AI and the cost of switching later, assessed before you commit

5 days ago
3 min read

Introduction

In a stable software market, lock-in is a manageable risk because you are unlikely to want to move. In this one, the assumption should be reversed: the product you choose this year is quite likely not the one you would choose in two, because the market is moving faster than the useful life of a purchasing decision.

That makes the cost of leaving a first-order consideration rather than a detail in the contract review. Two products with identical capability and price can differ enormously in how easy they are to abandon, and that difference is worth more than most feature comparisons. It is also entirely knowable before you sign, and almost nobody asks.

1. AI and the cost of switching later deserves weight in the decision

Treat it as a criterion.

Score it alongside capability and price. In a market where you are likely to want to move, the ability to move is a feature, and the product that scores slightly lower on function may still be the right choice.

2. Ask how you get your data out

The first question.

What formats, how completely, how quickly, and whether it includes everything or only the records you originally uploaded. Test the export during a trial rather than taking the answer on trust. An export that produces a proprietary archive nobody else can read is technically an export and practically not one.

3. Check whether the output is portable

Frequently overlooked.

Configurations, rules, templates, trained categories and prompts represent months of accumulated work. If these cannot be exported in any usable form, leaving means rebuilding them.

4. Count the integration work you would repeat

Where the real cost sits.

Every connection built into your accounting, scheduling or customer systems has to be rebuilt. This is usually the largest component of switching cost and it grows with each integration added.

5. Consider the retraining

The human side of lock-in.

A team fluent in one product loses time learning another, and adoption of the replacement starts from zero again. Deep familiarity is a real switching cost and it is the one that makes people reluctant to move even when they should.

6. Watch the contract term and notice period

The contractual component.

A three-year term with a ninety-day notice window means the practical earliest exit may be two years away. The discount for a longer commitment is a payment for accepting that, and it is usually a poor trade here.

7. Prefer standard formats and open interfaces

The structural protection.

Products that store data in recognisable formats and offer documented interfaces are easier to leave than those that do not. This is a technical question with a straightforward answer and it is worth asking.

8. Avoid depth of dependency you do not need

A design choice.

The more processes route through one product, the more expensive leaving becomes. Keeping the dependency proportionate to the benefit is a deliberate decision, not a default.

9. Reassess at each renewal

Where the option is exercised.

The alternatives, the pricing and your own requirements will all have changed. A renewal treated as automatic is a decision not to consider a market that has moved substantially since the last one.

Be careful about the opposite error. Optimising entirely for portability can mean choosing a weaker product and never getting the benefit at all, and a degree of lock-in in exchange for something that genuinely works is a reasonable trade made knowingly.

Conclusion

Score the cost of leaving alongside capability and price, because you will probably want to leave.

Establish exactly how data is exported and test it during a trial, check whether configurations, rules and templates are portable, count the integration work you would have to repeat, allow for the retraining, examine the contract term and notice window against how quickly this market moves, prefer standard formats and documented interfaces, keep the depth of dependency proportionate, and treat every renewal as a real decision.