From scoping to ops —
process, spelled out
The worst thing an AI project can do is hide its process. We put every step's deliverables, customer responsibilities, and exit criteria in writing — so every stakeholder in enterprise procurement can follow along.
№ 01 · Four phases
30 min → 1 week → 6 weeks → 12+ months
- 00:300130 min
Scoping call
Deliverables
- ·Scenario brief
- ·Data readiness assessment
- ·Feasibility verdict
Your side
Business owner + tech lead, 30 min each
Exit criteria
Both sides agree on scope — if it doesn't fit, we end here and respect each other's time
- W 01021 week
Proposal & quote
Deliverables
- ·PoC scope doc
- ·Data & API inventory
- ·Milestones + pricing
- ·Success criteria (hit rate / latency / human-in-the-loop rate)
Your side
Two 1-hour discovery sessions + sample data (can be masked)
Exit criteria
Customer signs off → PoC contract
- W 06036 weeks
Ship the PoC
Deliverables
- ·Deployable agent prototype
- ·Closed-loop validation report
- ·Go-live assessment (safety / perf / cost)
Your side
Domain expert 2h/week for pairing + test data
Exit criteria
Targets met → full delivery; targets missed → we rebuild or refund
- M 12+0412+ months
Long-term operations
Deliverables
- ·Monthly runtime report
- ·Skill iteration plan
- ·Incident SLA
Your side
No day-to-day effort needed; both sides align on anomalies
Exit criteria
Customer can exit anytime with full data / model / deployment handover
№ 02 · Principles
Three lines that decide what we do and what we decline
Transparent pricing
Project-based pricing, milestone-based payment. If PoC targets aren't met, we rebuild or refund — not pushing the risk onto you.
Stable team
The product, architect, and engineers at kick-off are the same people delivering to launch. No "pre-sales team A, delivery team B".
Customer owns everything
Data, model weights, deployment, runbooks — you own it all. We can hand the project over to your internal team anytime.
