The news
Writing in Demand Gen Report, Andrea Tarrell — CEO of Sercante and President of Technology Services at Trilliad — argues that underperforming revenue tech is a people and process problem, not a software problem. Research from Netguru puts the cost of disconnected data at 20–30% of annual revenue, while sales reps spend only 16% of their day actually talking to customers. The rest gets eaten by admin friction from a fragmented, half-used stack.
Our take
The article is describing an operator problem dressed up as a technology problem — and that distinction matters enormously when AI enters the picture.
Here's the mechanism: AI features in your CRM, MAP, or ABM platform are not self-configuring. They're multipliers. Feed them clean data, documented processes, and consistent inputs, and they return real leverage. Feed them a half-populated CRM and automation sequences no one is actively managing, and the AI layer compounds the mess. It doesn't fix it.
This is why so many AI rollouts stall at the "toggled off" stage. The feature exists. The license is paid for. But the underlying data model was never cleaned up after the last migration, the lead stages haven't been redefined since 2021, and no one can explain in writing how a contact actually moves through the funnel. That's not a deployment blocker the vendor can solve.
The instinct when results lag is to chase the next tool — or to assume you picked the wrong one. But the pattern Tarrell is describing isn't a purchasing problem. It's a readiness problem. Before any AI feature can do useful work, someone has to answer two questions: what decision does this output inform, and what does the data need to look like for that output to be trustworthy?
Most GTM teams haven't answered those questions for their existing stack, let alone for the AI layer sitting on top of it. The AI features aren't underperforming. They're accurately reflecting the state of the operations underneath them.
The so-what
Stop evaluating AI features until you've evaluated the process they're supposed to automate. If you can't write down the steps a human currently follows — and point to where the data lives at each step — the AI version of that workflow will fail on the same fault lines. The teams getting real output from their revenue tech aren't running better software. They're running documented, maintained operations that give AI something to work with.
Your stack isn't the problem. Your processes are. Fix those first, and the AI features you already have will start earning their keep.