Ask me which product is doing the most to reshape sales technology right now, and I won’t point to a sales technology company. I’ll point to Claude. Anthropic built it as a general-purpose model for knowledge work and coding, and the company keeps pushing it deeper into the enterprise.
Sales teams adopted it, and reps now lean on general-purpose assistants for prospecting research, account plans, competitive prep and deal strategy. When my company studied sellers’ day-to-day AI use, those assistants sat at the top of the list of tools sellers reach for, while software built specifically for selling was barely mentioned. Only one in seven sellers works with anything purpose-built for a sales team.
Gartner now predicts that 95% of sellers’ research workflows will begin with AI by 2027, up from less than 20% in 2024. Foundation models are already used for revenue work, so the question for sales leaders is what happens after the rollout. Most teams are learning the hard way that a brilliant general-purpose model and a working sales motion are two different things.
Handing Reps a Raw Model Isn’t a Strategy
In most companies, AI rollouts follow the same general process. The CEO signs the enterprise agreement, walks it down the hall, and tells the revenue organization, ” You have Claude now, go be better.”
I watched this happen at a well-known technology company, where a sales force of roughly 300 reps got exactly that directive and no further instruction. Within weeks, every rep was running a different homemade process, and the team’s best seller got pulled off quota to build prompts and automations for everyone else. The company converted its top revenue producer into an engineer. Multiply that across an industry, and you have the current state of AI in sales: enormous capability with zero process.
Sellers are feeling it. In our seller research, 76% said leadership is pressuring them to do more with AI, but 44% received neither tools nor training alongside the mandate. So they improvise with whatever they have, which is usually the general-purpose assistant their company already pays for.
Selling Isn’t Engineering
Customizing an assistant like Claude means writing instruction files, connecting data sources and maintaining it all as products and pricing change. That works well for engineers, because configuring systems is their job. It’s a bad fit for sellers.
The average account executive has no interest in learning what a markdown file is between customer calls. When companies push that work onto reps anyway, they break the thing that makes a B2B sales organization run in the first place: discipline. Great sales teams execute a consistent motion, with the same qualification, the same message and the same process across every rep. Do-it-yourself AI produces 300 private processes instead.
Most of the AI conversation has been driven outward from the engineering world, and engineers think in systems: inputs and outputs, ones and zeros. They also love showing off their automations.
Sellers are the opposite. Their work runs on live conversation, relationships and judgment, and plenty of them would rather nobody knew how much AI helps them at all. They want the results without the audience. Hand an engineer and a seller the same company data, and they’ll draw different conclusions, because they value different things. An assistant trained to satisfy everyone splits the difference, and the seller pays for it.
What Sellers Want From an AI Teammate
None of this means sellers dislike AI. We surveyed 500 B2B sales professionals about working with an AI teammate, and their top concern was accuracy, while fear of being replaced ranked last. Asked which kind of intelligence would make an AI teammate most valuable, 72% chose reasoning over content generation.
They want strategic advice and pattern recognition from the technology and will happily work alongside AI that knows their deals and gets things right. They have little use for a fluent writer with no idea what stage the opportunity is in.
The Engine Still Needs a Vehicle
The answer isn’t another foundation model, nor is it waiting for Anthropic to become a sales company. Anthropic builds engines, but an engine isn’t a vehicle. Nobody hands a delivery driver a crate engine and asks them to get the packages out, and nobody should hand a rep a raw model and ask them to make quota.
The engine needs a vehicle built around it for the job: purpose-built AI that connects to the CRM, knows the products and the pricing, follows the sales process and thinks like a seller about the deal.
Purpose-built doesn’t mean starting over, either. The strongest sales AI now runs on foundation models underneath, which means the enterprise Claude investment isn’t wasted. It’s just the beginning of the stack.
Keep The Model, Add The Layer
Sales leaders need to take a practical approach. Keep the foundation model and put purpose-built AI between the model and the rep. Then judge that layer by whether it knows your deals, fits how sellers work and produces answers accurate enough to trust.
Organizations that build the vehicle are already seeing the payoff. Gartner finds that sales organizations giving sellers AI-enabled next-best actions are 2.6 times more likely to achieve commercial growth.
The teams that pull ahead next year will close the distance between a general-purpose engine and a specific quota, with purpose-built AI that thinks like a seller, works like a teammate and lets Claude do the job Anthropic built it for.
Matt Darrow is the CEO and co-founder of Vivun, creator of Hero, an AI sales teammate designed to help sellers navigate and win complex deals. An engineer by training, Matt has more than 15 years of experience across enterprise software, sales engineering, and product management, including leadership roles at Zuora and Deloitte.
