Skip to main content

Colos: Collaborative AI for consultants on strategic projects

 by Felix Speulman

Colos was founded just over a year ago by former consultants Jester Koldijk (Boston Consulting Group) and Niek Simons (McKinsey & Company), and later joined by engineer Poonam Mishra. The company is building an EU-first AI platform that enables consultants and strategy teams to carry out their projects in collaboration with agents. Colos was recently nominated for the Tech Hero Award, which will be presented on November 24 at CIODAY.

What do you offer that others don’t yet have?
Jester Koldijk: “We’ve built a collaborative AI platform specifically for consultancies and for strategy and transformation teams. We tailor that platform to each company’s unique knowledge. We help clients build that knowledge and keep it up to date so they can use it in their work.

Our offering stems from three insights gained from our work with clients. First, we strongly believe that you should work based on your own knowledge and expertise. AI is certainly capable of performing tasks on its own, but the results must align with your knowledge and context. This is particularly important in consulting work. For legal work, for example, you can easily source knowledge from outside, such as case law. Consultants, on the other hand, work with what’s happening within a company: their insights from previous projects and their own data.

We capture that knowledge in what we call the “Company Brain.” Within it, we distinguish between the “how” and the “what.” The “how”—the way a company performs a particular task—is documented in playbooks. The “what” consists of insights from previous projects and the consulting firm’s own data.

Second, we believe that employees can do the work together with agents. To do this, we use a structured, proven problem-solving process: defining, structuring, and analyzing the problem, then communicating the outcome. At each step, the consultant delegates work to an agent, such as conducting research, holding interviews, and creating slides. The agent also asks critical questions at each step to ensure the quality of the result.

Third, we see at many companies that junior consultants struggle a bit more to learn because AI can take over much of their work. A junior’s learning process, however, involves trial and error: trying things out, learning from them, and understanding what works well and what doesn’t.

Consulting also has its own way of working. Just like a contract, a consulting deck has a specific structure, a specific look and feel, and specific terminology. What we’re seeing now with AI is that juniors often produce many pages that don’t really say much. Seniors spot this immediately because they have the experience and have gone through that learning process themselves. For many juniors, this is more difficult: with AI, they get something that seems logical and sounds good, so they use it. Then they’re asked tough questions by seniors or clients and don’t really know the answer because they used Copilot. That’s a risk, especially for consulting firms and transformation teams. It’s very important that what’s said is accurate.

We help with that by conducting joint work sessions and supporting the review process. In addition, the agents provide juniors with feedback at every step of the process.”

What’s the biggest challenge you’re facing right now?
“I can think of two things. First: how do we effectively gather all the knowledge that resides in the minds of people within organizations? And how do we store that knowledge in a logical and accessible way so it’s easy to use? That’s a major technical challenge. It also varies by domain: a strategy team, a consulting firm, and a transformation team operate in a highly project-driven manner. That’s very different from, say, a marketing organization.

The second challenge is keeping up with all the latest developments. That’s why we’re also optimizing everything internally and working with agents as much as we can, so that our own development and change process moves as quickly as possible.”

What does your AI landscape look like?
“We built our harness—the layer that controls the AI agents—entirely in-house. So it’s part of our IP. A very important aspect of this is that we’re model-agnostic: in principle, we can integrate any model. We also conduct evaluations to determine which model is best suited for which task.

That model independence has two major advantages. The first is cost efficiency. The second is that it makes it very easy to keep pace with the market. For example, you’re seeing a lot of open-source models now. They’re a lot cheaper, and you can train them yourself because the parameters are public. If you train such a model for your specific context, you’ll also achieve the desired quality.”

What are you proud of?
“I think I’m most proud of how we help clients. That we get people moving and encourage them to think about how to use AI. You can see that people are also struggling with it a bit: yes, they can do certain things much faster, but they shouldn’t always want to do them faster. We help them have the right discussion about that. We guide them through that process, train them, and help them grow along with us. In doing so, we ultimately help companies with their transformation process.”

Connect with your fellow
Digital decision-makers.

CIODAY is the annual network event for digital decision-makers in business, government, and everything in between. Organized by ICT Media B.V. in association with CIO Platform Nederland
contact
info@ictmedia.nl
+31 (0)73 614 00 70
Magistratenlaan 60
5223MD 's-Hertogenbosch
© 2026 All Rights reserved