When your CTO says "we can build this ourselves," here's what to ask next

When I talk with CTOs about encoding the communications profession and the system we've set up for that (The Brain), they get excited and want to know about the storage, how the agentic bit works, and how we set up the AI connection. Then they often think or say: “We should build this ourselves.”

Fantastic compliment. It means they agree: we need this, and this is a very smart way to make AI work for our business.

On the surface, they're absolutely right. With a solid IT-AI team, you can wire up n8n flows, spin up vector databases, build internal APIs. Automations and storage are widely available. Yet the risk is that you’ll end up with great plumbing and empty pipes. 

Here's what separates a generic AI content stack from a genuinely strategic one: whether you've encoded how professionals actually work. What they know. How that and what knowledge flows through an organization. How messaging gets adapted, negotiated, and approved in real life.

The hard part isn't what you think it is.

The chimney sweep problem

Watching a professional sweep a chimney, it looks like ten minutes of simple work and it's overpriced. But you have the chimney swept anyway because a chimney fire or an insurance debate would create bigger problems. There is value in a clean chimney... but how much?

You could do it yourself. These guys charge 80 EUR for 5 minutes of work. But you're not paying for five minutes. You're paying for equipment, training, inspections, and the judgment about when to go through the roof versus inside the house. You're buying the hundred hours of invisible expertise behind those five.

Corporate communications works the same way. The visible output is typically text. The invisible work is knowing how to present and structure organized messaging, make sure it's aligned with themes and topics, synced with stakeholders, market context, crisis patterns, internal politics, the psychological "feel" of a brand. That lives in decks, campaign retros, leadership instincts, and half-remembered hallway conversations. But it also lives in published materials, if you knew which ones to build on and which ones to disregard.

Can you map all that into a graph that AI can actually use? Probably not.

Communications leaders do not encode 

Here's something that rarely gets said out loud: most communications professionals can't encode how they work. And they don't want to.

And I say that without a trace of criticism. Communications is an intuitive profession. The best practitioners feel their way through stakeholder dynamics, sense when messaging lands wrong, know instinctively which proof point fits which audience. Asking them to write the ultimate rule book is like asking a jazz musician to score every improvisation in advance.

IT thinks in systems. Communications thinks in feel. That's not a gap you bridge by putting them in a room together for an hour.

You need a specific type of communications expert, one who has spent years doing the intuitive work and is wired to reverse-engineer it into structures. Someone who can translate "this doesn't feel right" into entity relationships and retrieval logic. That combination is rare. We've spent over two thousand hours building ours.

So yes, your IT team can build the pipes. The question is: who's encoding what flows through them?

What encoding looks like

In the systems we build, the goal is to set up a Brain: an organizational knowledge base that knows your messaging, positioning, voice, and tone in structured form.

A few things make this work:

• Separating knowledge from instruction. Claims, beliefs, facts live in one layer. Tone rules, templates, and protocols live in another. This prevents contamination, so your tone doesn't drift and your retrieval doesn't get confused.
  ‍
• Mirroring the real semantics of the profession. Entities like actors, topics, opinions, communication patterns. Relationships like EXTRACTED_FROM or SUPPORTS. The system reasons the way a senior strategist reasons.
  ‍
• Connecting to any LLM via MCP. People keep using ChatGPT, Claude, or CoPilot. The Brain does the heavy lifting in the background.

Agents don't just push content from A to B. They execute professional workflows: assemble messaging for a specific audience from approved elements, adapt tone for a particular spokesperson, cross-check against previous statements, surface issues for human judgment.

From a CTO's perspective, this is closer to a new layer of organizational infrastructure than a point solution. A semantic canon for how your company speaks and reasons in public.

And crucially: it's built with your communications and marketing leaders, not for them.

When it works

A few things shift:

Executives stop complaining that AI "doesn't sound like us" because the system finally knows what "us" means.

The authenticity debate quiets down. You're no longer asking "Did a robot write this?" You're asking "Would the CEO sign their name to this?" And the answer is yes, consistently.

Subject matter experts can contribute to thought leadership without the weeks of rewriting. Their thinking lands in the first draft at 90% quality.

Communications and marketing stop being the bottleneck for every piece of content. Their strategic know-how is encoded once, then reused safely thousands of times.

Ask a different question

So if you're a CEO hearing "we can build this ourselves," or a CTO thinking it, I'd invite you to change the question.

Instead of: "Could we build these automations ourselves?"

Ask: "Do we have someone who can systematically encode how our organization communicates, how our subject matter experts interact with communications and marketing, and do we have our themes, topics, and messaging so sharp and well done and organized that AI can actually use it?"

If the answer is "not yet," then you're probably automating the wrong thing, with time that should be spent on something that helps your company more.

The hard part is not the tech. It's the knowledge canon.

‍