About 11 months ago, a few of us started arguing.
Not about a specific product or a specific company. About a question that felt increasingly urgent the longer we sat with it: what actually happens to revenue when AI gets good enough to replace most of what revenue teams do today?
There were four or five of us in that first conversation -- product builders, GTM operators, people who'd spent careers inside sales organizations and product orgs, watching the same patterns play out across different companies at different scales. We'd all been consuming the same research. Reading the BCG reports. Digging into the Gartner predictions. Arguing with each other about what was signal and what was noise.
We disagreed on almost everything. Which is why we kept going.
The three questions we couldn't resolve
The first was about buyers. The dominant narrative -- that B2B buyers were going self-serve and sales reps would become irrelevant -- felt too clean. Our experience said something more complicated was true. Buyers want to self-serve. They also make worse decisions when they do. The research supported this contradiction and nobody was talking about it honestly.
Gartner's 2025 research predicted that by 2030, 75% of B2B buyers will prefer human interaction over AI -- a direct reversal of the self-serve trend that's been building for a decade. At the same time, Forrester confirmed that more than half of all large B2B purchases -- deals over a million dollars -- would flow through digital self-service channels by 2025. Both of these things are true simultaneously. They are in direct tension. Nobody had a clean theory for how they resolve.
The second question was about data. We'd all seen the same disaster from different angles: organizations investing in AI tools, only to discover the data those tools ran on was fundamentally unreliable. Validity's 2025 State of CRM Data Management report surveyed 602 CRM users and found that 76% of organizations had less than half their CRM data accurate and complete -- and yet 90% considered CRM data the cornerstone of their operations. You're building AI on a foundation of sand and calling it intelligence. Nobody in the vendor community wanted to say this out loud because it undermines the product narrative.
The third question was the hardest: what is the actual human advantage in selling, once AI can do everything mechanical? We had intuitions. We disagreed on the specifics. The research was genuinely unclear.
What the data kept showing us
The more we dug, the more a consistent pattern emerged across the research:
"AI adoption is growing fast. Results are not growing at the same rate. And nobody is asking why."
MIT Sloan's 2025 State of AI in Business report found that while over 80% of organizations had explored or piloted general AI tools, only 5% of enterprise-grade AI systems reached production. The core barrier wasn't budget, regulation, or talent. It was that most AI systems don't learn from context, don't retain feedback, and don't improve over time. They're reactive, not adaptive.
Meanwhile, Vidyard's State of AI in GTM report found that 93% of GTM teams say they're using AI in some form -- but only 49% use it in day-to-day operations. That gap is enormous. It's the difference between having a gym membership and actually training.
| Metric | Number | Source |
|---|---|---|
| Organizations that explored/piloted AI tools | 80%+ | MIT Sloan 2025 |
| Enterprise AI systems that reached production | 5% | MIT Sloan 2025 |
| GTM teams "using AI in some form" | 93% | Vidyard 2024 |
| GTM teams using AI in daily operations | 49% | Highspot 2025 |
| CRM data that is accurate and complete | <50% at 76% of orgs | Validity 2025 |
The disconnect between adoption and outcomes was the thread we kept pulling.
What we decided to do about it
We are not analysts. We are not academics. We are practitioners -- people who have lived inside these systems, built products for them, sold to them, and watched them fail in predictable patterns. We have opinions. We have research we've been sitting with for the better part of 10 months. And we have a lot of questions we're still arguing about.
We decided to write it down.
Not as a whitepaper. Not as a vendor perspective on the market. As a running debate -- our honest, sometimes uncomfortable, sometimes contradictory thinking about where revenue is going and what it actually takes to build organizations that thrive in that world.
Some of what follows will be wrong. We expect that. We're inviting disagreement. The questions feel more important than our answers.
Here's what we're going to explore across this series:
- Why the data foundation of most revenue organizations is structurally broken -- and why adding AI makes it worse, not better
- How the buyer changed in ways most revenue teams haven't absorbed -- and the counterintuitive implications
- Why AI copilots are a category-level design mistake -- and what the right architecture looks like
- What actually happens to the sales profession -- not the optimistic version and not the catastrophist version, but the honest one
- Why organizational memory is the most undervalued asset in revenue -- and almost no one is building it
- What the actual competitive moat looks like when every tool is available to everyone
We're not here to sell you something. We're here to think through something hard, in public, with whoever wants to argue with us.
If you've been wrestling with the same questions, we'd genuinely love to hear your pushback.
This is Post 1 of 13. The next post asks a question that should make every revenue leader uncomfortable: Is your CRM a system of record, or a system of lies?