What happens when you turn years of consulting knowledge into a conversation?

What happens when you turn years of consulting knowledge into a conversation?

This started as a question I couldn't stop thinking about.

Consulting firms — good ones — are essentially knowledge businesses. Over decades they build up methodologies, frameworks, and ways of asking questions that genuinely work. Senior practitioners carry all of that in their heads. Junior staff spend years trying to absorb it. Clients pay for access to it.

What if you could just... talk to it?

The problem with frameworks

Most consulting IP lives in documents. Slide decks, playbooks, methodology guides. They get written, stored, and referenced — but they don't do anything. They don't ask you questions. They don't notice when your answer is weak. They don't push back when you've given them a solution dressed up as a strategy.

A good practitioner does all of that. They know which questions to ask, in what order, and when to challenge versus when to accept. That judgment is the actual value — and it's almost impossible to write down in a way that transfers.

So I started trying to build it instead.

The experiment

Working with a management consultancy, I started taking specific stages of their methodology and encoding them into Claude-powered chat agents — each one running on AWS Bedrock in Sydney, each one designed to do one thing well.

The first agent was built around outcomes discovery. The methodology has a precise way of getting organisations to define what will actually change as a result of their work — not what they'll do, but what will measurably be different. It sounds straightforward. In practice, most people default to listing activities, or describing vague aspirations. A skilled facilitator knows how to unpick that.

I spent a long time thinking about how that facilitator actually behaves. What do they ask first? How do they handle someone who keeps giving them deliverables instead of outcomes? When do they push for more specificity and when is "good enough" actually good enough? What does a weak answer look like versus a strong one?

Then I wrote all of it down, with a precision that the methodology had never demanded of itself before, and put it in a system prompt.

The second agent was built around benefit identification. Different stage, different personality. This one is supposed to be direct — almost blunt. It takes each outcome in turn, asks a structured set of questions across several benefit categories, and keeps pushing until it's confident nothing has been left on the table. "You've found four benefits here. I think there are more. Let me probe."

That instruction is now in a model. It behaves accordingly.

What the process taught me

The translation was harder than I expected — and more revealing.

Building an AI version of a methodology forces you to make explicit everything that was previously implicit. The edge cases. The judgment calls. The things a practitioner does instinctively that they've never had to articulate. You can't leave anything vague, because the model will hit the vague parts and not know what to do.

In a way, the AI build became the best documentation exercise the firm had ever done. Things that lived in people's heads for years had to be surfaced, debated, and written down precisely. The model was almost incidental — the real output was the clarity.

The other thing I didn't expect: how usable the agents actually were. Not for replacing experienced consultants, but for extending what they could deliver. A client-side team member could now run themselves through a rigorous structured process without a facilitator in the room. The quality held up. The methodology was being applied consistently, at a scale that wasn't previously possible.

Where it goes from here

The same pattern is showing up across other types of work. Proposal generation grounded in a firm's real past work and delivery evidence. Document intelligence that can answer questions about a programme using a methodology as its interpretive lens. Research tools that prepare consultants for client meetings in minutes instead of hours.

Each one is the same underlying idea: knowledge that was previously locked inside experienced people — or buried in documents that no one reads — made accessible through a conversation.

That's not a small thing. The firms that figure out how to do this well don't just get more efficient. They change what they can actually offer.

If you're sitting on years of accumulated methodology and wondering what it would take to make it work for you at scale — that's worth a conversation.