Some vendors treat handover as churn. We treat it as a valid outcome. Everything we build for you is yours: repositories, infrastructure as code, evaluation harnesses, runbooks, dashboards and documentation.
Managed run is optional. Taking the system in-house is planned, not a fight over access.
Why clients take systems in-house
Common reasons:
- Internal platform team matured and wants direct ownership
- Contract end after a successful programme
- Policy requiring certain ops roles to be employees
- Cost optimisation at scale when FinOps shows managed run no longer fits
None of these mean the engagement failed. They mean the organisation is ready to operate.
What handover includes
Code and configuration. Application, prompts in version control, tool adapters, orchestration, IaC templates.
Evaluation assets. Harness suites, test data governance notes, CI integration.
Observability. Dashboards, alert definitions, log queries documented.
Runbooks. Incident tiers, rollback, escalation, corpus refresh, model upgrade playbook.
Architecture documentation. Data flows, residency configuration, integration map, known limitations.
Knowledge transfer. Pairing sessions, recorded walkthroughs, office hours window — scoped in the increment, not rushed on the last day.
Handover is not a cliff
We recommend a phased transition:
- Shadow — client ops watches managed run or vendor ops handle incidents
- Shared — client on-call primary, we backup
- Client-led — client owns pager; we on retainer for defined hours if desired
- Full client — optional retainer ends
Duration depends on criticality and team readiness. Discovery can note expected handover shape.
No lock-in by design
No client data trains shared models. No proprietary runtime that only we can host. Standard clouds, standard formats where possible.
If something is hard to hand over, we flag it in discovery — better before sign than at exit.
IP and contracts
You own what we build for you under agreement terms. Specifics live in contract, not blog posts. FAQ covers the plain intent: your code, your data, your prompts.
Preparing your team during build
Handover fails when internal teams meet the system at exit.
Better approach:
- Name an internal technical owner during discovery
- Include them in weekly demos and architecture decisions
- Allocate time for them to run deploys in staging before go-live
- Budget training for ops and engineering separately
We can embed alongside your engineers for part of the build if you want deeper transfer.
Evaluation after handover
The harness is the insurance policy. Internal teams must keep CI gates when they change prompts or models.
We recommend retaining red-team cadence at least annually or when material changes occur — internal or external help.
FinOps ownership transfers too
Who watches inference spend after exit? Name them. Show them the dashboards before handover, not after the first bill surprise.
When managed run still makes sense
Some clients prefer ongoing evaluation cadence, model upgrade labour and incident first response on a monthly fee — especially when AI is business-critical but internal headcount is thin.
Others hybrid: client ops runs day-to-day, we retain quarterly evaluation review. Scoped explicitly.
Taking over a system we did not build
Handover article describes our default. If you inherit a pilot from another vendor with no runbooks and no tests, discovery assesses what a stabilisation increment looks like before you operate it safely.
Questions before you plan exit
- Where is the repo and who has admin?
- Can your team deploy a prompt change through CI today?
- What is the rollback procedure — last tested when?
- Who owns corpus refresh?
- What breaks if we cancel managed run next month?
If you cannot answer, delay exit until handover completes properly.
Maturity assessment before handover
Score readiness: deploy frequency, mean time to recover, harness pass rate on internal changes, ops confidence survey. Gaps become a short stabilisation scope before exit.
Handover on a date without readiness score is ceremony, not capability transfer.
Retainer and advisory models
Some clients exit managed run but retain quarterly architecture review or upgrade assistance — hours bounded, not open-ended support ticket roulette.
Scope retainers explicitly in contract when useful.
Documenting known limitations
Honest limitation docs — languages not supported, case types excluded, model behaviour edges — help internal teams set user expectations and prioritise increment two.
Hiring and skills for internal ops
Teams need blend of platform engineering, evaluation maintenance and ops discipline — not only ML researchers. Handover includes role description suggestions for internal job postings if you are building a team.
Training internal hires during hypercare beats hiring after vendor exit cold.
Licensing and third-party components
Handover lists model providers, open-source dependencies and licence obligations. Internal legal reviews vendor terms on renewal — especially API data handling clauses.
When handover is not yet right
If harness coverage is thin or runbooks untested, extend managed run or scoped stabilisation increment rather than forcing exit on a calendar date. Forced handover creates operational debt you pay with incidents.
Part of how we work
"We stay" is a why point — managed run is valuable. "You can take it in-house at any time" is equally true.
Role pages for CIOs and engineering leaders repeat both messages. Engagements page describes managed run commercials without implying permanence.
Incubics formed in 2026 to ship production AI with adult operations. Handover is part of adulthood.
Intellectual property and continuity
Code review during build is the best preparation for handover. Internal engineers who reviewed prompts and tools already know where bodies are buried.
Export lists — repos, secrets locations, dashboard URLs, provider accounts — are delivered as checklist with owners, not discovered during exit call.
Continuity planning: if key vendor engineer leaves, documentation and pairing depth determine whether client feels pain. We structure teams to avoid single-bus-factor on client accounts.
Some clients want partial handover — internal team owns ops, vendor retains quarterly eval review. Hybrid models work when documented.
Open-source licence compliance for embedded libraries is part of handover pack for internal legal.
Training recordings have retention and access policy — often same as internal LMS standards.
Post-handover support windows can be bounded hour blocks — not ambiguous "email us."
Success metric for handover: internal team deploys prompt fix through CI without vendor within thirty days of exit.
If that metric fails, handover was early.
Taking systems in-house should feel like graduation, not abandonment.
Read How we work, Engagements and FAQ for process and commercials. Select your region — India, Middle East, ANZ or Other — when you write.
Closing note
Production AI is a programme of small disciplined choices: discovery before build, evaluation before users, APIs before agents, runbooks before scale, adoption alongside code. Skip any one and the demo survives while the business outcome does not.
Incubics works that way by design — two-week discovery, production by week twelve, optional managed run. Offices in Bengaluru, Pune and the USA. Legal entity IQLEXA Technologies Private Limited.
If this piece matches a problem you are living, the next step is a conversation, not another pilot.
Plan handover from discovery onward — internal owners in demos, ops in runbook reviews, engineers in CI from mid-build.
hello@incubics.com for discovery on a new build, or to scope a handover plan if you are approaching go-live with an internal ops team ready to step in.