude analysis for a seamless pipeline from disclosure to findings.

Input Output from a disclosure or finance tool — bank statements, asset schedules, valuations
Output Structured financial analysis, anomalies flagged, figures extracted and organised by category
The point Financial review work handled at the extraction stage. Lawyer effort focused on the analysis.

Training shows what’s possible.
The rest builds it in.

Training is the entry point. Around it sits the work that turns what the firm learned into how the firm operates.

Skills creation

Bespoke skills built on workflows the firm already trusts. The same task, done to the same standard, by any fee-earner who runs it.

  1. Due diligence reporting from document bundles
  2. Initial case review — summary, chronology, issues list
  3. Instructing counsel — AI-assisted brief preparation
  4. Contract redrafting from a house template
  5. Disclosure pipeline — tool output to Claude analysis

Compliance and AI documentation

The governance layer. Written before the firm relies on AI at scale — not after something goes wrong.

  1. AI protocol for client agreements and T&Cs
  2. Privacy policy updates for AI data processing
  3. Internal AI policy — who uses it, for what, with what safeguards
  4. Court compliance framework tailored to the firm's practice areas

Workflow review

A look at how the firm is actually using AI — not how it thinks it is. Where it is working, where it is not, and where the repeatable wins are sitting uncaptured.

  1. Observe live workflows across selected practice areas
  2. Identify where AI is being underused or misused
  3. Map the repeatable tasks that are candidates for skills
  4. Report with prioritised recommendations

Ongoing support

A named adviser available when the firm hits something — not a retainer, not a ticketing system. Direct access when the work requires it.

  1. Teams or WhatsApp — ask a question, get an answer
  2. No minimums, no retainer — just availability
  3. Covers prompting questions, skill issues, compliance checks
  4. The relationship that follows from a working programme

The questions
every firm asks.

Answered here, so the first session can focus on work.

No. Enterprise and Team accounts explicitly do not use conversations or documents as training data. Your client matter information stays private to the organisation. This is a critical difference from free consumer tools, and it is one of the first things the training covers — along with where to confirm it in the account settings.

On Team and Enterprise plans, no. Projects persist until you delete them. The 30-day limit applies to individual chats outside Projects on the free tier. For any work you want to keep — a matter, a client, a research thread — put it in a Project. The training covers Project setup in Session 1 and goes deeper on structure in Session 2.

Not for generating content. Court guidance across jurisdictions consistently prohibits AI from creating, altering, or embellishing witness evidence. Preparatory work behind sworn documents — research, chronologies, issue summaries — is generally permitted, subject to disclosure obligations. The training covers the current regulatory and court guidance and gives the firm a framework for reading whatever specific directions apply to its jurisdiction and practice area.

Transparency is the right position — and increasingly the required one. A formal AI protocol in your client agreements sets out how AI is used in the firm's work: AI review, human review, and client approval. It manages expectations, mitigates negligence risk, and positions the firm ahead of regulatory requirements rather than behind them. Silva can draft this as part of the compliance documentation work.

No. AI gets you to 70–80% in minutes. The last 20–30% is professional judgment. The skill is knowing which 70–80% to keep. AI handles the blank page, the data extraction, the first draft. Your fee-earners do the work that requires knowledge of the client, the strategy, and the matter — which is most of the work worth charging for.

The firms that build this in early will deliver better service at a higher effective rate. The ones that do not will be competing against those that have.

Fixed fee.
Per session or full programme.

The full programme is two sessions of two hours each, delivered remotely, to up to the agreed number of attendees. Priced per session or as a fixed programme fee. Wrap-around services — skills, compliance documentation, workflow review — scoped separately.

From Contact us for pricing

All programmes are subject to a brief scoping call to confirm fit and confirm attendee numbers. Bespoke programmes for larger teams scoped on enquiry.

Who this is for

Partners who need to understand the risk, governance, and commercial case before putting AI in front of their teams
Associates and paralegals who will use AI on matters day to day and need the habits and limits, not just the tools
Firms running off-the-shelf AI tools without an operating model around them — the training builds the model
Teams preparing for a larger AI programme — Discovery Day or Implementation — who want a working foundation first