Clean CPQ data is becoming the silent precondition for agentic selling, and the gap between companies with it and companies without it is about to widen.
The failure point for an AI sales agent in a complex sale is the product logic underneath it. Threekit’s own launch release pitches its agent on flagging conflicts before they become errors and catching a rule violation the rep would have shipped. In our reading, an agent can only catch an invalid configuration when the configuration rules and pricing logic it reasons over are complete and correct. Where CPQ data is inconsistent or constraints are missing, the agent inherits those gaps.
Threekit’s AI Sales Agent launch this week, aimed at manufacturers selling complex configured products, is a partial response to this dynamic. The agent accepts inputs from notes, voice memos, photos, RFPs, and line drawings, working toward a valid quote without waiting for a complete specification. Threekit CEO Matt Gorniak said revenue for the AI Sales Agent has grown more than 229% in the last year, and the company says its platform handles more than 10 million configuration-to-quote sessions monthly. But the agent’s accuracy depends on the underlying product data being structured correctly to begin with.
For sales leaders evaluating agentic investments in complex-sale environments, the CPQ layer is the prerequisite audit. Companies that have already modernized their product configuration and pricing logic, as documented in the recent wave of CPQ modernization, are better positioned for agentic deployment. Those that have not will find that agents surface the data quality problem faster than they solve it.
In our reading, infrastructure debt in CPQ is now a deployment gate for AI as well as a technical liability. The revenue teams with the cleanest product data will see the fastest agentic ROI. Everyone else has a sequencing problem.
