assess: decide what to build and why.

01 · The problem

what we hear

• “We know AI matters but don't know where it fits.”

• “We ran a pilot because we felt we should, and nobody can say if it worked.”

• “Every vendor says their tool is the answer.”

02 · Diagnosis

why it happens

The expensive mistakes happen before any code is written: solving the wrong problem, choosing an architecture that is hard to leave, or automating steps that should stay with a person.

03 · Outcomes

what we deliver

Scoped solution

A business problem translated into a defined Claude solution, with the architecture, model, context strategy and entry point chosen and justified.

Delegation map

Which steps Claude takes, which stay human and which are collaborative, judged on reversibility, stakes and accountability.

Trade-off brief

The options and their consequences, written so non-technical stakeholders can make the decision.

Roadmap

What to build first, what to wait on, and what would change the plan.

04 · Engagement

how it runs

1. discovery
2. analysis
3. recommendation
4. readout

05 · Honest limit

what we won't do

Sometimes the right answer is not to build an agent, or not to use AI for a step at all. When that is the answer, we will say so.

06 · Proof

see how we think

From our writing on Agentic workflows and automation.