We bought AI licences, now what? How an organisation becomes AI native.

What we see time and time again is that companies have bought tools and been pitched on the fact that those tools are going to make them AI-native on their own. They see really good adoption metrics, with lots of people using them, but they aren’t necessarily using them to actually get work done. Where we see the massive unlock is when you can automatically bring not just the users to the tools, but also the context the user needs.

One of the reasons this is less of a bottleneck with developers is that they are technically adept enough to figure out how to do this for themselves, whether by creating MCPs or using other methods. Also, most of the context a developer needs lives right in a repository. They are a very unusual type of knowledge worker because they have access to basically all the information they need in one place.

For most other knowledge workers, the information and context they need sits across multiple systems. It is spread across knowledge management systems like SharePoint, email, spreadsheets, ERP systems, and other tools.

When we think about adoption, we like to focus on an end-to-end workflow. We aim to prove that we can bring all the context together to satisfy that workflow in a way that is 95% to 99% automated.

This means not requiring the user to go off and hunt for information halfway through the workflow. It also prevents the workflow from making up stuff or making assumptions just to fill gaps. While total hallucinations have partially disappeared, the kind of hallucination where the AI makes “reasonable” assumptions through a complex process is quite dangerous because it gets difficult to check and fix.

When we think about that enablement journey, it is primarily about giving users permission to dream big, ensuring we have the right infrastructure and connectors in place, and enabling end users to capture and share skills. Rather than relying on a central enablement team that tries to understand every single part of the business, end users can build skills that reflect their actual business needs.

They can iterate on those skills with central supervision for security and data privacy risks, without being forced to figure out the underlying infrastructure for sharing skills. It can be very intuitive and easy without creating tons of duplicate files, ensuring that teams are able to share things centrally.

Harriet brings together a unified AI platform, a deployment team and a strategy so that your company becomes AI-native.