Adoption

Adopting AI isn’t giving everyone Copilot

We have just started Astrolabe, and we were lucky enough to do it as two people who are native to AI, with superpowers: we have a swarm of bots that does a lot of repetitive work for us, and we call it los galácticos. They run social media, help with accounting, summarise what happens in our inbox each day and book meetings in our calendar. There are departments we haven’t created yet because that swarm runs them. Why we call it that we will tell you in another article, because it is quite funny. Our company is digital-native and AI-native: it is connected to Foundry and Notion, everything feeds into the agents’ context and everything is modelled in an ontology. That lets us grow much faster and spend our time on what matters, which is making the real decisions and creating value for our customers.

Transferring that to a large company doesn’t happen overnight, and we understand that. Adapting a company to this world is a titanic task: you have to decide how much to invest in a paradigm that can change with the next model capability, there is uncertainty about jobs, and there is a huge cultural side, with employees, middle managers, leadership and customers who sometimes don’t want to talk to a robot and sometimes prefer to. The revolution is moving very fast, but change in companies lags behind, and we are right in that gap. The usual answer is to hand out ChatGPT or Copilot licences and hope people stop doing repetitive tasks. The problem is that it makes it very hard to measure the return and the quality of the investment. Changing a company takes years, but proving that AI is worth something has to be done as soon as possible, because cultural change needs an example and, if it takes too long, the solution is out of date.

What we propose is to give robots the decisions that are robots’ decisions, without human supervision whenever possible, and to leave the important decision to a person, but pre-digested: with all the pros and cons and with executing it as frictionless as possible. And the most important thing of all is traceability. For a person to decide well, they need to see every decision the AI has taken underneath and dig into them until they understand how it got to this point. That is why those automated decisions are also part of the ontology. It is what makes it worth letting robots perform many automated operations: you can always review what they did and why. If every AI step needs a human to supervise it, the bottleneck is the decisions themselves. That is why, when we map your processes, we decide with you which ones are critical for the business and must be taken by people, and where operators have reasoning that can be automated. Those decisions are modelled in the ontology as actions, measured and compared with control groups, so you get one more decision layer that tells you how profitable it was to automate each one. Which AI to use for that, commercial or sovereign, we covered in AIP or n8n. This is the part that can change your company the most, the one we have worked on the most in ours and the one we most want to accompany you through: getting our boots muddy, respecting your idiosyncrasies and setting you the challenge needed to reach the best decisions.