From The Agents, our weekly podcast on how we deploy and run AI agents at SaaStr, and how to do it yourself.
The other night I was talking with an AI CEO and their head of marketing, and they told me the same thing I hear constantly: “We tried AI agents on outbound. It doesn’t really work.”
It almost always comes down to one mistake. They pointed the agent at the wrong leads.
So I walked them through the one heuristic that fixes it, the same one we use at SaaStr. It’s simple enough to explain in a sentence, and it’s the difference between an agent that looks like a toy and an agent that prints money.
Here it is: sort your leads into A, B, C, and D. Then put the agent on the B leads. Not the A leads. The B leads.
Your A Leads Don’t Need an Agent #
Your A leads are so hot a human falls out of bed for them.
Someone fills out the form and says “I have a million-dollar budget and I’d like to sign this week.” Your laziest rep, the one at the mall, the one half-watching a movie, responds to that email in 60 seconds. Nobody on your team is letting a million-dollar inbound sit.
So what does an agent add there? Nothing. You’re automating the one part of the funnel that was never broken. Worse, you risk an agent stepping on a deal a human was always going to win anyway. Leave your A leads to your humans. They’ve got it.
This is where most teams start, because it feels safe and obvious to put the shiny new agent on the best leads. It’s exactly backwards. You see no lift, you conclude agents don’t work, and you turn it off.
Your B Leads Are Where the Money Is #
B leads have real signal and a real score. They engaged with something, they fit the profile, they’re not random. They’re just not hot enough to be worth a human’s time.
And here’s the structural truth about every company of any size: humans never follow up with the B leads. Not because your reps are lazy, but because the incentives are rational. A rep with a quota and limited hours is going to spend those hours on the A leads that close this quarter. The B leads sit in the database forever. There’s signal in them, but not enough to justify a human picking up the phone.
That pile is exactly what an agent is built for. It has infinite patience, no quota anxiety, and no preference for the easy win. It will work a lead that’s worth working but not worth a human’s afternoon. It will follow up three times when a human follows up zero times.
The C and D leads are a longer conversation. Some of them have gold, most don’t, and sorting that out is its own project. Don’t start there. Start with the B leads, because they’re already sitting in your CRM with contacts you already have. You don’t need to buy new data. You need to work the data you’ve been ignoring.
What This Is Worth: $500K for Us, Millions for You #
We run an agent (Artisan, in our stack) on our B leads. Past attendees and contacts with a score, a signal, some reason to believe, but not enough for a human to chase.
That agent is $500K for us. We’re a small team, that’s a small slice of our database, and it’s not even our core business. $500K is the difference between catering the team lunch and bring-your-own-sandwich.
Now scale that. We’re three humans working a niche events database. If you’re a real go-to-market org with tens or hundreds of thousands of scored-but-ignored contacts, the same motion isn’t worth $500K, it’s worth millions. The B-lead pile grows with the size of your funnel, and at scale almost none of it gets touched. That’s not a rounding error. That’s a second pipeline sitting in your CRM that nobody is working.
The math is simple: take your scored leads that no human follows up with, multiply by even a low conversion rate and your average deal size, and you’ll find the number is a lot bigger than you’d guess. For most companies at scale, it’s the largest pile of free pipeline they have.
How to Actually Make It Work #
Pointing the agent at the right segment is most of the battle. The rest is training:
Segment tightly. Don’t hand the agent “the database.” Hand it a specific campaign with a specific story. For us that’s something like “alumni of SaaStr Annual 2024, here’s exactly what’s different in 2026.” Narrow segments with real context produce specific outreach. Broad segments produce generic spam.
Feed it context, not just contacts. The agent should know why this person is a fit, what changed since they last engaged, and what’s relevant to them now. Fresh, specific context is the entire difference between a reply and a delete.
Let it find lookalikes. A good outbound agent is also good at “find me more companies that look like the ones I already closed.” That’s how the B-lead motion expands into net-new pipeline without you building a single list.
Pick a tool and train it. Artisan is what we use, but it’s not the only vendor that does this well. The choice of tool matters less than the discipline of training it on the right segment with the right context.
Why Humans Will Never Really Do This (And Why That’s the Point) #
The reason this works isn’t that the agent is smarter than your reps. It’s that the agent does the work your reps are structurally incentivized to skip.
A human rep, when a deal is slipping, panics and discounts. A human rep, faced with 500 B leads, works 5 of them and lets 495 rot. A human SDR quits every few months and takes the context with them. None of that is a character flaw. It’s what the quota and the comp plan produce.
The agent has no quota anxiety and no preference for the easy win. Give it the leads humans won’t touch, the patience humans don’t have, and the context humans never bothered to load, and it goes and finds the gold everyone else walked past.
It’s Not Just Us: Kyle Norton at Owner and the ICONIQ Data #
If you think the B-lead motion is a SaaStr quirk, look at the operators running it at scale. Kyle Norton, CRO of Owner.com, is putting up numbers that only make sense if you stop letting pipeline rot. Owner runs $2M+ in ARR per rep per year, roughly 4x the per-rep ARR of their direct SMB competitors, and over $100K in closed-won revenue per outbound BDR per month. Not pipeline. Closed revenue, per BDR. And they sell roughly $10K deals to mom-and-pop restaurants, the kind of small business that supposedly doesn’t buy software. His team built their entire cold outbound infrastructure in three weeks, and his booked revenue per dollar spent, on a per-rep basis, runs 3x any team he’s ever managed.
Kyle’s framing maps directly onto the A/B/C/D model. He’s blunt that AI agents are now better than a mid-pack AE or SDR, the 50th-percentile rep. That’s exactly the rep who was never going to work your B leads anyway. The leverage isn’t replacing your best closers. It’s covering the work the median rep skips.
His one warning is the one that matters most: don’t let reps run their own agents. Kyle maps companies on a ladder from Level 0, reps using ChatGPT as a smarter search bar, up to Level 4, a system that improves itself. Most companies stall at Level 1, where individuals build their own thing and great ideas never scale. The leverage comes from centralizing it: one GTM AI owner training the agent for everyone. Same lesson we learned the hard way. Pick a tool, train it, and don’t hand it to ten reps to figure out alone.
ICONIQ’s 2026 GTM data says the same thing from the top down. Across 150+ B2B revenue leaders, teams with strong AI adoption are hitting quota at 67% versus 59% for everyone else, and running materially leaner on GTM headcount. And here’s why the B-lead pile matters more than ever: the funnel is breaking at the bottom, not the top. Top-of-funnel held steady, but demo-to-close conversion is down 5 to 10 points year over year, sales cycles run 3 to 4 weeks longer, and pipeline coverage slipped. When every deal is harder to close, leaving scored leads untouched costs more than it ever has.
The Playbook in Bullets #
Sort leads into A, B, C, D. Put the agent on the B leads.A leads already get a human response in 60 seconds. An agent there adds nothing and risks stepping on a sure thing.** B leads have real signal but aren’t worth a human’s time**, so humans never work them. That’s the gold.** The B-lead pile is already in your CRM.You don’t need new data, you need to work what you’re ignoring. For us it’s $500K off contacts we already had.At real scale, with tens of thousands of ignored scored leads, the same motion is worth millions. Segment tightly and feed real context.Narrow campaigns with a specific story beat “the database.” Let the agent build lookalikes to expand B-lead outreachinto net-new pipeline. It works because the agent does the unglamorous follow-uphumans are incentivized to skip, not because it’s smarter. Owner runs $2M+ ARR per rep and $100K+ closed-won per BDR per month**. The upside of working the pipeline humans skip is not theoretical.** AI agents are now better than the median rep (Kyle Norton, Owner). That’s exactly the rep who ignores your B leads. Centralize the agent. Don’t let reps run their own**. One owner training it for everyone beats a thousand local experiments that never scale.** ICONIQ 2026: AI-forward teams hit quota at 67% vs 59% and run leaner**. With conversion down across the funnel, ignoring scored leads is costlier than ever.
We go deep on this and the rest of our agent stack every week on The Agents. If you sell to operators who are actively deploying agents, that’s the room, and we’re taking a small number of sponsors: saastr.ai/media-sponsors.