Human-assisted consulting agents (1/n)

For the last two years, one of my guilty pleasures was asking lawyers and doctors how they were using AI, given that they were about to be replaced by it. When the question came back to me, I answered confidently. I run a data analysis consulting firm. We’ve been using predictive models since day one. We use LLMs to write code faster, generate charts, proofread documents, and rewrite emails so they don’t sound passive-aggressive. ChatGPT, Copilot, Claude, NotebookLM hourly.

I was serene. We were “AI-conscious” and AI-assisted.

Clients pay us for framing problems, choosing the right angles of analysis, and extracting insights that actually mattered. And because REKOLT is a boutique firm, the model was partner-dependent. Clients came to work with Karim and Meryem. Difficult to scale because knowledge transfer, methodology, and style were hard to replicate.

For a long time, I told myself this was our moat but I also knew it wouldn’t last forever.

Last Monday, I started using Claude Code seriously. Yes, I was late to the party. And that was the “aïe aïe AI” moment. I realized it became very hard to keep repeating the usual consulting BS about being “smarter than AI”. Our clients don’t care whether the insight came from a human brain or a machine. They want their problem solved. Faster. And cheaper.

I used to say that LLMs were our intern but Claude Code behaved like a senior consultant.

Here’s how I used it

At REKOLT, we have an offer called All You Can Analyze: We build a dataset (operational or market) -> we start with a vague business question -> and we torture the data until sharp insights emerge.

Last week, I used Claude Code as if it were a consultant on that offer.

I gave it: the dataset, told him to clean it and harmonize it, the analyses I had in mind, the client’s visual and told him to be proactive.

It generated the analyses, produced the charts, and assembled them into a PowerPoint.

On that basis, I did what Project Director or Partner does in a consulting firm : annotates slides, kills useless charts, reframes titles, asks for deeper analysis and challenges conclusions,

We iterated very fast, a project that used to take 10 days (senior consultant full time + partner half time) now takes 5 days, with circa half the cost base.

Strategy shift?

We are not AI-assisted consultants anymore. We are humans supervising AI execution.

And we won’t be using that productivity gain to increase our margins, we will pass these economies on to clients and increase throughput. More projects, more iterations, more value delivered.

The real unsolved problem

Now comes the hard part. How do we sell and market this ? How do we propose an offer where the value is higher, the delivery is faster, the cost is lower but the differentiation is harder to explain without sounding like hype?

So far, netiher AI or us haven’t cracked it.