Why I Don’t Believe in AI Sales Prioritization as a Silver Bullet
(and what we’re actually building instead)
It’s been almost five years since I last wrote anything long-form (and that was about a Chinese Autonomous Driving startup so buckle up because times have changed). Maybe that's because I've been heads down, or maybe because I’ve been surrounded by too many people confidently typing LinkedIn essays while their GTM systems burn quietly in the background.
Either way—this felt worth breaking the silence for.
Over the last few months, I’ve been watching a new wave of “AI for GTM” tools hit the market. The promise is familiar: stop chasing volume, start using AI to ruthlessly focus on your best accounts. They pitch a world where your reps wake up to perfect lists, complete with context, steps, and messaging—no thinking required.
I don’t doubt that the tech is improving, we’re seeing it in front of our eyes. I just don’t buy the narrative.
The pitch sounds right. The reality rarely is.
Let’s be clear: the pain these tools are solving is real. Reps are overwhelmed, lead scoring is outdated, and most sales teams are guessing who to focus on. AI GTM startups have built serious infrastructure to help cut through the noise. They don't just give you a score—they suggest the “why” and even what to do next.
But having sat on both sides of the GTM house— ops and sales —I’ve learned this: AI-driven prioritization only works when you already have a handle on the messy details of your own system. Otherwise, you're layering intelligence on top of noise.
Context still matters more than intelligence
At Plaid, roles, segments, products, or teams do not equate to specific functions. A “Partnerships” title might be responsible for part of core Enterprise pipeline. A “Lending” account might mean HELOC, BNPL, credit building, or all three.
Role ≠ function. Function ≠ influence. And influence isn’t a Salesforce field1
You can’t model your way around that. Not unless the system understands your org, your motion, and your edge cases—and most tools don’t. They abstract away context in favor of confidence. Not maliciously. Just… optimistically.
Which is why we’re building something different.
What we’re actually doing (and why it feels more honest)
We’re in the middle of building a Clay-powered GTM decision engine that connects to Salesforce, OpenAI, Gong, and more. It’s not finished yet, but we’re aiming for something more transparent and flexible than what the market is selling:
A taxonomy we define: vertical, sub-vertical, product relevance
Signals we control: hiring, funding, website visits, customer language
Workflows that reflect how we sell, not how a vendor thinks we should
It’s not “smart” out of the box. It’s manual. It takes work. But it’s designed to reflect our judgment—because that’s what actually drives good prioritization.
You can’t outsource prioritization if you haven’t done the thinking
There’s a pattern I’ve seen repeat: companies buy AI tools because they don’t have conviction on their ICP, or can’t align on segmentation, or don’t know which product to lead with. Then they expect the tool to tell them what to do.
That’s backwards.
The best GTM teams use AI to accelerate a strategy they already believe in, and iteratively improve on hypotheses that their expertise tells them they should test. The worst use it to delay the hard conversations.
Closing thoughts
If you’re buying an AI GTM platform thinking it’ll tell your reps exactly who to talk to, when, and what to say—you’ll probably be disappointed. Not because the tools are bad, but because GTM isn’t a data problem, it’s a context problem.
That doesn’t mean AI won’t help. It absolutely can, it’s helping us every day. But the magic doesn’t come from the model—it comes from how you design your system around the specifics of your company, your market, your motion.
We’re still early in building ours. But at least we’re honest about what it takes.
And I’ll take that over a “superintelligent pipeline agent” any day.
Note: Salesforce does have a “Campaign Influence” field but you get the point