Yes AI has captured quite a bit of press over the last couple of years, but it has been more within the last year that it appears to be finally turning into the next-gen platform from vendors forcing it down their customer’s throat to new companies being formed to provide agents for niche ecosystems. From a company that will be acquiring someone else, there is a pretty good chance you’ll get enmeshed in this as well.

In the M&A space, most of the focus is on the deal-makers and the service providers providing tools to help analyze the collection of financial docs or write the contracts that will bind the deal. Not so much for how to deal with it if you’re the one doing the acquiring or left with the implementation work. It makes sense, between the tooling, data, and access to corporate data, this is still uncharted waters.

Looking at what’s currently out there, there are at least 3 focal areas to keep in mind:

  1. The generative capabilities from the vendor using their own/public data (native ChatGPT or the various dictation/doc generation/meeting summary apps).
  2. The doc generation and analysis based on your data uploaded to the vendor (think Microsoft Co-pilot).
  3. Custom apps, data, and agents based on your data and ecosystem.

This first area is specific to ensuring there is policy alignment between your company and what is being acquired. For the data that is already out there, the proverbial horse had already left the barn.

The other two areas will be more nuanced as it will depend on how you’re interacting with the vendor, the data they have access to, where it is being indexed (or the training data generated), what are the rules behind it (ownership and access policies), and what guardrails, if any, are in place from a policy and enforcement standpoint.

If the training data is local to one company and not the other, then there will need to be some considerations as to what it would take to ensure the models accurately reflect the data for both organizations. When divesting, then how do you carve out and integrate just that piece?

Needless to say, there is quite a bit to consider. From the integration standpoint, you will be looking at the retraining of the models and finagling the agents and vendor hooks to use the new bits. The rest will be very much policy heavy and dependent upon not just the terms you use for the deal and the TSAs, but also may entail creating or updating your own company process and policies.

cab