:focal())
Today Anthropic, the company behind Claude, goes public. Peter Verrykt, Business Director Professional Services at Xylos, shares what he sees of this with customers and in his own workplace. And why AI in your company needs an architecture and a policy today.
A lot has already been written about Anthropic's IPO, and even more has been assumed about where it is heading. To me, it mostly feels like confirmation of something we have been seeing up close for a while, with our customers and in our own workplace.
Convincing then, keeping up now
In the past, you still had to convince colleagues to try AI, and sometimes it fell short of expectations.
Today that has flipped. Everyone gets access to models that go far beyond summarising a text or preparing an email. Those who dare to push a little further suddenly see an enormous amount become possible. With a few good prompts, applications seem to do what we had long failed to get from IT.
And where that access does not come officially? There, employees sort it out themselves. With their own credit card, and neatly expensed afterwards.
What that means for your company
For your company, that means:
Customer data in tools nobody knows.
Costs you only see in the expense run.
Unanswered questions. Which model is used for what? Where is your data processed? And is it used to train a next model?
Anyone can call a model
A hundred people doing that at the same time, each with their own goals, permissions and budgets, is another story. Then you need an architecture and a policy. That starts with four questions:
Who is allowed to use what?
Which data may go where?
Which request needs a large model, and which can a smaller, cheaper model handle?
What do you do when a provider goes down or raises its prices?
How we tackle it ourselves: our AI gateway
We have been working on this at Xylos for more than two years. The biggest piece of work in that period is the development and delivery of our AI gateway. All AI traffic in our company runs through it. It does three things:
It checks what comes in and what goes out.
It manages who may use which model and which tool.
It sends every request to the model that fits.
Far from finished
With the speed at which AI is evolving today, we are of course not done with it. Every new model and every new feature shifts something somewhere.
Xylos is an Anthropic partner, and that indeed means we have certified Anthropic developers in house. Nice, but for me the value of the partnership lies mainly here: we see early what is coming and absorb it in the gateway, before it ends up with the first colleague using a private card.
Do you know what is happening in your company?
An IPO puts Anthropic in the spotlight of many more people once again. For our customers, the question stays the same: does AI happen in your organisation with overview and control? I'm curious how that works for you. Do you know today which AI tools are used in your company, and what for?
Curious how our architecture answers both governance and cost structure? Let us know, we are happy to take a look with you.
About the author
Peter Verrykt is Business Director Professional Services at Xylos. He helps organisations turn data into concrete business value. He helps companies look beyond technical implementations and uses data and AI as a foundation for better decisions, greater agility and sustainable growth.