Agency AI client work rescue — escape Base44, Bolt, Lovable

Why agencies get locked into AI builders

The lock-in is not accidental. Agencies pick Base44, Bolt, or Lovable because the first client project ships 3x faster than building from scratch in Next.js. The margin on a fixed-price client engagement is material when the scaffolding is free. The second and third client projects go the same way. By project five, the agency has a portfolio of client apps living on a platform it does not control, with source code that cannot be exported cleanly, data in a database owned by the platform vendor, and no way to guarantee the uptime its client SLAs require.

The tipping point is usually a client compliance review. A healthtech client asks where the PHI lives and discovers the AI builder is not BAA-compatible. A UK fintech client asks about data residency and discovers the platform is US-hosted. A public-sector client asks for the source repository and discovers the agency cannot deliver a standalone build. At that point the escape is not optional — it is the next engagement on the backlog, and the agency either does it carefully or risks losing the client. See the Base44 developer hub, Bolt developer hub, and Lovable developer hub for the platform-specific failure modes.

What client contracts actually require (and why proprietary builders violate them)

Three contract clauses routinely trip up agencies working on AI builders. The first is source-code ownership. Most client engagement contracts include a work-for-hire clause transferring all source code to the client on acceptance. An agency cannot transfer source code it does not fully own, and AI builders retain some share of the stack — the visual editor, the managed database connectors, the deployment runtime. At minimum the agency is delivering a codebase the client cannot self-host.

The second is data residency. UK, EU, and increasingly Asia-Pac clients require customer data to live in specific geographies. AI builders default to US hosting with no per-customer regional option. Agencies delivering into regulated verticals (see the fintech vertical hub and healthtech vertical hub) cannot honour data residency clauses on a US-only platform. The third is the audit log: enterprise client contracts require seven-year retention of access logs, schema changes, and administrative actions. AI builders log for their own support purposes, not for client compliance, and the logs are usually not exportable.

When a compliance review catches any of the three, the agency has to escape or lose the account. The escape engagement is specifically designed for that moment — fast, fixed-price, zero-downtime migration to a Next.js codebase the agency fully owns, with data in an infrastructure the client can audit.

Our agency migration programme

1. Discovery. Week zero. We inventory every route, every integration, every database table, and every cron job on the AI-builder version of the client app. Output is a migration spec the client can sign off on before we write a line of code. No surprises at cutover.
2. Parallel build. Weeks one to three. We build the Next.js project in parallel against the live AI-builder version, matching routes and behaviour one to one. Data is mirrored from the source database via read-replica or nightly export, so the Next.js project is always within 24 hours of production.
3. Cutover. Scheduled weekend window. DNS flips, final data migration, and smoke tests. The client sees continuous service; the agency moves from renting the stack to owning it. Four-hour monitoring window after cutover to catch any edge cases.
4. Handoff docs.A written runbook for the new stack (deployment, secrets, database, Stripe, auth), a per-module developer guide, and a recorded walkthrough. The agency’s in-house team or a follow-on retainer takes it from there.

The programme is built to match the app migration service shape but scoped specifically for agencies migrating client work. Multi-client engagements get a volume rate; the retainer support service covers ongoing engineering after the escape is complete.

Agency escape pricing

Fixed price from $9,999 for a single client-app migration of typical scope (10-30 routes, standard Supabase + Stripe integration, Vercel hosting). Larger apps with custom integrations, multiple tenants, or regulated data are quoted between $15,000 and $35,000 after discovery. Multi-client engagements — migrating three or more client apps off the same AI builder — run on a volume rate.

Optional retainer after cutover: $3,499 per month for ongoing engineering against the migrated stack. The retainer is the bridge between delivering the escape and handing back to the agency’s in-house team, and it covers maintenance plus one significant feature per month. Most agencies keep the retainer running for three to six months after cutover while they onboard the new stack internally.