Claude.ai
01Research, drafting, and analysis in the browser, with Projects, files, and web search in one place.
Enablement · The programmes
Genuine AI fluency at every level of the firm, from analyst to partner.
Hands-on Claude sessions that take an entire firm from first principles to building with the most advanced tools in the suite.
A team trained once is a team falling behind. The technology underneath this work shifts on a weekly cadence, so a firm that trained six months ago is already operating on stale knowledge.
Our programmes are monthly precisely because the technology is weekly. Each person learns to recognise where Claude belongs in their own work, and stays current as the market keeps moving.
Always on the frontier
The frontier moves weekly. We keep your team on it.
New models, tools, and capabilities arrive on a weekly cadence. Our sessions land at a steady rhythm so the team is never more than one step behind the edge.
Frontier shifts, weekly
Solven session, monthly
How it works
On your own workflows, not generic demonstrations.
- i Hands-on, on the real work.Sessions run on the firm's actual deals, models, memos, and reports, so each person leaves knowing where Claude fits their day.
- ii Month one is breadth.Orientation across the full Claude ecosystem, from first principles to building, so the whole team shares a map of what is possible.
- iii From month two, we go deep.Later sessions concentrate on the areas the team finds most valuable. The session's arc stays; the path adapts.
- iv Trainers read the room.The topics below are guidance, not a script. Pace and emphasis follow the people in the room, analyst to partner.
What we cover
The full Claude toolkit.
Each person learns which tool fits which part of their work, so the whole suite is in play, not just the chat box.
Claude Code
02The most capable tool in the suite, and far more than a developer tool: organising files, automating recurring work, building real systems.
Session fourClaude for Chrome
03Claude working alongside the pages and tools your team already uses, without leaving the browser.
Claude Cowork
04Claude that takes actions across the desktop, running a workflow end to end rather than only generating text.
Session threeClaude for Excel
05Formula generation and audit, data cleaning, and model building, with Claude reasoning across the whole workbook.
Session threeClaude for PowerPoint
06Decks, charts, and IC-ready memos drawn from outlines, notes, or source documents, aligned to the house format.
Session threeThe four-session arc.
From how a model actually thinks, through conversation and production, to building agentic systems with Claude Code.
01
- The history of AI through the transformer breakthrough
- Tokens, embeddings, vectors, and attention
- How generation works: probability, sampling, run-to-run variance
- Where Claude is sharpest, where it fails, and why hallucinations happen
- The analyst-first principle and the habit of verification
How a model turns text into output
Textraw input
Tokenscomputable units
Vectorsmeaning as geometry
Attentionwhat matters
Outputprobabilistic
02
- The interface: chat, Projects, Memory, files, web search, Artifacts
- Structured prompting: XML tagging, role framing, output specs
- Few-shot prompting, chain-of-thought, and metaprompting
- The devil's advocate pattern: Claude as critic, not author
- Context-window management, token economics, knowing when to stop
Anatomy of a structured prompt
Context
Instructions
Examples
Output spec
03
- Claude for Excel: formula generation, audit trails, financial models, cross-sheet reasoning
- Claude for PowerPoint: decks from outlines, brand-aligning, IC-ready memos
- Claude Cowork: desktop automation and multi-app workflows
- The bigger idea: tools and function calling
- Why Claude + tools is fundamentally more capable than Claude alone
Tools around the model
Excelmodels & data
PowerPointdecks & memos
Coworkdesktop agent
Claude + toolsfunction calling
04
- The agentic loop, version-controlled prompts, and markdown fluency
- Skills & SKILL.md: packaging team expertise as deployable capability
- MCP: connecting Claude to databases, APIs, and internal systems
- The infinite-horizon problem and explicit done-conditions
- Automation patterns: structured extraction, batch processing, report generation
The agentic loop
Plandecompose
Actexecute
Verifycheck work
Iteraterefine
Iterate feeds back into the plan — the loop runs until the done-condition is met.
The thread
Every session carries the analyst-first principle.
Analysts produce the formative work first, then bring Claude in as a critic, a red-teamer, and a verifier.
The risk in any rollout is that talented people become editors of drafts they cannot fully defend. We design every session the other way round. Verification habits, the devil's advocate pattern, AI as critic rather than author — the same skills that unconstrained use erodes are the ones we build.
That is what we mean by augmentation without cognitive atrophy, and it runs through all four sessions and everything that follows them.
Programme structure
Built to flex around the firm.
Month to month
Programmes run monthly with no minimum commitment. Stay as long as the frontier keeps moving — which is to say, as long as it is useful.
Sessions plus build hours
Training pairs with optional build and consulting hours, so confidence from the room turns into real systems on real workflows.
On site or remote
Delivered at your office in Lisbon, remotely, or a mix. The format follows the team, not the other way round.
Open to the whole firm
Analyst to managing partner. Fluency is a firm-wide capability, not a specialism held by a few. Each engagement is scoped and priced bespoke.
Member of the Anthropic Partner Network · Phase 1
A conversation to tune the programme to your firm's priorities.
Tell us where your team is today and what the next quarter looks like. We will shape the arc around it.
Start a conversation →