McKinsey’s April 2026 survey of 419 pricing executives puts it precisely. Discount approval and governance: a top-three impact opportunity for 62%, a top-three investment priority for 22%. Renewals, 60% against 13%. The area ranked lowest for impact takes the most funding.
Why speed did not fix it
The constraint at the point a deal is structured was never speed. It is authority. Can the person setting the price see the margin consequence before committing, test the discount against what they may grant in that entity on that term, and know the revenue treatment while the clause is negotiable?
Restatements are falling, not rising. Total restatements fell 18% in 2025 to 391, the second lowest in 20 years of Ideagen Audit Analytics data. But revenue recognition climbed back to second most common cause at 14%, and large accelerated filers took their biggest share since 2019. Volume down, concentration moved toward bigger companies.
What did not move
EU AI Act high-risk obligations shifted to 2 December 2027. Year-end close did not. The regulator gave the market 16 more months. Your close gives you 19 weeks.
Nineteen working weeks to 31 December. Your organization does four things inside them at once.
It builds and locks the FY27 plan, so a project not in the plan by October waits a year for funding. Your team structures more deals, at deeper discounts, on more non-standard terms, than in any other quarter. Year-end close begins. And whatever the business reads this autumn shapes next year’s priorities.
Four calendars, one file
Those are not four calendars. Every quote your team structures between now and 31 December becomes revenue you recognize, defend, and audit in the first quarter of next year. The concession your team grants in the last week of December still sits on the books when the auditor arrives in March.
That file is being written now, by people who do not think of themselves as writing it.
What the money bought
In July, Gartner published something worth sitting with: a survey of 204 finance leaders taken in March, splitting finance AI investment by what it aimed at. The returns tracked the allocation almost exactly, as CFO Dive reported.
Gartner found the reverse case in the same data. Functions investing in what it calls Upend initiatives, the ones creating new value propositions, products, or markets, were more than twice as likely to report high realized value. The functions seeing value changed what the organization could decide, not how fast it could type. The split is not a story about AI failing. It is two portfolios behaving differently, with most of the money in the one that flattens.
I read that as a diagnosis rather than a disappointment. If your AI spend has not shown up in your numbers, the technology probably did not underdeliver. Instead it went to a part of the process that was never the constraint.
Where discount authority actually decides the number
Your commercial teams structure deals. They configure a bundle, set a price, apply a discount, agree a term, commit to a service level. That moment decides margin, sets revenue treatment, and creates audit exposure. Everything after it in the quote-to-cash chain, the contract, the invoice, the ledger entry, the disclosure, is bookkeeping on a decision already made.
So what does a faster quote buy at that moment? It gets you there sooner. It does not make the number right.
Take a seller who could not price a four-year bundled agreement correctly on Monday. Give them an assistant that drafts the quote in 90 seconds. The same flawed number arrives Monday morning instead of Wednesday afternoon. You have compressed the cycle time on a decision you could not defend. That is why the return did not appear.
Speed compresses the time to a decision you could not defend. It does not make the decision defensible.
Why discount authority is the constraint, not speed
Most finance functions are not blind to this. But they meet it too late. You can know the margin on a bundled four-year agreement when you structure it, but most enterprises calculate it afterwards, from a quote already sent. By then you can accept it or reopen a negotiation you have closed. Neither is a control.
The offers you cannot price from history
A margin guardrail is a claim about a cost you already know. Set a floor at 34 points and you are asserting that someone has calculated what delivery costs at the volume the customer will actually consume. On products you have sold for years, that assertion holds, learned by attrition. On an offer introduced in the last four quarters, priced per outcome or across a consumption range, it does not. Nobody has costed it at the top of the range.
Why discount authority inherits the wrong number
The gap is widening from both ends. Buyers keep pulling terms toward consumption and outcome models. AlixPartners examined 65 SaaS and AI-native companies and found only four had fully adopted outcome-based pricing, while 72% now meter part of delivery through consumption or credits. That hybrid middle is where the unpriced variance sits.
The guardrail came from a different product. The cost base moves during the term.
Read that condition as a control statement. Whether recovery is permitted was settled when the deal was structured, by someone who probably did not know they were settling it.
So the new offer enters the quoting motion carrying discount thresholds inherited from a product with different economics, sold on terms whose cost base moves during the term. Authority is not absent. It is calibrated on the wrong product.
The money went to the easier half
McKinsey asked 419 pricing executives where agentic AI would have the greatest impact and, separately, where the money was going.
The three governance-heavy areas are the three starved of funding. Supporting data below.
Where pricing leaders say the impact is, and where the budget goes. Percentage ranking each area in their top three. Source: McKinsey Agentic AI in Pricing Survey, November 2025, n=419, published 7 April 2026.
But read the last row against the four above it. The area ranked lowest for impact attracts the most funding. The areas ranked highest, all of them discount authority and governance work, get the least. McKinsey’s verdict is the sentence worth carrying into FY27 planning: the budget alignment improves productivity but not growth.
Where discount authority is still unbuilt
Only 5% to 10% of organizations have fully scaled agentic AI in any pricing use case. Discount approval, renewals, and contract compliance sit in early development, because they need safeguards nobody has built. So the market knows where the value is. It has not worked out how to govern it, so it spent elsewhere.
What is the governed half worth when someone does it properly? McKinsey documents a $15 billion distributor that spent 18 months replacing manual pricing across 1.5 million stock keeping units with governed guidance. That delivered more than 200 basis points of margin, and a price adviser and discount manager found a further 50 inside ten weeks. More than 250 in total.
What discount authority means, mechanically
Decision quality at the moment of deal structure has a specific shape. The person structuring the deal sees the margin consequence before committing, not after. Their discount authority binds in the moment, testing what they may grant in this entity, for this product, on this term length, which is what governs variable pricing is for. The revenue treatment of a non-standard clause is known while the clause is still negotiable. And when the auditor asks in March why that price was approved, the answer is a record rather than a reconstruction. None of that is achieved by making the quote faster. All of it comes from putting authority into the moment the deal is structured.
Those three words describe a commercial control problem, not a technology one. Clear rules means discount authority exists as enforced structure rather than a policy document. Auditability means the decision itself captures the approval and the margin position. Reversibility means you can still see the consequence while the term is open to change.
What the governance data says about ownership
McKinsey’s 2026 AI Trust Maturity Survey, across roughly 500 organizations, found governance and agentic controls lagging everything else. The finding I keep coming back to is ownership. Organizations that assign clear accountability scored 2.6 on average. Those without scored 1.8.
I ask question four in most conversations with finance leaders. Can you produce the approval record, with the margin position as it stood, without asking anyone. Almost nobody says no. They say it would take a few days, then start naming who would have to be involved.
The few days is the answer.
Because a quote is a decision with a wallet attached. If the discount authority around it is a policy document and a deal desk queue rather than an enforced structure, speed only raises the rate at which undefended decisions enter the book.
The deadline that moved, and the one that did not
Plenty of enterprises built their AI governance work around 2 August 2026, when the EU AI Act was to apply to high-risk systems. That date moved. Regulation (EU) 2026/1744, the Digital Omnibus on AI, entered into force on 27 July 2026, and high-risk obligations for stand-alone Annex III systems are deferred to 2 December 2027, with Annex I systems to 2 August 2028. Article 50 transparency obligations were not deferred.
Read as breathing room, and for compliance planning it partly is. Notice what did not move. Your year-end close did not. Nor did your auditor. BDO’s February 2026 assessment of ASC 606 pain points reads like a description of a complex enterprise quote: variable consideration from discounts, rebates, and contingent pricing; principal versus agent analysis driving gross or net treatment; contract modifications arriving through change orders and amendments. Each turns on how you allocate the transaction price, a standalone selling price question settled when the bundle is structured, not at close.
The regulator gave the market 16 more months. Your own close gives you 19 weeks. Anchoring governance work to a regulatory date was always the weaker argument. The stronger one is that the number you defend in March is being created by the quotes going out this month.
Six questions worth answering before October
Diagnostic rather than rhetorical, and all answerable from your own last quarter. If you can answer all six, you are in better shape than most enterprises we see. If you cannot answer three, that is not a reporting gap. It is an authority gap. We have written separately on why that is an audit trail question rather than a reporting one.
Four bands. None of them is a verdict on your team.
What your answers usually reveal
The pattern is usually the same. The information needed to decide well existed somewhere in the business. It was not present at the moment of the decision, in the hands of the person making it, with the authority to constrain it.
So take the two you could not answer into the FY27 conversation in October, rather than all six. A quantified gap gets funded. A described condition does not, and attempting all six at once is the main reason this sits unfixed for another year.
What to do with those answers
Start from your own two weakest, not from a demo. This does not put you in a sequence.
Not faster quotes. Defensible ones, at the point the deal is structured.
A seller opens a four-year managed services deal for one of your German entities. Their discount authority is checked before the quote can go out, for that product, on that term, in that entity. The margin appears while the term is still negotiable. When the approval lands, the record writes itself, including where the margin stood at the time.
Nobody has to remember to do any of that. The design does it, so nobody has to enforce a discipline.
Then in March, when the auditor asks, somebody retrieves that record. Instead of reconstructing it.
That happens without standardizing every entity first. An enterprise that grew by acquisition cannot put every entity on one vendor’s pricing model, so the governing layer sits above the entities instead, which is what a composable revenue architecture means in practice. Our CFO guide to composable stacks covers the assembly. A layer that only works once everyone migrates is not governance. It is another migration.
Discount authority, proven in public
We build this for technology service providers. The roster is public: Dell, Telefónica, TierPoint, Park Place Technologies, telent. Ensono used its own newsroom to explain why it chose us for complex hybrid managed services. telent runs £50 to £60 million of annual quoting through the platform and went live in eight weeks. Dell EMC cut proposal changes from a day to 15 minutes. Both are public. Both measure speed, and we do not stretch case studies.
The discount authority question worth asking any vendor
So ask us the harder question. How long does it take one of our customers to produce that approval record, with the margin position as it stood. Then ask our competitors the same question.
Is this an argument against investing in AI for finance?
No, and the distinction matters for the FY27 plan. The productivity use cases are not wasted. They plateau, which is a recoverable failure. Gartner’s own projection is that CFOs who get portfolio deployment right add 10 points of margin growth by 2029. The question is whether your portfolio is weighted toward the half that plateaus, and whether you can say what the split is without going and asking someone.
The EU AI Act deadline moved. Does that not buy us a year?
For AI Act conformity work, it buys 16 months on stand-alone Annex III high-risk systems, to 2 December 2027. It does not move Article 50 transparency obligations and it does not touch your reporting calendar. The exposure in this article is created by commercial decisions and settled under revenue recognition, not under the AI Act. Your close date, your auditor’s questions on variable consideration and bundled performance obligations, and your FY27 funding window all sit inside the same 19 weeks.
We already have a deal desk and an approval matrix. Is that not the same thing?
Same intent, applied after the fact. The test is question four. Can you produce the approval record, with the margin position as it stood, without asking a person to find it? A deal desk reviewing a structure someone else built is a check. Discount authority at the point of structure is a constraint. The difference shows when volume rises, which is what Q4 does.
We are mid-integration after an acquisition. Is this the wrong time?
Gartner does not endorse any vendor, product or service depicted in its research publications and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner research publications consist of the opinions of Gartner’s research organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this research, including any warranties of merchantability or fitness for a particular purpose.
GARTNER is a registered trademark and service mark of Gartner, Inc. and/or its affiliates in the U.S. and internationally, Magic Quadrant is a registered trademark of Gartner, Inc. and/or its affiliates and is used herein with permission. All rights reserved.
Gartner, Magic Quadrant for Configure, Price and Quote Applications, By Mark Lewis, Luke Tipping, 22 January 2026.
