Back to Perspectives

The Consulting Pyramid After AI: Research Without a Junior Bench

18 November, 20269 min read
Jan Beránek

The junior pool used to set what an engagement cost and how many a firm could run. Once research runs on the firm's own checked method, a senior-only boutique can take on breadth that used to need a bench – and still bill day rates.

A boutique consultancy can now take on research-heavy work without a junior bench – as long as its own method is encoded and every answer is checked before a senior reads it. In the consulting pyramid, price ≈ hours × juniors per partner, and the same juniors largely decided how many engagements the firm could run at once. That model is under strain: billable utilization in Deltek's professional-services benchmark has dropped to 66.4%, its lowest point in nearly two decades, while headcount still grew 2.8% in 2025 (Deltek).

My take is 1 consultant, many agents, productized methodologies and data. On calls I put it more bluntly: the army of juniors are the agents. That makes the senior-only boutique, the firm that never had an army, the one best placed for this shift. None of this means hiring fewer juniors. It means a different shape of firm, and a longer list of work a small senior team can say yes to.

How the consulting pyramid set price and capacity

A partner at one consultancy put it simply on a call with us: the junior pool used to define what an engagement cost, and that is going away.

I know the pyramid from the inside. We spent fifteen years in the consulting chair, much of it as U+, the corporate venture builder FifthRow grew out of. You scale a consultancy by adding consultants, and at some point it's basically doing acquisitions. That was not a life I wanted to have.

The large firms are adjusting, and not all in the same direction. Around 1,300 graduates and school leavers joined PwC UK in 2025 (PwC), 200 fewer than the year before, and its UK head wrote that "graduate intakes everywhere are under pressure" (Fortune). McKinsey, meanwhile, hired about 12% more consultants in North America last year and expects growth "just below that number" this year, Eric Kutcher said in October 2026 (Bloomberg via Yahoo Finance). Both are still working out what juniors will do.

What changes when research runs on your own method

When research runs on the firm's own method – its questions, its trusted sources, its order of steps – capacity stops tracking headcount. HBR's authors call the new shape an obelisk, "a tall, narrow model: fewer layers, smaller teams" (Harvard Business Review). Others call it a "diamond" – "wider in the middle and top, thinner at the base" – or a flipped pyramid, with "more senior and mid-level consultants with minimal junior support" (Consultancy.uk). A senior-only boutique is close to that last shape already. Much of the research junior hours used to pay for can now run in minutes or hours.

Two conditions. First, the method has to be yours, written down step by step – a generic prompt gives output every competitor, and every client, can produce. Second, the output has to be checked before a senior sees it, because agents break. They can hallucinate, one agent can't do the whole job, and even when it does, the quality isn't consistent. If your senior has to check every citation, your senior has become the bench. How to verify AI research covers the checks.

Why the boutique gains more than the pyramid firm

First – there is no bench to keep busy. In a pyramid firm, a research hour that moves to a system is often a junior hour no longer billed. A boutique doesn't have that conflict – for it, the same system is simply more capacity.

Second – the senior part is already there. The Consultancy.uk piece argues that "clients want senior judgement, not junior analysis" (Consultancy.uk). In practice: a partner at a brand consultancy told us their clients struggle to know when they need behavioral insight, when a predictive model and when human research. Getting that call right is senior work. What held senior-only firms back was breadth – a twelve-market scan or a weekly read on every client needed people they didn't have.

Pyramid firmSenior-only boutique with an encoded method
What sets capacityJuniors per partnerHow often the method can run, checked
What sets revenueJunior hours times ratesSenior days on more client cycles, fixed-fee phases, standing services
Where research happensJunior analysts, over weeksThe firm's own method, in minutes or hours
What seniors spend time onReviewing junior draftsFraming and judgment on checked evidence
How people learnBy doing the researchBy designing, running and challenging the method

What a boutique can take on now

The clearest case we have is a boutique global brand strategy consultancy. It is expert-led with no junior research bench, and its own leads put discovery research at 70 percent of the work on a project (FifthRow proof). Its four-lens discovery method now runs as a system. Given only the brief for an engagement its strategists had already worked by hand, it returned the same lead territory they had taken to the client, in 48 minutes, with 597 source-verified questions answered in that run.

That gave a small senior team breadth it could not staff before:

  • One discovery framework reused across ten engagements and pitches, instead of rebuilt for each.
  • 70 client and prospect accounts reviewed every week, unattended.
  • A decade-horizon scenario study that runs in an afternoon instead of occupying a team for weeks.

Nearly half of all the research the firm now gets arrives without anyone asking for it (FifthRow proof). It arrives already checked: quality control is a judge agent that reads every output (FifthRow), so a strategist opens a reviewed read on a client's market, not a pile of links.

What do you bill for when discovery takes 48 minutes?

Most boutiques bill by the day. The day rate doesn't have to go – what changes is which days you sell. Three options a day-rate firm actually has:

  • Keep the day rate, sell more client cycles. On our page for consultants we call it "More client cycles per year": "Hours that used to go to production and fact-checking return to seniors." (FifthRow) The same senior team runs more engagements.
  • Price discovery as a fixed-fee phase. Sell the evidenced read and the territory it supports, not the weeks it used to take. A method that runs the same way every time is easier to price than a team's time.
  • Sell what happens after the project. A standing weekly read on a client's market is the service the brand strategy firm built – in its own words, "a reason to go back to a client after a project ends" (FifthRow proof). That is "Offers you can run again" (FifthRow).

Our post on how to productize consulting services with AI covers pricing a productized offer in detail.

What doesn't change

The client relationship, the framing of the question and the final recommendation stay with people. The brand strategy firm drew that line itself: its route from research to a positioning territory has "a human decision point between each one", because "direction-setting is where its strategists earn their fee" (FifthRow proof).

My version of it: it's never one thing and done. It's one thing, then some human judgment, then another thing and some human judgment.

The bottleneck moves, though. Once research stops being the constraint, the constraint is how many good decisions a partner can make in a week. I feel this personally – the agents finish, and the queue of decisions lands on me. Without changing how the firm decides, a lot of the new capacity just turns into a longer inbox.

Frequently asked questions

Can a small consultancy compete with the Big 4 on research-heavy work?

On research breadth for a defined question, yes – as long as its method is encoded and the output is checked. Large firms keep implementation capacity, global reach and the name a nervous board likes to see. The research gap is far narrower than it was.

Will AI replace junior consultants?

No – AI changes who does what. McKinsey's North American hiring is still growing, and the brand consultancy partner above told us they use AI to augment their advisory work, not to replace it. The way I'd put it to a junior today: you will no longer do this work, but you need to be the designer of the system that does it.

How do future senior consultants learn if juniors don't do the research?

From the method itself. An encoded method writes down what the best people ask, and in what order. They learn by designing and running the system, checking its output against the sources, and arguing about the recommendation with a senior.

Two ways to work with FifthRow

What we're building at FifthRow is the Large Consulting Model, a consistent, repeatable, fact-checked way to run consulting work. The firm puts its own method in, and it runs inside Copilot, ChatGPT and Claude, where your team already works. We do not train on, resell, or package your data.

You can license the LCM for your own delivery – encode how you deliver once, reuse it across engagements, and sell more expert work from the same senior team. Or partner with us in delivery, where we co-own the work and agree up front who owns the client and the quality bar. Either way, you keep the client and the last mile. Both options are on our page for consultants.

If you want to see what your own discovery method looks like as a system, book a slot with me.