Methodology
Three engagements. Three deliverables. You own the output.
We don’t do open-ended retainers. Each engagement is a discrete piece of work with a defined deliverable. You decide whether to proceed to the next one.
01
Diagnostic
A written assessment of where AI sits, what’s exposed, and what’s worth building. Useful on its own. You can stop here.
02
Build
Production AI, single-step or agentic depending on the job, integrated, evaluated, owned by your team. Not a pilot. Not a POC.
03
Handoff
Your team operates the workflow. We document, train, step away. Optional ongoing only when you actually want it.
01. Diagnostic
What you walk away with
A written report you can act on, internally, with or without us. The diagnostic is complete in itself. It surfaces where AI sits, what’s exposed under the AI Act, and what’s worth automating.
- Inventory of AI systems already in use, with risk classification under the EU AI Act
- Annex IV gap analysis for any high-risk systems identified
- Use-case map: which agentic workflows would deliver measurable ROI, sequenced
- Governance lite: AI register template, transparency notice patterns, oversight loop sketches
- The decision: build, defer, or done
View a sample Diagnostic →
02. Build
What you walk away with
Production AI, running in your infrastructure, integrated with your existing tools, documented, evaluated. The shape calibrated to the work, single-step automation where that’s right, agentic orchestration where it’s warranted. Not a prototype that needs a follow-on engagement to ship. The thing itself.
- The system: code, prompts, tool integrations, orchestration logic where applicable
- Evaluation harness: held-out test set, accuracy and safety thresholds, drift monitoring
- Annex IV technical documentation for high-risk systems where required
- Operator runbooks and two training sessions for your team
- Connection to your existing tooling stack (Microsoft 365, Google Workspace, Notion, Slack, your vertical software)
03. Handoff
What you walk away with
Operational independence. Your team runs the workflow without us. The artefacts are yours: code, prompts, evaluations, runbooks, IP. Ongoing engagement is optional, never assumed.
- Full transfer of code, prompts, evaluation suite and documentation
- 30-day post-launch support window for operational questions
- If you want, an evaluation cadence (monthly or quarterly) as a separate engagement
- If you want, regulatory horizon scanning as a separate engagement
- If you don’t want either of those, you operate independently. That is the default.