The new model for management consulting

The new model for management consulting

In March 2026, Sequoia published Services: The New Software.
The thesis: the next trillion-dollar company won't sell you a tool, it'll sell you the work. The models are now good enough to do the job, not just assist the person doing it.

For most professions, Julien Bek had a clear view on how this plays out. For ours, he didn't. On management consulting, a $300-400B market, he wrote:

"The interesting question is whether AI can disaggregate consulting into intelligence components (data gathering, benchmarking) and judgement components (strategic recommendations), with the intelligence layer getting automated and the judgement layer staying human. Best candidates TBD."

"Best candidates TBD."

That open question is the one we built Fendra to answer.

Can AI replace management consulting in the near term? No. Can it replace parts of a consulting project? Yes, and some parts more than others. We believe the best place to start is the data processing, where analysts burn much of their time: pulling Excel exports, joining tables, cleaning, and crunching numbers by hand. In our experience that work is heaviest in data-intensive projects like pricing and procurement optimization, where you're on transactional datasets running to millions or billions of rows. It's complex but mechanical, rules-based work, and it's where the traditional model breaks most. That's the part we automate. The judgement layer, what the numbers mean and what to do about them, should stay human.

So that's the line we draw. Machines do the intelligence. People do the judgement.

That's why we built Fendra. The platform does the data work directly, processing through those datasets and hundreds of tables comfortably, so the analyst spends their time on the business side, where they're actually worth their rate.

It still has to be validated and controlled, so the experts stay in the loop. And unlike the usual setup, we hand over our queries, so your own team can re-run and check everything themselves.

So why is this better than the traditional model?

Tokens are cheaper than analysts, and we pass the saving on. We automate the expensive part: the manual data crunching. Tokens cost a fraction of analyst hours, so the engagement gets lighter and we share that saving with you. The whole point is profit optimization, going through your transactional data to check three things: whether the logic of how you price and how you spend is sound, whether that logic is actually followed in day-to-day sales and purchasing, and whether it's monitored over time through contracts and the like. It's profit-driven work, and we want the profit to stay with you. For a mid-market company, the engagement has to be priced so the upside is clearly worth it.

We work on the full data and run more analyses, not fewer. A senior consultant carries a handful of analyses in their head, the ones they've run enough times to trust. A platform carries a whole library of them, pre-built and ready to fire. So we cover far more ground in the same engagement. And we run every one against your complete dataset, not a sample or a one-off extract someone pulled because the full file was too big to open. When you look at everything, you find what hides in the tails: the handful of customers, SKUs, or contracts quietly leaking margin that a sampled view would never surface.

Once we're connected, staying is easy. The hard part is the first connection, getting access to the data and mapping it. After that, the incremental cost of one more analysis is almost nothing. Updating last quarter's view or spinning up something we've never looked at before is a small task, not a new project. And you never have to adopt, license, or train your team on yet another piece of software to get there.

So where does this leave us?

Management consulting is here to stay, for a while at least. But automating whatever can be automated is inevitable as the foundational models keep getting better. We run as service-as-a-software, not software-as-a-service, and we think that's the model that holds until a better one comes along. Sell outcomes, not licenses. Don't force a tool onto the buyer. Automate what you can, and share the saving with the customer. Let humans do the thinking and machines do the grunt work.

You wouldn't trust an AI to set the price points that go straight out to market. You want a human on that. But you don't want that human spending three days fixing a broken Excel to get there. You want them spending those three days on your actual business problem and how to solve it. That is the Fendra way.

Fendra. AI-native services for pricing and procurement optimization.