The person who decides your AI deal is not in your demo
If you sell AI software into B2B accounts, the hardest seat in your next deal belongs to someone who has probably never opened your product. Selling AI to the CFO is no longer a late-stage procurement formality — finance has moved to the front of the evaluation, and it is not evaluating on the same axis your champion is.
The evidence arrived on 9 September. Salesforce published its CFO Priorities Report, a double-blind survey of 865 finance leaders across France, Germany, Japan, the UK and the US, fielded 4–15 May 2026. Two numbers in it should reset how you plan outbound this quarter:
- Nearly three-quarters of finance leaders say the scope of their role increased in the last year, and the single most-cited reason is managing AI use and expansion, at 49% — ahead of managing a growing tech stack (44%), more complex revenue models (43%), and increased involvement in sales strategy (43%).
- Half of finance leaders now say they are primary decision makers in their company's AI strategy.
That is a structural change in the buying committee, not a mood swing. The function that used to arrive at the end to negotiate a discount now helps decide whether the category gets bought at all. Sam Chung, Chief Customer Officer for Agentforce Revenue Management, put the shift bluntly in the report: AI decisions reach finance as questions about risk and control rather than as budget requests, and the CFO has moved from signing off on AI to being answerable for it.
For a seller, "answerable for it" is the operative phrase. People who are accountable for an outcome buy differently from people who are merely paying for one.
Why the CFO's remit expanded, and why it lands on you
The report describes three pressures converging on the same desk at the same time. Each one changes what a finance leader needs from a vendor.
Revenue got harder to see
Seventy-one percent of finance leaders say their company sells through more channels than it did twelve months ago, and the average company now runs three. Direct field and phone sales still lead at 69%, but web self-service, partner and indirect, social, and mobile all now contribute meaningfully. Meanwhile 65% track deals across more than one revenue model — recurring subscriptions (46%), consulting services (46%), usage-based (43%), one-time transactional (38%) and hybrid (35%) all coexist.
A deal that starts in one channel, passes through a partner, and gets billed on consumption has to be reconciled across systems that were never designed to talk to each other. Finance is not managing more transactions. It is managing more ways revenue can be created, priced, billed, collected and recognised.
The manual tax got expensive enough to notice
Roughly seven in ten finance leaders (67%) say their team still completes at least one in five workflows manually, usually in spreadsheets. They believe as much as 40% of their team's work could be automated. That gap is why finance is genuinely interested in AI rather than merely tolerant of it — and it is why a vendor who shows up with a credible automation story gets a real hearing.
AI became a governance problem before it became a productivity win
Finance leaders are being asked to govern AI's rollout across the whole company while simultaneously using it to modernise their own function. Decisions that used to sit with IT — where AI gets deployed, what data it reaches, what controls surround it, how impact is measured — now carry financial, control and audit implications. So they land on finance too.
This is the part most AI GTM vendors have not adjusted to. Your product is not being assessed as a tool. It is being assessed as an exposure.
The four blockers that actually decide the deal
Ask a seller what kills an AI deal and they will usually say price or ROI. The data says otherwise. Here is what finance leaders name as their primary blockers to expanding AI use:
| Blocker | % naming it | What the CFO is really asking | What answers it |
|---|---|---|---|
| Security concerns | 46% | What data does this reach, and who else can see it? | Data-flow documentation, retention terms, sub-processor list, named certifications |
| Limited internal subject matter expertise | 42% | Who operates this once the champion leaves? | Implementation scope, admin burden in hours, training path, support model |
| Governance and compliance concerns | 42% | Can I show an auditor what this did and why? | Action logs, approval gates, human-in-the-loop controls, exportable audit trail |
| Difficulty integrating with existing tech stack | 42% | Will this create another island of data? | Native integrations, sync behaviour, field-level mapping, failure handling |
| Lack of funding or resources | 39% | Is this worth displacing something else? | Cost per outcome, displacement case, measurable baseline |
Notice the ordering. Funding sits last. The four blockers ahead of it are all questions about control, and none of them are answered by a better demo. They are answered by documentation, by defaults, and by the ability to explain what your system did after it did it.
There is a second signal in the report that points the same direction. Among finance leaders already using AI, 90% report positive ROI, and among those using AI agents specifically, more than 90% report benefits across time savings, productivity, cost savings and forecast accuracy, with roughly half calling those gains significant. Time savings led at 52% "significant benefit" plus 42% "some benefit."
So finance is not sceptical about whether AI works. It is sceptical about whether your AI can be governed. That is a much better objection to receive, and a completely different one to handle.
The evidence gap: why capability claims stopped converting
Put the CFO data next to the other major benchmark published this month and the picture sharpens considerably.
Salesloft's 2026 U.S. Revenue Benchmark, a survey of 500 U.S. sales and revenue decision makers at companies with 200+ employees released on 1 September, found that 100% of respondents use AI somewhere in the revenue process, but only 20.6% describe their deployments as production-ready with measurable outcomes. Another 28.2% are still experimenting.
Every finance leader you sell to has now watched that happen inside their own company. They have approved AI tools that got switched on, got demoed enthusiastically, and never produced a number anyone could defend in a QBR. That experience is the real competitor in your deal — not another vendor.
It also explains the autonomy preference in the same benchmark: 38.4% favour a guided model where AI can recommend or act while humans keep oversight. Buyers are not asking for less capability. They are asking for capability they can supervise.
The same report shows why supervision is hard to achieve with the data most teams have. Updating CRM records is the most-cited administrative bottleneck at 37.6%, 31.4% call manual CRM administration the greatest barrier to pipeline generation, and 55.6% say the loss-reason information entered into CRM is mostly subjective seller reporting. Only about 32% can immediately pinpoint why a deal stalled.
A finance leader being asked to sign off on an AI system that runs on that data is being asked to underwrite guesswork. Their caution is correct.
What this changes about outbound
The practical consequence is that the account you are prospecting has a new gate, and your sequence probably does not acknowledge it exists. Four adjustments matter.
Multi-thread into finance earlier, and differently
Most multi-threading advice tells you to add the CFO late as an economic buyer. That is now backwards. If half of finance leaders are primary decision makers in AI strategy, finance is a category gatekeeper, and category gatekeepers need to be engaged before the champion has committed political capital.
But do not send finance the champion's message. The champion wants leverage and speed. Finance wants control and auditability. The same value proposition, sent to both, reads as a productivity pitch to one and as an unquantified risk to the other.
Lead with operating evidence, not capability claims
"Our AI writes personalised messages at scale" is a capability claim. Finance discounts it automatically, because they have heard it and they have seen the 20.6% number in their own building.
Operating evidence is different, and it looks like this:
- What the system does and does not touch, stated in data terms rather than marketing terms
- Where the human approval gate sits, and whether it can be enforced rather than merely offered
- What gets logged, for how long, and whether it can be exported for an audit
- How a rollback works when a campaign goes wrong
- What the number was before, what it is after, and who measured it
None of that is exciting. All of it is what an accountable buyer needs to say yes.
Use signals that indicate a finance-governed AI decision
This is where signal-based outbound does real work, because the trigger you want is not "company exists in my ICP." It is "this account is mid-way through an AI governance decision right now." Those moments are externally visible if you are watching for them:
- Hiring signals for RevOps, AI governance, data engineering or FP&A systems roles — a company staffing AI oversight is a company about to standardise its AI stack
- Posts and comments from finance and RevOps leaders about consolidation, tool sprawl, AI ROI measurement or audit requirements
- Consolidation language in public commentary, which matters given that the Salesloft data shows 26% actively consolidating revenue technology, 32.6% evaluating where consolidation makes sense, and 26.4% still preferring point solutions — a genuinely split market where timing determines whether you are an addition or a displacement
- Funding and revenue-model changes, since 87% of finance leaders say their company plans to add more consumption-based products, and every pricing-model change forces a systems review
- Competitor complaints and renewal-window chatter, which tell you an evaluation is already open rather than hypothetical
A generic list of finance titles at target-size companies gives you none of that. The signal tells you when, and the CFO data tells you what to say when you get there. This is the core of how Updately is built — capture the warm signal, enrich and score it against ICP, research the account properly, and write the message in your voice rather than a template's — and it matters more in a finance-governed deal than anywhere else, because the cost of an early, badly-aimed touch into finance is a closed door rather than an ignored email.
Bring the measurement plan into the first conversation
Finance leaders are accountable for AI ROI across the whole company. Handing them a measurement plan is handing them part of their own job description already completed.
A usable one is short: the baseline metric, the observation window, the attribution method, the threshold that counts as success, and the decision you will both make if it is not met. Vendors who can commit to a kill criterion are dramatically more credible than vendors who cannot, precisely because almost nobody offers one.
A 30-day plan for a finance-governed account
If you want to act on this before the quarter closes, this sequence is realistic for a small team.
- Week one — audit your own evidence. Write down, honestly, what your product touches, logs, retains and exports. If you cannot answer the four blocker questions in the table above in plain language, you are not ready to be in front of finance, and no amount of sequencing fixes that.
- Week one — segment your open pipeline by gate. Mark every open AI-related deal with whether finance has been engaged. The unengaged ones are your stall risk, not your late-stage forecast.
- Week two — build the finance-specific message. Same product, different proof. Control, auditability, integration behaviour, admin burden. No capability adjectives.
- Week two — set up the signals. Hiring, consolidation commentary, pricing-model changes, competitor complaints. Score them against ICP so volume does not drown relevance.
- Week three — multi-thread the top twenty accounts. Champion gets the leverage story, finance gets the control story, and the two messages must not contradict each other, because they will be compared.
- Week four — run the measurement conversation. Propose the baseline and the kill criterion before anyone asks. Track how many finance stakeholders reply, not how many emails you sent.
Takeaways
- Finance is now a category gatekeeper for AI, not a late-stage signature. Half of finance leaders are primary decision makers in their company's AI strategy, and 49% cite managing AI as the reason their role expanded.
- The deal is decided on control, not cost. Security (46%), internal expertise (42%), governance (42%) and integration (42%) all outrank funding (39%) as blockers.
- Scepticism is about your system, not about AI. Ninety percent of finance leaders using AI report positive ROI — but only 20.6% of revenue teams have production-ready deployments, and every CFO has watched that gap firsthand.
- Capability claims lose to operating evidence. Logs, gates, rollbacks, baselines and a kill criterion do more for conversion than any feature list.
- Signals tell you when finance is in motion. Governance hiring, consolidation commentary, pricing-model changes and renewal chatter are all visible from outside, and they are what separates a timely first touch from a wasted one.
- Multi-thread with two different stories. The champion buys leverage. Finance buys supervision. Send the same email to both and you lose one of them.
The teams that win AI deals for the rest of 2026 will not be the ones with the most impressive demo. They will be the ones who show up to the finance conversation already able to answer the question the CFO is actually accountable for: not "does this work?" but "can I explain what it did?"
Sources: Salesforce CFO Priorities Report (9 September 2026, n=865 finance leaders) · Salesloft 2026 U.S. Revenue Benchmark (1 September 2026, n=500 revenue decision makers)