The AI Discount Question

I The Signal

Clients have started asking a question that used to stay unspoken: if AI made the work faster, why is the price the same?

It is no longer hypothetical. Over the past nine months it has shown up in contracts, renegotiations and filed accounts:

• It is in contract terms. The Australian Financial Review reported in April that some companies are hard-coding a 10% "AI efficiency" discount into contracts. Revium, a boutique advisory firm with about 40 staff, described a tick-box at the final phase of a contract worth more than $1 million: tick "we used AI", and 10% comes off the price immediately. Its managing director said the softer version is now routine: "are you using AI?", and if yes, straight to a push for lower cost. (AFR, via Revium)
• The biggest buyers are renegotiating. Internal Amazon documents seen by Business Insider show major AWS ProServe customers, including Booking.com and Danske Bank, "proactively" seeking to renegotiate for "AI-delivered pricing". One document summed up the client view: "If GenAI enhances or speeds things up, they should get more value faster and shouldn't pay the same rates." ProServe's response was to aim for up to 75% fixed-price projects. (Business Insider, Feb 2026)
• The professionals are doing it to each other. KPMG International pressed its own auditor, Grant Thornton UK, to pass on AI savings, and threatened to switch if it didn't. Its 2025 audit fee fell to $357,000 from $416,000, a 14% drop in dollar terms. (Financial Times, via The Irish Times)

Against that, the best buyer-side data available points the other way. Source Global Research's survey of UK and US consulting clients found that 82% expect AI to speed up projects, but only 48% expect it to make the work cheaper. The other 52% expect fees to rise. The clients who expect better work are the ones who expect to pay more. (Source Global Research, Feb 2026)

Read together, this is not a market moving to "cheaper". It is a market sorting work into two piles: work the client sees as effort, which gets discounted, and work the client sees as judgment, which doesn't.

II What's Being Overweighted

The common prescription is "move to outcome-based pricing". The large firms are the evidence offered. BCG's CEO says three-quarters of the firm's largest AI cases now carry variable fees (WSJ, republished by BCG), and Business Insider reports Accenture and others moving the same way (Business Insider, Jun 2026).

That evidence says less than it seems to. In the same interview, the BCG CEO puts outcome-based work across the whole firm at "significantly less than a third". He also says his first case as a partner was a variable-fee deal: "this is not new." Variable fees are a risk-sharing tool for large transformations with measurable targets, sold by firms with the balance sheet to absorb a miss. For a solo advisor or a five-person firm, copying gain-share is mostly a way to do unpaid work while the client controls the variables. Even the buyers aren't asking for it. West Monroe's chief AI officer told Business Insider that most clients want fees "tied to outcome metrics, not pure gain share."

The underweighted point is simpler. The discount request is triggered by disclosure, not by outcomes. In the Revium example, the price moved the moment the box was ticked. How you describe your use of AI is now pricing language, whether or not you write it that way.

III Counterfactuals

If the tick-box AI discount spreads from large-contract procurement into standard mid-market terms over the next two to three quarters, then any engagement quoted in days or hours will be reopened at the point of disclosure. Advisors who quote a defined deliverable and the decision it supports will be asked a different question: "what does this decide?", not "how long did it take?"

If the Deloitte Australia episode becomes the reference case buyers cite, then verification will become a line item. Deloitte repaid A$97,587 of a A$440,000 government fee after a report was found to contain AI-related errors, including non-existent references and fabricated quotes (AFR, Oct 2025; Australian Greens release on Senate estimates). In that world, contracts will ask who checked the output and how, and advisors who can show their review process will hold a premium that AI speed cannot erode.

IV Implications for Advisors & Operators

• Write your AI answer before the client asks. One paragraph: where you use AI (research, first drafts, synthesis, interview summaries) and where you do not (the recommendation, the trade-offs, the sign-off). Said first, it frames the conversation. Discovered later, it opens a renegotiation.
• Stop quoting time for work AI has compressed. Price the deliverable and the decision it enables. The only rate data on independents we could find points the same way. A June 2026 survey of 56 independent consultants found an average hourly-equivalent of $179 for hourly billing, $222 for retainers and $275 for time-limited monthly fees. Value-based pricing was the only rate-setting method associated with higher rates. The sample is small, self-selected and social-impact-heavy, so treat it as direction, not a benchmark. (Chorus 2026 Rate Report)
• Give the saving back visibly, as scope rather than as a discount. In Source's survey of senior US tax clients, 66% expect to pay less for like-for-like AI-enabled work, but 70% of those would redirect the saving to other advisory work. Revium used AI-run interviews to reach about 20 more stakeholders for roughly the cost of three or four face-to-face interviews. That turns the efficiency into a bigger engagement, not a smaller invoice.
• Use milestones tied to outcome metrics, not contingency. If a client wants skin in the game, offer a fixed fee with a modest bonus on an agreed metric. Do not make your base fee depend on variables you cannot control.
• Name your verification step and price it. "Every source checked, every figure traced, reviewed by me before delivery" is now worth writing into the proposal. It is the part of the job the Deloitte case showed clients will pay for, or claw back.
• Do not underprice to look modern. When Revium quoted a rebuild at about 60% of its original cost, the client's board initially pushed back. It worried the lower price meant lower quality. A large, unexplained AI discount can signal less judgment, not more efficiency.

V What We're Watching Next

• Whether "AI efficiency" discount clauses appear in mid-market master service agreements and procurement templates, not just in large public and enterprise contracts.
• The next round of audit fee disclosures, for a sign of whether KPMG's 14% cut was a one-off or the start of a pattern other buyers cite.
• Whether Source or another buyer-side researcher repeats the client survey. The 48/52 split on fees is the most important number in this issue, and one data point is not a trend.
• Whether any independent-consultant rate survey starts asking directly about AI disclosure and price changes. Right now that evidence is anecdotal.