Why exit-points matter. Most consulting engagements are sold whole. Ours are sold in stages because the second stage should not exist until the first one has paid for itself. About 1 in 3 Discovery Weeks ends with us telling the founder they are not ready for AI yet — and we mean it.
Fifteen-minute questionnaire, one 30-minute call within 24 hours, one shadowed working session. We map your current workflow inputs, the cost per output, the human hours involved, and the systems-of-record where the work lives. Output: a written roadmap, ranked by ROI, with the top two or three automations specified to MVP scope.
About 1 in 3 Discovery Weeks ends with us recommending you do not proceed with AI — usually because the bottleneck is upstream of where AI helps (CRM hygiene, missing process, broken handoffs). The roadmap is yours either way.
If you proceed, we spend two weeks building the integration backbone — CRM connections, document indexing, permissions model, observability scaffolding. Most engagements have hidden plumbing debt that would surface mid-MVP and kill the timeline; we surface it now while the cost is small.
Output: a working data + permissions layer the MVP can stand on, plus a finalised technical spec for the first agent.
One agent shipped to production for the highest-ROI workflow from the roadmap. Built in 4–6 weeks, deployed under human review, measured against a baseline you set in week one. We instrument it from day one — every action is logged, every output reviewable, every cost trackable.
Most MVPs land at €12–18k all-in (build + first month of ops). We will tell you on the discovery call where in that range we expect yours to fall.
The MVP is in production; now it has to stay in production while the world changes around it. Source documents drift, regulations update, your own workflows evolve, model providers ship new versions. The maintenance retainer covers monitoring, eval-harness re-runs, source updates, and quarterly tuning sessions.
Three tiers — Light (monitoring + monthly eval), Core (Light + tuning + new use-case discovery), Plus (Core + dedicated engineer). Cancel with 30 days notice.
Once one agent is humming and trusted, the second is faster to ship — the foundation is reused, the team knows the workflow, the eval harness is portable. We typically ship the second agent in 60–70% of the time of the first, the third in 40%. By agent four or five, you have an operating system, not a project.
This is where the 2–3× workforce-capacity number comes from. Not from one heroic agent, but from a thoughtful set of them owning specific lanes of work.
Want to see the full pricing in one place? Investment overview →
