The question the invoice cannot survive unprepared
Session 1 established the evidence: inside the frontier, AI-assisted consultants worked 25.1% faster at quality rated 40% higher. Clients read the same study. Procurement teams read it twice. So the question is not whether "why is the invoice unchanged?" gets asked - it is whether your firm's answer was written calmly in advance or improvised in the room. Today runs firm-level and commercial: the arithmetic of the leak, the three honest answers, and a re-pricing exercise on a real engagement. Your teams' engagement-level craft - what the faster, better deliverable actually looks like - lives in the consultant track with Northbridge Advisory and its client Cadence.
One framing rule before we start: "honest" is doing a lot of work in this session. There are dishonest answers to the invoice question - silent discounting, silent hour-padding, quiet non-disclosure - and they all trade long-term trust for short-term comfort. We price the other way.
Where the leverage model leaks 8 min live
The classic pyramid: partners sell, managers run, analysts produce - and the firm bills by the hour at every level, with margin concentrated in the wide analyst base. Generative AI compresses exactly that layer. The analyst work it accelerates - research, first-draft decks, summaries - is the pyramid's revenue engine, and McKinsey's own Bloomberg-reported use is the tell: faster PowerPoints and proposal drafts. If the hours drop and the rates stay, the arithmetic does the rest.
LiveThe arithmetic of a 25% faster engagement4 min▶
Run the study's own speed number - 25.1% faster inside the frontier - through a plain hourly engagement. The hours below are round illustration numbers; the percentage is the evidence.
| Line | Hourly model, before AI | Same engagement, 25.1% faster |
|---|---|---|
| Analyst hours on the workstream | 400 hours | ~300 hours |
| Fee at an unchanged hourly rate | 400 x rate | 300 x rate - a quarter of the revenue gone |
| Quality delivered (per the study) | baseline | rated 40% higher |
| Who captured the improvement | - | the client, silently, at your expense |
Read the last two rows together: under hourly billing, the firm delivers a better product and books less revenue for it. That is not a technology problem - it is a pricing model built for a world where effort and value moved together, meeting a tool that split them apart. The leak is structural, so the fix has to be structural too.
Self-studyWhy "just bill the same hours" is not a plan2 min▶
The tempting non-answer is to keep the invoice shape and let the hours quietly stop meaning anything. It fails on both flanks. Clients benchmark: when competitors deliver in two weeks what your invoice says took four, the hours read as padding, and padding reads as dishonesty. And your own people know: juniors asked to record hours that no longer reflect the work learn that the firm's numbers are negotiable - a lesson that does not stay confined to timesheets. Every durable answer starts by conceding the true sentence: the hours went down. The fee logic has to stand on something else.
The three honest answers 7 min live
Three answers survive the invoice question with trust intact. They are not mutually exclusive - mature firms run all three at once, on different engagement types.
LiveValue, product, transparency5 min▶
Answer 1 · Value pricing: price the decision, not the hours. The client was never really buying hours - they were buying a recommendation they can act on, someone accountable for it, and getting it in time to matter. Recommendation quality, accountability, and speed are exactly what an AI-augmented firm delivers more of, so the honest move is to attach the fee to decision stakes: what the choice is worth, not what the drafting cost. The invoice answer becomes one sentence: "You are paying for the recommendation and the accountability behind it - the drafting was never the product."
Answer 2 · Productized and subscription offerings: sell the playbook, not just the project. When AI compresses production, the assets the firm built - playbooks, diagnostics, tools - can carry revenue as products with recurring pricing that never references hours at all. The market signal here is unmistakable: PwC became OpenAI's first reseller and its largest ChatGPT Enterprise customer, with licenses for over 100,000 employees. A Big Four firm now retails AI capability as a product line - the canonical marker that firms sell AI assets, not only AI-assisted hours.
Answer 3 · Transparency clauses: disclose the AI use, keep accountability priced in. State in the engagement letter where AI assists, what the firm verifies, and that a named human owns every fact that reaches the client. Disclosure is not a discount trigger - it is what lets you charge for verification and judgment as explicit line items. The cost of the opposite is on the public record: Deloitte Australia partially refunded a government contract after a report was found to contain fabricated references, which Deloitte attributed to generative AI use it later disclosed. Session a3 owns that story in full; here it is simply the price tag on non-disclosure.
Self-studyWhat NOT to do: the two silent moves3 min▶
Two moves feel safe in the moment and destroy trust on a delay. Silent discounting: shaving the fee without naming why, teaching the client that your price was soft and inviting the next shave - you have conceded the leak and bought nothing with the concession. Silent hour-padding: keeping the invoice shape while the hours stop being real, which converts a pricing problem into an integrity problem the first time a client benchmarks your timeline against a competitor's. The pattern behind both: they avoid the conversation instead of winning it. Every honest answer starts by saying out loud that the work got faster - then explains what the fee actually buys.
What clients now expect 5 min live
The demand side moved too. Faster turnarounds are becoming the default expectation, not a premium - and the old suspicion ("did you use AI on my work?") is inverting into a selection question: which firms use AI well. Sophisticated clients increasingly treat visible, governed AI capability as a reason to hire, not a reason to haggle.
LiveThe arms race, and your talking points4 min▶
The scale of firm investment tells you what the firms believe clients will pay for. EY launched EY.ai with a US$1.4B investment. KPMG committed US$2B to its Microsoft AI alliance. BCG's ~90% employee AI usage - a press-reported figure from the company's own statements, label it that way - signals the same conviction from strategy consulting. Nobody spends billions to bill fewer hours; they spend it because AI-augmented delivery is becoming the qualifying bar for the work itself.
Your talking points for the fee conversation, in the order they land:
| Client says | You say |
|---|---|
| "AI made this faster - why is the fee unchanged?" | "It did, and you received it three weeks earlier at higher quality. The fee prices the decision and the accountability, which have not gotten smaller." |
| "Couldn't we just run the model ourselves?" | "You could get a draft. The verification, the judgment about your situation, and a firm standing behind the number - that is what you are hiring." |
| "Do you disclose where AI is used?" | "Yes, in the engagement letter - along with what we verify and who owns each fact. The disclosure is part of the quality system, not a caveat." |
Re-price one engagement 💷 ★ 20 min · everyone works
One real engagement, re-priced end to end. If you cannot use a live client, borrow the consultant track's case: Northbridge Advisory's six-week go-to-market engagement for Cadence works fine as neutral ground.
Take a real (or the Northbridge x Cadence) engagement. One you know the economics of - scope, team, duration, fee. Write those four facts at the top of the page.
Write its hourly-model economics. Hours by level, rate by level, the fee as the sum. This is the "before" column - the version the invoice question attacks.
Estimate the AI-compressed hours honestly. Which workstreams sit inside the frontier (session a1's map earns its keep here), and what does the analyst layer actually shrink to? Resist both flattery and panic - a defensible number beats a dramatic one.
Re-state the fee as value-priced with three anchors. Decision stakes (what the choice is worth to the client), speed (what arriving earlier is worth), accountability (what a firm standing behind the number is worth). If an anchor has no sentence behind it, it is decoration - cut it or defend it.
Draft the two-sentence answer to "why is the invoice unchanged?" Sentence one concedes the true fact (the drafting got faster). Sentence two names what the fee buys (the decision, the speed, the accountability). No hedging words - session a4 will hold your deliverables to the same standard.
Stress-test with the meanest procurement question you can imagine. "Your competitor quoted 40% less and says they use the same AI." If your two sentences survive that, they will survive the real meeting. If they do not, the weak anchor is the one to rebuild.
Homework 📝 ◐ 25 minutes
- Finish the re-pricing one-pager - before economics, compressed hours, value-priced fee with three anchors, the two-sentence answer - and show it to one partner you trust to argue back.
- Audit your engagement letters for AI language. Do they disclose AI use and verification ownership anywhere? Bring the honest yes/no to a3, which builds the policy those clauses come from.
- Price one playbook as a product. Pick the practice's most reused asset and write one sentence on what it would cost as a subscription - no delivery hours attached. You are collecting raw material for a5's operating-model session.
Three questions before you go 🎯 ◐ 90 seconds
1 · Hourly billing leaks under AI because...
The study's 25.1% speed gain, run through an unchanged hourly model, means the firm delivers higher quality (+40%) and books roughly a quarter less revenue for it. The leak is structural: effort and value no longer move together.
2 · The three honest answers to "why is the invoice unchanged?" are...
All three start by conceding the true fact - the drafting got faster - and then attach the fee to what did not shrink: decision stakes, reusable assets, and verified accountability. The silent moves (discounting, hour-padding) are the ones that destroy trust.
3 · The PwC-OpenAI deal matters to this session because it signals...
First reseller + largest ChatGPT Enterprise customer (100,000+ licenses, May 2024, per CNBC): PwC changed what it sells rather than defending the hour as the unit of account. That is answer 2 - productization - executed at Big Four scale.
What this session covered from the official universe
Certificates, videos, and graded assessments stay with the official providers - this track teaches the working core.