Written by:
Editorial Team
DSG.AI
Audit fee compression is here, and AI is the lever. In February 2026, KPMG pressed its own auditor, Grant Thornton UK, to cut fees on the grounds that AI was making the work more efficient, and the fee fell 14%, from roughly $416K to $357K (Irish Times). The lesson for every internal audit buyer is simple: if the firm auditing the Big Four is demanding AI-driven discounts, you are entitled to the same conversation. The buyers who do not have it are quietly overpaying. The buyers who do, and ask for the right things, will not just shave a line item, they will get more audit for less money.
This article is about that second move. A cheaper Big 4 contract is the obvious play and the weakest one. The stronger response is to insist on audit that AI plus humans actually performs, with the coverage and evidence to prove it. Here is how internal audit fees are moving, how to tell real efficiency from theater, and what to demand at the table.
Why AI is structurally compressing audit fees, not just trimming them
For two decades, audit pricing rested on one fact: testing controls and gathering evidence took a lot of billable hours, and most of those hours were junior. AI removes the floor under that model. Tools now scan entire transaction populations instead of samples, automate routine testing, and draft documentation. PwC's US assurance transformation leader has said the firm expects AI to automate the full audit cycle within calendar year 2026, with a tool for every step from planning to evidence collection to tie-out (Accounting Today).
When the labor that justified the bill gets automated, the bill has nowhere to hide. That is why this is compression, not a one-off discount. Across professional services, 79% of firms report that AI is changing pricing conversations and 42% say clients are now questioning the pricing model itself (Accounting Today). The billable hour was the unit of value precisely because hours were the constraint. Once a population test runs in minutes, charging by the hour rewards the firm for being slow.
The KPMG episode matters because it removed the deniability. The OnlyCFO newsletter put it bluntly: KPMG just handed every audit client the playbook for renegotiating fees (OnlyCFO, June 2026). You no longer have to argue that AI should lower your fee. A Big Four firm already made that argument, on the record, to win its own discount.
Internal audit fees vs. external audit fees: where buyers have the most room
One distinction first, because it changes your tactics. The KPMG story is about a statutory external audit, and external audit fees are sticky: rotation rules, regulator scrutiny, and audit committee politics slow renegotiation. Internal audit fees, the co-sourcing and outsourcing spend a CAE controls directly, are far more elastic. You set the scope, you pick the provider, and you can move the work. That makes internal audit the place where AI-driven fee compression hits first and hardest.
If you co-source internal audit with a Big 4 or mid-tier firm, your contract is the easiest one in the building to reprice on AI grounds. Before you negotiate, know what you are paying for: our Internal Audit Sourcing Cost Reference lays out current in-house, mid-tier, Big 4, and subscription ranges, and Co-Sourcing vs. Outsourcing Internal Audit covers which model fits which function. Walking in with the numbers is the difference between asking for a favor and citing a market.
Real AI efficiency vs. fee theater: how to tell them apart
Here is the trap. Every firm will now claim AI savings, because every client is now asking. Most of what you will hear is positioning. Some of it is real. The difference is testable, and you test it by asking what changed in the work, not in the slide deck.
Real efficiency shows up as more coverage and faster cycles at lower or equal cost. Theater shows up as the same audit, the same sampling, and the same report, with an AI logo on the cover letter and a modest discount offered before you even pushed. Use this to sort them:
| Signal | Real AI efficiency | Fee theater |
|---|---|---|
| Testing scope | Full-population testing replaces sampling | Still "we sampled 25 of N" with an AI mention |
| Cycle time | Findings land in days, measured and reported | Same 10 to 12 week cycle, no time data offered |
| Evidence | Pulled from source systems automatically, with a trail | Still client-prepared screenshots and exports |
| Pricing unit | Per audit, per outcome, or subscription | Same hourly rate, slightly fewer hours |
| The discount | Tied to a named efficiency you can verify | A round percentage, offered to keep the account |
| AI governance | The firm can show how its own AI is controlled | "Our AI is proprietary" and nothing more |
The tell is the pricing unit. A firm that has genuinely automated the work prices the work, not the hours, because hours no longer describe what it does. A firm offering you 10% off the same rate card is protecting the billable hour and hoping you accept a tip instead of the savings. As one finance commentator noted after the KPMG cut, the open question is no longer whether AI makes audits cheaper, it is when the savings reach the customer (The Finance Story).
One honest caveat, because the firms will raise it and they are partly right. The Institute of Internal Auditors points out that AI does not eliminate judgment: experienced auditors still validate outputs, refine scope, and own the conclusions, and partners often review more anomalies, not fewer, because AI surfaces more of them (The IIA, May 2026). That is true. It is also not an argument for keeping fees flat. Judgment was always the expensive, defensible part of the bill. The clerical hours around it were the padding. Pay for judgment. Stop paying senior rates for evidence-gathering a machine now does.
What to demand when fees come down
A discount on a bad model is still a bad model. If you are reopening the contract, do not settle for a smaller version of what you already overpay for. Demand a better unit of work. Ask for these, in writing:
- Coverage, not just a cut. If AI made the audit efficient, the deliverable should be broader. Insist on full-population testing where it is feasible, and make the provider state the share of testing that is full-population vs. sampled. That single number separates automation from marketing.
- Cycle time as a contractual metric. "AI-efficient" means nothing if findings still arrive twelve weeks late. Put a measured cycle-time target in the engagement and report against it.
- Audit-grade evidence with a trail. Workpapers and evidence that survive external review, regulator questions, and your audit committee. If the "efficiency" is the firm asking you for more screenshots, it is your efficiency they captured, not theirs.
- A pricing unit that reflects automation. Per-audit, outcome-based, or subscription pricing, not a trimmed hourly rate. If the firm will only sell hours, it is telling you the work is still manual.
- Governance of the AI doing the work. Any AI performing audit procedures needs its own controls, monitoring, and audit trail, or you have swapped one assurance gap for another. A credible provider answers with a framework. The IIA, the standards body for the profession, publishes guidance on exactly this for a reason.
If a provider balks at all five, you have learned what their AI story is worth.
The stronger move: audit that AI and humans actually perform
The point of fee compression is not to win a cheaper contract for the same thin audit. It is to get audit that is genuinely performed, which is a different product from audit that is staffed and billed by the hour. That is the model behind assureIQ. assureIQ performs compliance work, it does not just track it: agents pull evidence from source systems, test controls across the full population rather than a sample, and assemble workpapers, with auditors scoping, judging, and signing off. Humans stay where judgment lives, which is exactly where the IIA says they belong. Automation absorbs the volume that used to dominate the bill.
The economics follow from the design, not from a negotiation. Across 250+ production AI deployments, this model delivers audits at 40-60% below typical Big 4 co-sourcing rates, with 50%+ reductions in audit cycle time and 3-5x increases in coverage. The coverage gain is the part the discount conversation misses entirely: full-population testing is not a faster version of sampling, it is a categorically better audit. DSG runs this ISO 27001 certified and EU-headquartered, which settles the data-residency question most European buyers raise first.
This is the difference between audit as a service and a repriced co-sourcing letter. One changes what the audit is. The other changes the invoice and leaves the work alone. When a Big 4 firm offers you an AI discount, the right counter is not "make it 20%." It is "show me the coverage and cycle time, priced per audit," and then put that quote next to a model built to deliver it. Start with our audit services overview.
The bottom line
AI is compressing internal audit fees structurally, and the KPMG precedent means no firm can credibly claim otherwise. Internal audit is where buyers have the most room to renegotiate, because the CAE controls the scope and can move the work. Use it. But do not stop at a cheaper contract for the same sampled, slow, hourly audit. Tell the real efficiency from the theater, demand coverage and audit-grade evidence rather than a round-number discount, and price the audit by the outcome. The buyers who treat fee compression as a chance to upgrade the audit, not just discount it, are the ones who will come out ahead. The rest are negotiating over the padding while the work stays the same.
<!-- related-links:start (auto-managed by seo/sync-internal-links.mjs) -->
