In my opinion, we see that 2025 is all about AI agents. And I don't see there's any other way of doing AI agents without having integration. There's no future without having an AI connected to your company data, to your personal data, depending on what that agent should do and make it in the perfect spot to really serve that need. Again, because we're in a great partnership,we've done some work together on how can we use AI together with Make and how can we use different LLM models inside Make.
We have a customer where we started six months ago working with them on their AI assistance. When we started, everything was coded. They had their data uploaded directly into the OpenAI databases. And by applying Make, by applying a better filter strategy, by applying a better RAG strategy, we ended up improving their assistance by 55 % more accuracy on their answers. By applying Make into this coding, normally environments, we were a lot more flexible in testing different strategy, selecting the right strategy for doing the rack pipelines, which really means we were actually testing 11 different strategies in terms of a few weeks, like two, three weeks, versus doing that in code might maybe take six months to do.
So by doing this in a visual way, we were able to define the way the AI should answer and be able to deploy that to their customers. In a matter of months, when the owner of the, of the company saw the latest results based on their testing, they were testing on a couple hundreds of questions. They were going for from 34 % to 92 % in some cases, and they were blown away
