Mixer Mode for consultants — a thought experiment, not advice

Disclaimer up top: I am not a senior McKinsey consultant or a BCG partner. This is a pedagogical thought experiment, not advice. But if I look at consulting through the Mixer Mode lens, something worth naming shows up: the senior partner isn't alternating between "domain expert" and "reader of internal politics" and "seller of the next phase". They're holding all three at once, modulated.

The Disclaimer First (Because It Matters)

I'm not a partner at a consulting firm. I haven't carried a book of business, sat in a steering committee on the firm side, or shepherded a sponsor through the politics of a transformation. This is a translation exercise — a demonstration that the frame travels from software to other domains where senior judgment is structured and tacit — not an assertion about how consulting works from the inside.

I'm writing it because the shape repeats across enough professions that pointing at the shape is useful, not because I have a diagnosis to sell anyone who actually runs an engagement. If you're a partner reading this and the channels are wrong, please cut them differently in the comments. That's more valuable to me than nodding agreement.

The Senior Partner's Channels

From outside, the senior partner seems to hold at least five things at once. Client domain — the industry, the metrics, the leverage points, the language the client uses internally. Internal politics — who blocks, who sponsors, who pays, who'll quietly torpedo the recommendation in week six. Buyer expectation — what the CFO told you they wanted, which is rarely identical to what they actually need. Deliverable quality — the deck, the argument, the data backing it. Firm reputation — don't burn the next engagement, don't take a stance that closes the door on the renewal.

Five channels, modulated, simultaneous. Not five hats picked up and put down between meetings.

Why the Junior Pipeline (Process-Only) Kills the Mixer

This is the part I think translates most cleanly from the software case. The traditional path to partnership ran through years of analytical work — model the cost structure, build the slides, run the benchmark. In doing that work, the associate absorbed the other four channels by being in the room: sitting in the steering committee while the partner read the sponsor, watching how the deck got reshaped between the dry run and the final, learning when a recommendation gets softened because the firm has another engagement coming.

The analyst learns to push out slides, not to read the room. The associate runs the analysis, not the calibration of the sponsor. The two years making decks weren't the job. They were the preceptorship where the other four channels got absorbed by osmosis. The deck was the excuse for the preceptorship, not the point of it.

If LLMs absorb slide production — and they're starting to, on the analytical layer, not yet on the narrative or the framing — then the preceptorship structure breaks. The associate produces fewer decks, learns less by osmosis, and the pyramid narrows exactly where the Brynjolfsson and Acemoglu work measures the entry-level compression. The channel that was being learned silently goes unlearned, and no one notices for five years because the existing partners still have it.

What Changes With Agentic + Meta-Software

The five channels don't disappear. They modulate faster. The partner who could hold three live engagements before can hold five, because the analytical work that used to consume associate hours is now consuming agent compute. That's the optimistic version, and it's real.

The pessimistic version is the one the paper warns about: without Meta-Software — governance of the agent's work on the deliverable — the predicted slowdown shows up. The partner ends up reviewing line by line what the agent produced, because no layer is enforcing the firm's voice, the engagement-specific tone, the things that absolutely cannot be said for political reasons. The reviewing-line-by-line move eats the capacity gain and then some. The firm "adopted AI" and got slower per deliverable, because the contract for what the agent should produce was nowhere — it lived in the partner's head, the same place the prompt used to live.

The Meta-Software layer for consulting work doesn't exist yet in a meaningful form, and that's an opportunity worth naming. Versioned style guides per client, per industry, per engagement. Auditable rubrics for what a deliverable has to satisfy before it lands on the partner's desk. The shape of the layer is recognizable from other domains where it's further along.

What Doesn't Translate

Three things, at least. CEO trust isn't delegated to an agent. The relationship is presence, history, and the specific feel of someone who's been in the room before. No model substitutes for that, and pretending otherwise is the kind of mistake that ends an engagement.

Judgment about what gets said and what stays unsaid in the room is pure Pillar 1. The senior partner who knows that the CFO is about to be replaced, and adjusts the recommendation accordingly without saying so, is running a channel that doesn't externalize. It can be described after the fact. It can't be specified ahead of time. The Mixer Mode frame names that channel exists; it doesn't replace the cognition that runs it.

The framework doesn't make a junior partner. Mixer capacity is earned over years of cases, and the cases have to be real cases with real stakes. No simulation produces the same channel expansion that a live engagement produces, and the firm that thinks it can shortcut this with training is the firm that runs out of senior capacity quietly over a decade.

Mixer Mode, in consulting, names the simultaneity. It doesn't substitute for the simultaneity. The frame helps describe what the senior does and helps design the pipeline that produces the next one. The work itself is still the work.

If you've worked in consulting — at any tier, on any side of the table — which channel took longest to learn? Especially curious about the one that wasn't in any training program, the one that nobody named explicitly, the one you only noticed you had after watching a partner who didn't have it. Send me a DM or reach out via the contact channels at rlabs.cl — I want to calibrate the model against real cases, not against my own outside reading.

#MixerMode #FutureOfWork #DigitalTransformation #TechLeadership