Clients often ask what happens after they sign, before the first line of code lands. Week one is deliberately boring in a good way: clear access, a written plan, and a rhythm you can forward to your co-founders without another meeting.
I run business operations at PAI Technologies, Upwork contracts, EU onboarding, weekly status, while Gyanender’s engineering lead owns technical milestones. Rekha joins for strategy and scope questions on larger engagements.
This is the same playbook we use for Gordon (M-agi-c Solutions), TerraSenti / MythraCore hardware, and PlaywithDB builds: adjusted for GDPR, HIPAA, or lighter commercial scope.
Day 0–1: Kickoff and access
Kickoff call (30–45 minutes): goals, out-of-scope, compliance constraints, and who approves scope changes. We confirm the single engineering lead accountable for delivery.
Access checklist sent same day: repo (or greenfield invite), staging if any, sample data or anonymised exports, API keys via your secret manager, never email. For hardware: KiCad project, BOM draft, fab preferences.
Commercial: Upwork milestone or direct SOW, timezone overlap windows, and escalation path to Rekha or me for contract and scheduling, engineers stay out of rate negotiations.
Day 2–3: Written plan and architecture sketch
You receive a one-page plan: milestones, acceptance criteria per slice, risks, and subprocessors if EU. For AI: retrieval approach, model choice, eval plan. For agents: tool list and logging.
Architecture sketch is not a 40-page PDF: diagram plus bullets: data flows, region (e.g. eu-central-2), and what we will demo in week two.
If you sent a strong brief (see Rekha’s article on briefs), this step is fast. If not, we still ship the plan but may narrow week-one scope to discovery and a spike.
Communication rhythm
Weekly written update every Friday (your morning / our evening depending on zone): done, next, blockers, decisions needed. No surprise demos.
Loom when visuals matter: UI flows, hardware assembly, eval dashboards. EU clients often prefer async video over late-night calls, we record in English with chapters.
Slack or email, your choice. We do not require daily standups unless you want them; we do require timely answers on blocking access or scope questions.
End of week one: first demo
Typical week-one demo: deployed spike, retrieval on sample docs, PCB schematic review, or n8n flow calling Claude on test data, depends on engagement type.
We show what we measured: latency, retrieval hit rate, DRC summary, or automation success rate. Honest gaps are listed with options for week two.
You leave with a recording, written notes, and explicit asks if you owe us data, legal sign-off, or design feedback.
EU and regulated clients
GDPR: DPA alignment, data minimisation in staging, region locks documented before production traffic. Gordon-style work uses EU regions and subprocessors lists clients can paste into vendor reviews.
HIPAA or PCI: we narrow week-one data to synthetic or client-redacted sets until BAA or scope is signed, no “just use production for speed.”
Rekha or I schedule legal/IT calls when needed; engineers join only for technical subprocessors and logging questions.
What you can do to make week one smooth
- Name one decision-maker for scope and one for technical access.
- Send anonymised sample data early, real shape beats perfect data late.
- Share brand/design constraints before we polish UI.
- Block 15 minutes Friday for written update review.
- Tell us your launch or investor date, we sequence milestones backward from it.