The Nano Method: How AI Systems Actually Ship

Typical duration: 1–2 weeks — this is the AI Readiness Sprint. We run stakeholder interviews, a data and systems access audit, use-case scoring, an ROI model, and a risk and compliance screen against PDPL and sector rules.

You walk away with an AI Readiness Report, a prioritized use-case matrix, an ROI model, and a risk register. The stage ends when one pilot scope is signed with written acceptance criteria and an eval plan.

Typical duration: 2–6 weeks. We build version one with instrumentation from day one, run golden-set evaluations in Arabic and English, red-team the guardrails, and open the system to limited real traffic.

You receive an eval report (version 1), access to a staging environment, and weekly build notes. The stage ends when the eval threshold on the golden set is met, the guardrail checklist passes, and the client signs off on user acceptance testing.

Duration: 30 days in production. The system runs on real traffic with full measurement, humans stay in the loop per the agreed policy, and you get weekly numbers — calls answered, carts recovered, notes drafted, whatever matters to your business.

You get a live monitoring dashboard you can log into yourself, plus a day-30 outcomes report — for example, "214 calls answered that you would have missed, 37 appointments booked." The stage ends when the SOW's acceptance criteria are met, satisfying the guarantee, and a go/no-go decision is made for Scale.

Typical duration: 2–8 weeks. We extend the system to more branches, channels, or languages; deepen integrations with your CRM, calendar, or EHR; and train your staff on the new workflows and standard operating procedures.

You receive an integration map, a recorded training session, and a runbook — what the system does, what to do when it misbehaves, and who to call. The stage ends when every in-scope unit is live and the error budget has been stable for two consecutive weeks.

Ongoing, under an ops retainer ($1,000–3,000/month). This covers monitoring, re-running evaluations on every model or prompt change, versioned prompts, incident response under a written SLA, a monthly report, and a quarterly business review.

You receive a monthly ops report, an incident log, and a quarterly business review deck with next-quarter recommendations. This stage has no fixed exit — it is governed by the SLA and renewed annually.