AI tools in Business Central support: what does it mean for you?

More consultancies use AI tools in their support work. That can make the work faster, but it raises questions: what is sent where, who takes responsibility for the answer, and what do you actually save? This article goes through what AI is typically used for, what a person still has to do, and which questions about data and responsibility you should have answered. It is written for those who buy support, not for developers.

What AI is typically used for

AI tools are often used to search documentation, suggest an explanation of an error message, draft replies and summarise long cases. Some partners also use them to write code or run tests.

They are aids. How well they work depends on the tool and on how the consultant uses it. Ask the partner which tools they use and for what.

Ask the consultant to tell you whether AI is used to write code for you. If so, the code must be reviewed and tested like any other code, and you should know about it.

What a person still has to do

A consultant must understand your setup and your processes. An AI suggestion can sound right but be wrong for your environment. So a person should read the suggestion, check it against your setup and take responsibility for the change.

Particularly for changes that affect posting, VAT or inventory, you should expect a person to have reviewed and tested the solution before it runs in production. Test in a sandbox first if in doubt.

Ask how the partner ensures quality: review by a colleague, testing in a sandbox and a plan for rolling back. These are the same requirements that apply to changes without AI. AI does not change who is responsible towards you.

Ask the consultant to state in the case if a suggestion comes from an AI tool and what has been done to check it. That makes it easier to trace an error afterwards.

Questions about data

Ask for the answers in writing, and read them together with the data processing agreement. If the partner cannot answer precisely, that is information in itself.

If a consultant pastes parts of your case into an AI tool, it may contain personal data or commercially sensitive information. Ask:

Which AI tools are used, and who supplies them
Is your data sent to an external service, and where is it located
Is data used to train models, and can that be switched off
Are the tools listed in the partner's data processing agreement as sub-processors
How is it ensured that personal data and customer data are not pasted in without reason
Can you prohibit AI on certain cases

What you can expect to save

Do not let a promise of faster answers turn into a lower price without an agreement. Ask how any time saving benefits you: lower hours used, faster answers or better documentation. It should be in the agreement and not only in the sales material.

Microsoft has its own AI features in Business Central. Check Microsoft Learn for what is available in your version and region and what requires activation.

It can also be an advantage for you that answers are written clearly and in a structured way and that earlier solutions are found quickly. Ask whether you get a share of the documentation that results.

Agree also what happens if the AI tool gives a wrong answer that leads to an error on your side. Who fixes it, and who pays for the time? That should be clear before it happens.

Data processing and legal matters

Have your own adviser assess what is required in your situation. See also the article on data processing agreements.

At Addverk

If you want to know how we work with support, you can read more on the page about services and prices. Ask us the same questions as listed above.