notes · GTM

The empty seat at the bottom of the price sheet

May 5, 2026 · Benjamin Dysin

I spent the last two weeks auditing every company I could find that sells AI-moderated qualitative interviews. There are nine of them serious enough to argue with. Listen Labs has raised $100M at a $500M valuation. Outset is at $51M total. Strella did $1.6M in ARR in their first year of monetizing, with 150% net dollar retention. Conveo is YC-backed with 400+ enterprise customers. The category is not a graveyard. It’s hot.

Six of those nine refuse to tell you what they cost.

That should be the first thing you notice. Not “what do they charge,” but “they don’t want you to know unless you book a call.” Listen Labs’s /pricingpage returns a 404. Outset says “get custom pricing.” Strella, Glaut, Feedbk, Inca, same shape. Of the three that do publish numbers, one (User Intuition) is hybrid pay-per-interview, and only two (Merren at $39/mo, Koji.so at €29) have anything that looks like a real self-serve plan a credit card can finish.

For a category whose entire reason to exist is replacing the human cost of running 1:1 interviews, the sub-$50/mo tier is essentially empty.

I want to argue that this is not a coincidence, and I want to argue that the standard explanation (“enterprise sales is more profitable”) is the one place this conventional wisdom is actively wrong.

The standard story is incomplete

The story you’ll get from any GTM consultant is: B2B SaaS converges on sales-led for high-ACV verticals because (a) the deals are big enough to support a sales team, (b) procurement gates exist anyway, and (c) you make more money the less the price is anchored in public. All true. None of it explains why a category whose core differentiator is doing something cheaper and faster than humans would refuse to put a number on it.

If your product is “AI moderator instead of a $200/hour researcher,” the price tag is the demo. You want it visible. The fact that nobody in this category does that suggests something else is going on.

What I think is actually going on, after staring at the landing pages of all nine of them long enough that I started to dream in their hero gradients:

They’re priced as panel-bundled enterprise software because they’re scared to be priced as software.

Listen Labs, Outset, Strella, and Conveo all bundle a participant panel: 8M to 30M people they can recruit on your behalf. That’s the part that isn’t software. That’s a marketplace. Marketplaces have to be sold by humans because the value depends on which segment of the panel the buyer needs reach into, and that conversation can’t be a pricing table. So they wrap the AI-moderation work inside the panel-bundling sale and the whole thing goes opaque.

That’s the structural reason. And once you see it, you can predict who in any “AI replaces $X” category will go opaque: anyone whose product is wrapped inside a higher-friction sale. AI legal review firms wrap into white-shoe-firm relationships. AI medical scribes wrap into EHR procurement. AI coding tools that bundle compute go opaque; AI coding tools that don’t (Cursor, Claude Code) publish prices on the homepage.

The opinion you can disagree with

Here’s the first one. Most “demo to learn pricing” gates are not strategic. They’re vestigial. They were necessary back when the founder did need to learn whether the prospect was real before quoting. Now they exist because nobody wants to be the first to put a number down. It forces the others to compete on it, and the in-category etiquette is to all stay quiet together. This is a coordination failure, not a pricing strategy. It is breakable by literally one company breaking it.

Here’s the second. The companies most loudly arguing that “you can’t sell qualitative research as self-serve, the buyer needs handholding” are mostly arguing this because their COGS structure (panel-bundled, priced like a service firm) doesn’t survive self-serve. A pure-software AI interview platform has a per-session cost in the high-cents-to-low-dollars range. Mine clocks ~$0.31 on free tier and ~$0.72 on Pro at a 15-turn baseline, on Anthropic May-2026 pricing. That’s an 80%+ gross margin at $5/session, and there is nothing about having a credit card form on a /pricing page that breaks the math.

These two together are the wedge. Not “we have a feature they don’t have.” We do, but features in this category get cloned in a quarter. The wedge is “we will name our price, and they structurally will not.”

What this means if you’re not building an AI interview tool

The interesting thing about pricing-opacity-as-wedge is that it generalises badly to any category where the leaders look like they’re winning. They are! They’re funded! They have logos! It’s hard to look at Outset and Strella and Listen Labs and not assume that the whole shape of their business (sales-led, panel-bundled, demo-gated) is the correct shape for this category. So everyone smaller copies the shape. The smaller they are, the dumber this is, because they don’t have a panel and they don’t have a sales team and they’re imitating the cosmetic surface of a business model whose actual machinery they don’t run.

This is a very common pattern. I’d argue it’s the most common pattern in B2B SaaS. Small companies modeling themselves on the surface morphology of large companies, missing that the surface morphology is downstream of operations they don’t share. Stripe Atlas and Pilot and Mercury all looked, for years, like they were “modeled on JPMorgan but smaller.” They weren’t. They were software companies, and the software-company shape (transparent pricing, self-serve, public docs) was the correct shape for them all along, and it’s how they compounded past the incumbents.

If you’re sitting in a category right now where every funded competitor has a “book a demo” button and you’re trying to figure out whether to copy them, the question to ask is: what part of their stack is not software?If you find one, the gate is load-bearing. If you can’t find one, the gate is vestigial, and your wedge is to remove it.

The bit where I admit what I don’t know

I haven’t run this experiment yet. I have a /pricing page with four tiers on it, the cheapest of which is $19/mo (BYO-Anthropic-key) and the most credible of which is $99/mo Pro. I have shipped six vertical landing pages. I have done zero discovery calls.

The thing I might be wrong about: maybe the buyers in this category have been so trained to expect “book a demo” that a published price reads as “not serious.” Maybe the panel-bundled players have already taught the market that the un-bundled version isn’t a real product. Maybe (and this is the genuinely scary one) the people who would buy a $99/mo AI interview tool have already settled for a Google Doc and a Loom, and the published-pricing wedge wins zero net new buyers because the buyers who’d convert weren’t going to demo anyway.

I’ll know in about six months. The plan is to publish, run discovery calls, and watch what actually converts. I’ll write that one too.

But I am fairly sure that “everyone in the category is opaque” is not the same as “the category requires opacity.” And the gap between those two is where the next wedge in any opaque-by-default market is going to come from.