AI Lab: I have 15 AI employees. And now I spend my time training them.

WELCOME TO MY LAB ON ARTIFICIAL INTELLIGENCE.

I'm building a company where I'm the only human and every employee is an AI agent.

Follow me!

THE LAB · EXPERIMENT 05

I have 15 AI employees. And now I spend my time training them.

My company now has 16 people. Fifteen AI agents and me.

For the past few weeks, I've spent a lot of time automating almost every operation in the company.

From identifying prospects all the way to the sale. From delivering transformation projects with COLAB all the way to invoicing. Not to mention continuously optimizing processes, and even token consumption.

But I realized something.

Automating the work is one thing. Making sure the agents really know what they're doing is another.

So I started training them.

Not just with a few instructions.

I built and enriched a real knowledge base: best practices in management, sales, marketing, finance, human resources, AI transformation, regulation, technology, case studies and data on the results companies have achieved.

General knowledge, but also knowledge specific to our environment and to the way we work.

For example, our agents can now look up real transformation cases, understand the results that were achieved and draw on that knowledge to better advise our clients.

I also wrote 77 learning capsules.

These animated micro-presentations cover different topics related to business transformation.

Zoé, our agent in COLAB, can use them to explain a concept to a client at the moment they need it.

Calculating return on investment, change management, Quebec's Law 25, process automation, human resources...

It's a bit like having a consultant available at all times to explain the different aspects of the transformation.

But the most interesting part is teaching them to work together.

I enriched their instructions to spell out how they should communicate, share their knowledge, collaborate, interact with clients and bring important decisions to me.

In Kate, for example, the agents use a shared approval process. They have to prepare their recommendations, explain the impacts and determine when a human decision is needed.

I'm also adjusting their behaviours, their working methods and their interactions so they gradually become more efficient and more autonomous.

And I'm working on optimizing their usage costs. Because tokens end up costing money too!

What am I taking away from this experiment?

It's relatively easy to create an AI agent.

But building a real team of 15 agents able to collaborate, learn, make good decisions and keep improving is a whole other adventure.

And I'm starting to believe that the next big transformation won't just be automating companies.

It will be creating companies that can learn and improve on their own, all the time.

What do you think? Share your thoughts or ask me questions in the comments. Follow me to see what comes next.

Paquin & Co.: https://www.paquinco.com/en

All the AI Lab experiments: https://www.paquinco.com/le-laboratoire-ia