You are about to hire someone to set up or build an AI tool for your business. One question is easy to miss in the excitement: if this goes well, or badly, could you move away from them? Lock-in means you cannot change supplier, tool or approach without a costly rebuild. It is avoidable, mostly by settling a few things in writing before the work starts. This guide shows you what they are.
Key takeaways
- Lock-in is cheapest to prevent before you sign and most expensive to fix afterwards.
- Keep your data, settings and records somewhere you control, and make sure you can export them.
- Agree which changes your own team must be able to make without the supplier.
- Ask how the AI model underneath could be swapped, and what that would cost.
What does lock-in look like in a small business?
It rarely looks like a trap at the start. It looks like convenience. The supplier hosts everything, makes every change for you and keeps things running. That feels like a good service, and often is.
The problem shows up later. You want to change something small and it needs a paid ticket. A cheaper or better tool appears and you cannot move, because your history, settings and customer records live inside the old one. The person who built it leaves, and nobody else understands it. Each of these is a form of lock-in.
AI adds a twist. The technology changes quickly, so the tool that suits you today may not be the best choice in a year or two. Being able to switch matters more than it does with software that changes slowly.
Where does lock-in hide?
In four places. Check each one.
| Where | What it looks like | What to ask for |
|---|---|---|
| Your data | Records, documents and history stored only inside the supplier's tool | A full export in a common format, on request, at no extra charge |
| Your settings | The rules, instructions, templates and examples that make the tool behave as you want, held only by the supplier | Copies of all of it, written down, as part of handover |
| The AI model | Everything built around one model from one provider, with no way to change it | An explanation of how the model could be swapped and what that would involve |
| Skills | Only the supplier knows how it works or how to change it | Training for named people on your team, with the changes they should be able to make |
The second row is the one people forget. With AI tools, much of the value sits in the instructions and examples that tell the tool how to handle your work. If those exist only in the supplier's systems, you would have to rediscover them from scratch if you moved.
What should I settle before I sign?
Ask these questions before you agree to anything, and get the answers in writing.
- Can we export all our data, settings and records in a common format, whenever we ask?
- Will our data be stored somewhere we control, or only inside your systems?
- Which changes will our team be able to make without you after handover?
- Who will train our people, and how will we know they can make those changes?
- If a better or cheaper AI model appears, how would we switch, and roughly what would it cost?
- What happens if your business, or the provider you rely on, raises prices or stops a service?
- What do we receive at the end: documentation, settings, access details and logins?
- Is there anything we would not own or be able to take with us?
A good supplier will have answers to all of these and will not be offended by the questions. Some will have a standard handover pack. If a supplier is vague about exports or ownership, that is worth noticing early.
How do I make handover real rather than a promise?
Handover is often treated as the last day of the project. Make it a thread that runs through the whole thing instead.
Name your people early: Pick one or two people in your business to work alongside the supplier from the start, even for a few hours a week. They will learn more by watching the tool being built than from any final training session.
List the changes that matter: Before the build starts, write down the changes your business will need to make over time. For example: updating prices, adding a new type of customer enquiry, changing when a case is passed to a person.
Test it before the supplier leaves: Ask your team to make one or two of those changes themselves, without help, while the supplier is still there to watch. If they cannot, the handover is not finished.
Write it into the agreement: Make the training and the handover pack named deliverables, not extras.
Is some lock-in acceptable?
Yes. Every choice ties you in a little, and chasing total freedom can cost more than it saves. A small business using a well-known product with a clean export option is in a reasonable position, even if switching would take some effort.
The test is whether you could leave if you had to, at a cost you could bear. If the honest answer is "only by starting again", think carefully before you commit. Think hardest when the tool sits at the centre of how you serve customers.
It also helps to know whether the person advising you has a product to sell. If they resell or earn from a particular platform, that is fine as long as they say so. Ask them how they would advise if that platform were off the table.
What if we are already locked in?
You are not stuck, though moving may take some planning. Start with what you can do now:
- Ask for a full export of your data and settings, and check that you can open it.
- Write down every change you have asked the supplier to make in the past year. That list is your picture of what you would need to be able to do yourself.
- Ask the supplier to train someone on your team to make the most common changes.
- When the contract comes up for renewal, use the questions above as your conditions.
None of this needs a fall-out with your current supplier. A reasonable supplier will help, and clearer expectations can improve the relationship.
For more on what to look for before you hire, see red flags when hiring an AI consultant and questions to ask an AI consultant.