Vendor lock-in with an external engineering team is rarely the result of a contract clause. It accumulates quietly: repositories that live in the vendor's organisation, infrastructure that only one engineer understands, decisions that were made in calls and never written down, and rights of use that were never explicitly granted. By the time a CTO wants to bring work in-house or change provider, the cost of leaving has become a hidden part of the price. Lock-in is prevented by four disciplines applied from the first week: the client owns every account, rights of use are granted explicitly, documentation is part of the Definition of Done, and knowledge is distributed rather than concentrated. This briefing turns those disciplines into a checklist.
The numbers: dependency is already the default
Most German companies already depend on outside technology and capacity. Bitkom's 2025 digital sovereignty study found that 96% of companies in Germany source digital technologies or services from abroad and that 90% of those cannot do without them. Replacing lost knowledge internally is slow: in Bitkom's 2025 survey, an open IT position took 7.7 months on average to fill, and the 2026 edition still counted around 79,000 unfilled IT roles. At the same time, the way knowledge is created is changing. Google's DORA 2025 report found that 90% of technology professionals use AI at work, with a median of two hours a day, while 30% have little or no trust in the generated code. Code that is produced faster but understood by fewer people is exactly the kind of knowledge that disappears with a vendor unless it is written down.
Where lock-in actually comes from
It helps to distinguish four kinds of dependency, because each needs a different countermeasure.
| Dependency | Typical symptom | Countermeasure |
|---|---|---|
| Control | Code, CI or cloud resources sit in accounts owned by the vendor | Client-owned accounts, vendor access by invitation |
| Legal | Unclear who holds rights of use in code, documentation and models | Explicit, comprehensive grant of rights in the framework agreement |
| Knowledge | Only the vendor can explain why the system works the way it does | Architecture decision records, runbooks, documentation in the Definition of Done |
| Concentration | One engineer holds most of the operational knowledge | Rotation, review across areas, knowledge spread as a tracked goal |
Discipline 1: the client owns every account
The simplest and most effective rule: every system that holds code, configuration, secrets, build pipelines, infrastructure or project documentation is owned by the client, and the vendor's engineers are invited into it with individual, named accounts. This includes the Git organisation, the CI system, cloud accounts and subscriptions, the secrets manager, the issue tracker, the wiki, container registries, domain and DNS settings and any third-party API keys.
This rule has a security benefit as well. When an engagement ends or an engineer rotates out, you remove access in systems you control, rather than requesting a transfer from someone else's.
Discipline 2: rights of use are granted explicitly
In Germany, computer programs are protected under §§69a ff. UrhG. Where software is created by an external provider, the client does not automatically receive comprehensive rights. Under the purpose-of-transfer principle in §31 Abs. 5 UrhG, the scope of rights granted is, in case of doubt, interpreted narrowly according to the purpose of the contract. A framework agreement should therefore grant, explicitly and comprehensively, the rights the client needs: exclusive, unlimited in time and territory, including the right to modify, to sublicense and to transfer, covering source code, documentation, test assets and configuration.
When the provider's engineers are employed abroad, the provider should also be able to show that it holds the rights it is granting. In India, for example, Section 17 of the Copyright Act, 1957 generally makes the employer the first owner of works created in the course of employment under a contract of service, unless agreed otherwise. A clean chain from engineer to provider to client is what makes the client's rights robust. Ask for a clause in which the provider warrants that chain, and clarify how open-source components and their licences are documented.
Discipline 3: documentation is part of done
Documentation that is planned for the end of a project does not get written. It must be part of the Definition of Done for each sprint item, and it must live in the repository, next to the code, where it is versioned and reviewed like code.
- Architecture decision records (ADRs): a short, numbered record for every significant design choice, with context, decision, alternatives considered and consequences.
- Runbooks: how to deploy, roll back, rotate secrets, restore from backup and respond to the most likely alerts.
- Onboarding guide: how a new engineer, on either side, gets from a fresh laptop to a running system.
- System map: current context and container diagrams, updated when the architecture changes.
- Interface documentation: generated where possible from code, so that it cannot drift.
A practical test: once per quarter, ask one of your own engineers to follow the onboarding guide and one runbook without help. Every point where they get stuck is a documentation defect, logged and fixed like any other defect.
Discipline 4: knowledge is distributed
Lock-in also arises within the vendor team. If one external engineer is the only person who understands the deployment pipeline, you depend on that person, not on the vendor. Counter this deliberately:
- Rotate review responsibility so that every area is reviewed by more than one person, including at least one of your own engineers for critical components.
- Pair a client engineer with the external team for selected sprint items each quarter, so that knowledge flows in both directions.
- Track a simple knowledge-spread indicator: for each critical component, how many people on each side could make a safe change today.
- Record important technical discussions as short written summaries in the relevant ticket or ADR.
The lock-in checklist
- All repositories, pipelines, cloud accounts and tools are owned by the client.
- Vendor engineers use individual, named accounts, never shared credentials.
- The framework agreement contains an explicit and comprehensive grant of rights of use, plus a warranty on the chain of rights.
- Open-source components and licences are listed and kept current.
- ADRs, runbooks and an onboarding guide exist in the repository and are part of the Definition of Done.
- A quarterly documentation test with a client engineer is scheduled.
- Every critical component has at least two people who can change it safely, and at least one of them is on the client side.
- No proprietary vendor tooling is required to build, test or deploy the system.
Why this is in the vendor's interest too
A provider that resists these disciplines is signalling that its retention strategy depends on switching costs rather than delivery quality. A provider that adopts them competes on the work itself. The checklist items belong in the framework discussion from the start. The checklist above is useful with any provider, and it is easiest to apply before the first line of code is written.
The notes on copyright and rights of use are general information, not legal advice.
Lock-in is prevented by daily habits more than by contract clauses. In a niogod™ pod, keeping architecture decision records and runbooks current is the responsibility of the Senior Consultant, who owns architecture and quality, so documentation is part of done rather than a project at the end. If you want an external team you could hand over at any point, that is the standard worth setting from the first sprint.
Talk to niogod™
Planning engineering capacity in DACH? Book a 30-minute consultation with the niogod™ team.
Book consultation