Ungoverned AI in Sales: The $10B Risk No One Is Talking About

Forrester’s 2026 B2B predictions report landed late last year with a number that should stop every sales leader cold. Ungoverned generative AI in commercial applications will cost B2B companies more than $10 billion in enterprise value this year. Declining stock prices. Legal settlements. Fines. Client refunds. And ungoverned AI in sales is one of the most exposed parts of that equation.

Around the same time that the report was published, a global consulting firm refunded a client hundreds of thousands of dollars for a project deliverable packed with AI hallucinations and, as Forrester put it, “slop.” Not a startup experiment. A major professional services firm. The kind of firm that charges a premium precisely because buyers expect expert-level accuracy.

That refund is what $10 billion looks like at the individual deal level.

Most coverage of this prediction has focused on marketing content teams and product development workflows. Both are real risk surfaces, and both deserve attention. But there’s a third one nobody is writing about, and it may be the most exposed of all.

It’s your sales floor.

The most unprotected AI surface in your business

Think about where AI-generated output actually reaches a buyer. In marketing, content goes through review. An editor reads the blog post. Legal reviews the case study. Someone approves the campaign before it runs. In product, features go through QA. Engineers test the output before it ships.

Now think about what happens when a rep gets an AI-generated answer during a live sales call. There’s no review step. No editor. No approval queue. The rep reads what the tool surfaces and says it out loud to a buyer who is actively evaluating your company, in real time. And if the AI hallucinates a product capability or gets a competitive comparison wrong, the buyer doesn’t think “that AI tool made an error.” They think your company doesn’t know what it’s talking about.

They wonder what else you got wrong. They go do their own research, find the inconsistency, and mention it to your competitor on the next call.

That’s not a hypothetical. It’s how deals die on calls every day.

Buyers are already losing trust

Forrester’s State of Business Buying 2026 report quantifies something sales leaders are starting to feel but haven’t been able to name. Nineteen percent of buyers using AI tools to research purchases feel less confident in their purchasing decisions because of inaccurate or unreliable AI-generated information. For procurement professionals, that number climbs to 28%.

Nearly one in five buyers is actively less confident because of AI. Not despite it. Because of it.

This is what makes the sales floor so dangerous. Buyers are doing their own AI-assisted research before your rep ever gets on a call, encountering unreliable information along the way, and arriving already calibrated for skepticism. Then your rep delivers an AI-assisted answer that contradicts something the buyer heard from your competitor, or cites a feature that shipped differently than described, and that skepticism turns into a decision.

The irony is almost too clean. The tools designed to make reps more credible are, in a meaningful percentage of deals, making buyers trust them less.

There’s a generational dimension here, too. Sixty-four percent of B2B buyers in 2026 are Millennials or Gen Z, according to Forrester. These are digital-native buyers who can recognize AI-generated content and are more likely to pressure-test claims than any previous buying cohort. When they catch an inconsistency, they don’t give the rep the benefit of the doubt.

What Forrester actually identified about AI governance, and why sales is most at risk

Forrester’s core finding isn’t just “AI is risky.” It’s more specific, and more useful.

The problem, as they frame it: existing top-down governance models were built for internally developed applications. They don’t work for AI embedded in third-party commercial tools.

That distinction matters. When a company builds an internal application, governance is baked into the development process. Requirements are reviewed. Outputs are tested. There’s a chain of accountability. But when that same company deploys a third-party AI sales tool, those governance assumptions don’t transfer. The tool ingests content, runs it through a model, and surfaces answers. Who reviewed those answers before the rep delivered them? Usually nobody.

In sales, this shows up in a specific, repeatable pattern: a tool ingests unreviewed documents, generates answers on the fly, and surfaces them to reps during live calls with no validation layer between the model and the buyer. The company hasn’t approved what the AI says. The rep doesn’t know to question it. The answer leaves the building.

The scale problem

One rep, one call, one wrong answer. Recoverable. The rep follows up, corrects the record, moves on.

Fifty reps, 300 calls a week, all powered by the same unvalidated knowledge base. That’s the version Forrester is pricing at $10 billion across the industry.

When AI hallucinations propagate across a full sales team, three things happen simultaneously. Wrong competitive positioning becomes an organizational habit, with reps confidently delivering incorrect differentiation on every call. Stale competitive information becomes organizational fact, with the AI surfacing answers based on last quarter’s landscape because nobody updated the knowledge layer. And the damage becomes invisible, buried in closed-lost notes that never trace back to a root cause.

There’s no firewall. You approved the tool. You didn’t approve what it’s saying.

The questions a VP Sales should be asking right now

This isn’t an argument against AI in sales. It’s an argument for knowing what governance actually looks like before your next bad call becomes your next lost deal.

Forrester Chief Research Officer Sharyn Leaver framed the mandate clearly: “Success will hinge on investing in AI governance, balancing human expertise with AI tools, and empowering teams to deliver clear, validated outcomes.”

For a sales leader, that translates to four specific questions worth asking of any AI tool that touches your reps’ live conversations.

1. Who reviewed the source material before the AI surfaces it to reps? Not who uploaded it. Who reviewed it for accuracy and approved it for use on calls?
2. When your product changes, how does the AI’s knowledge update, and how fast? If the answer involves someone manually editing a document, you have a stale-knowledge problem waiting to happen.
3. Can you audit what the AI told a rep on a specific call? If you can’t trace an answer back to a source, you can’t fix wrong answers. You’re governing blind.
4. Does the system generate novel answers, or deliver pre-validated ones? There is a meaningful difference between a system that produces a fluent-sounding response from a language model and one that delivers an answer a human expert reviewed before any rep ever heard it.

These aren’t a product evaluation checklist. They’re the minimum for running real-time sales enablement at scale without creating exactly the liability Forrester is describing.

The buyer at the other end of the call

Buyers are getting better at this faster than most sales teams are. They’re researching vendors before the first call. They’re reading the same analyst reports your reps are reading, sometimes more carefully.

When a purchase involves AI-powered features, Forrester data shows buying groups can swell to more than 20 stakeholders. More people in the room means more people who heard the wrong answer. More people will remember it.

When your rep delivers an AI-hallucinated competitive comparison that contradicts what the buyer heard from the competing vendor directly, you don’t just lose the deal. You lose the account, the reference, and the next intro they might have made.

The $10 billion isn’t an abstract figure. It’s the accumulated cost of confident answers that nobody validated, delivered to buyers who were already watching closely.

The bottom line

The hype era of AI in sales is over. What replaced it is an accountability era, and most sales teams haven’t noticed the shift yet.

The question is no longer whether your reps are using AI on calls. The question is whether anyone validated what that AI says before a buyer heard it. If you can’t answer that clearly, you already have a governance problem. The only variable is how many deals it’s in.