Enterprise onboarding in 30 days

Every onboarding promise is a lie until you actually deliver it. We promise 30-day onboarding, and we mostly keep it. Here is the week-by-week recipe, plus two outliers: one where we went live in 9 days, and one where 30 days became 11 weeks.

The 30-day structure

Week 1 — data audit + tenant provisioning

• New tenant row in the central backend's Postgres under FORCE RLS. Slug, legal name, host, billing settings.
• Sealed Secret rotates on NIP (Nortinia Internal Platform); Flux pushes it to the nodes automatically.
• We run a data audit against the customer's existing systems: what ERP, what API surface, which identity provider, how we want to SSO.
• Output: a signed scope document and the tenant in provisioned state on the central backend.

Week 2 — storefront branding + first webhook

• A new storefront app gets deployed (Next.js 16, scheduled to one of our XCP-ng hypervisor nodes as a Kubernetes pod).
• The customer's logo, palette, typography land in the tenant-config JSONB.
• The first webhook integration ships: usually the customer's existing ERP → us. Of 14 design partners, 11 had ERP as the first integration; 3 had CRM.
• Output: working storefront at <customer>.nortinia.com + live ERP webhook.

Week 3 — first workflow + first AI assistant tour

• We set up the first Nortinia Engine workflow for the tenant. Usually an order-processing flow or a quote-routing flow.
• The AI assistant (chat-panel) ships to both the storefront and admin. The customer's staff meets it in a 90-minute guided tour. SITE_NAV MCP tools introduce themselves.
• Output: a live workflow, and the staff knows what the chat agent is and what it is good for.

Week 4 — training + go-live

• Two role-specific trainings (admin, operator), 3 hours each.
• Go-live checklist: all monitoring green, all webhooks have passed test orders, graceful shutdown rehearsal passed, Sentry project active, backup pipeline triggered.
• Cutover.

The 9-day outlier

A logistics customer came in October 2025 with a working in-house ERP that handed us a clean OpenAPI 3.1 spec. They wanted SSO via Microsoft Entra ID, which is standard for us. Branding was minimal. Their workflow was an almost 1:1 mirror of our existing order flow. Two roles, eight users.

We went live in 9 days. Not because we were fast. Because every input was unambiguous and there was no internal politics on their side.

The 11-week outlier

A financial advisory firm whose audit-trail compliance was so strict that every mutation needed an internal legal review first. By week 4 we realised their audit_event masking needs were deeper than the regex+ML masker we had built. We needed a custom rule set. Then in week 6 they asked for a never-built GDPR data-export pipeline ("give us every data point on a user as one CSV within 30 days"). The last three weeks were writing a custom job runner that walks Postgres by hashed user-id.

The lesson is not "don't take financial customers". The lesson is: ask for the compliance document in week 1, not week 4.

What we made non-negotiable

• The 90-minute tour is mandatory. Tenants who skipped it generated 5x more support tickets two months in.
• The week-1 scope document is signed. No signature, no deploy.
• Sealed Secret rotation runs from a checklist. Not manual. We have not had a secret leak in six months.

How many onboardings have we run

14 design partners + 9 paying customers = 23 onboardings in 2026 so far. 19 stayed within 30 days. 2 outliers (9 days and 11 weeks). 2 currently in flight.

The pipeline is predictable now. That does not mean every tenant looks the same. It means the pipeline knows what to ask them in week 1.