AI Promised Sales Teams More Time to Sell. Right Now, They Have More to Manage.

At the quarterly vendor review, the AI line items tend to look healthy. Seats are filled and token usage climbs month over month, showing that the company is buying and using AI while leaving open whether any of it helps reps sell. That’s the unresolved problem at the heart of the sales tech boom: quota attainment has slid as tools multiply, and reps can end up spending more time keeping software running than talking to buyers. AI’s value depends on whether it frees up those conversations, once the time spent checking its output is counted.

Robert Blackler is Chief Revenue Officer at the data company Beauhurst, where he oversees new business, account management, customer success, and operations after previously running marketing. He’s building his operations function for what he calls "territory that is beyond anyone’s experience," and he wants AI judged by whether it puts reps back on the phone with buyers.

"There’ll be a decluttering of wasteful resource and token usage across companies that will need to be trimmed back when understanding the value," Blackler says. He expects AI to go through the same recurring cost-and-value review that finance applies to vendor contracts, with revenue operations running it. That review has to account for a cost that rarely appears on an AI invoice: the hours people spend checking and supervising the tools’ output.

AI inherits the CRM’s maintenance problem

Blackler has led enough CRM cleanups to know that no system stays fit for the business it was built for. Every company configures its CRM around its own qualification methodology and pricing, and then the business keeps moving. "Inevitably these systems also evolved as the business evolved and got cluttered with needless and useless requirements on sales teams, or actually missing and incomplete requirements that they should be capturing,” he says. “They never stay fit for purpose without this kind of ongoing maintenance." Larger companies hire dedicated admins to keep tweaking indefinitely, while most smaller businesses let the system sit “until something’s a pain enough to then go and deal with it.”

"This is similar to how I think AI will need to be managed," he explains. Under his approach, AI gets the scrutiny finance gives a vendor list, where every team justifies its spend and the thumbscrews tighten wherever the value is thin. "We’re going to need to do the same thing on AI usage, token usage, and adoption across different teams. We have to sense check the value, and that’ll be a forever recurring problem. It will not be fixed in one big swoop."

Over 40% of agentic AI projects are forecast to be canceled by the end of 2027, with escalating costs and unclear business value among the reasons cited. Blackler’s proposed review addresses those same pressures at the revenue-team level.

Blackler sees operations taking on centralized responsibility for AI testing and implementation. "What’s been added is, I think, a centralization for the ownership of AI within a business, because they are typically the most skilled to test and build things where sales teams can be really varied," he says. He wants ops aimed at the problem he watched the last tooling wave create, when "reps were doing an awful lot to run tools instead of to pick up the phone." He acknowledges that measuring ops’ contribution here is "very, very hard and subjective, but it doesn’t land anywhere else as strongly as in operations."

Output still needs a human check

Cutting the AI bill only helps if the tools that survive the cut produce accurate results, and AI output can read as polished and confident while the facts underneath it are wrong. Catching that takes the kind of checking Blackler’s analysts at Beauhurst already do as a matter of routine, much as a copywriter reads a draft for narrative and structure. "That kind of skills base is also going to be required to avoid the slop misdirecting people’s efforts within a business," he notes.

"That sounds like a big thing, but it doesn’t have to be." One person with an accountant’s habits can reconcile manually held facts against what an AI tool produces. When the output runs to tens of thousands of lines, the job is to "cherry pick and look at outliers," because going through every line in a foolproof way isn’t realistic. "You need a kind of QA part to any strategic decision making that’s based on AI driven data," he adds. Those spot checks have a more useful reference point when the team has benchmarked its data changes beforehand, giving it a baseline against which to assess what the tool produces.

Managers are back on the calls

On the sales floor, AI now helps shape call lists and generates activity around them. ARR and TCV still measure a sales team’s success, Blackler points out, while the early indicators that explain how a team gets there are "changing massively," and his team has gone back to the conversations to understand what the numbers alone no longer reveal. "We for example here have brought back a heavy emphasis on things like call listening," he says. Managers review live calls to make sure reps aren’t missing opportunity signals or dodging objections and are explaining the value proposition well. "It used to be much more telling from the numbers and measurables than it is today."

AI can cut the time it takes to produce sales activity while raising the human supervision that makes it effective, and that second kind of productivity is limited by how many managers a team has. Sitting alongside reps is "manager intensive time," as Blackler puts it, and it strains under a high rep-to-manager ratio. He has watched the practice slip at many companies during growth and considers it more important now than it used to be. That makes manager headcount part of the cost of AI, alongside the token bill.

Onboarding is where Blackler sees AI working well, since it surfaces answers a new hire would otherwise dig out of tens of thousands of words. Senior skill-building stays with a person, though he has found AI useful for his own development. "Personally, I’ve already found it wonderful for picking up two or three things I’ve forgotten that I have done in the past," he recalls. Before an important meeting, he feeds his key points into an AI tool and asks what question he hasn’t asked yet. The judgment about what to do with the answer still belongs to the person in the room.