A full teardown of two go-to-market machines built on the same bet: that an autonomous agent can replace the sales development rep. One is the buzziest AI-native challenger of 2024 (11x), the other the category-defining incumbent it set out to kill (Outreach, once worth $4.4B). What's actually AI-native, where the venture story and the GTM reality split, why the disruptor's own product churned four of five buyers, and the three moves I'd hand 11x before the category writes it off.

In late 2024, 11x was the most exciting company in go-to-market. It had a $50M Series B from Andreessen Horowitz, a $24M Series A from Benchmark, and a manifesto that opened with a line built for screenshots: "In 1999, Salesforce invented cloud-based software. In 2024, 11x is starting to kill traditional software." Its product was an AI sales rep named Alice. Its pitch was that you would stop hiring SDRs and start hiring software that does the SDR's job. The company name is the pitch: each digital worker, the founder said, "will replace the work of 11 full-time employees."

Eighteen months later, the CEO of 11x went on a podcast and said this: "I don't think AI SDRs work. I don't think they work in the current form. And you're kind of hearing that from the CEO of the company that invented the word."

That is not a competitor talking. That is Prabhav Jain, the person running 11x, recanting the category his own company coined.

Meanwhile, the company 11x was built to replace is having its own bad decade. Outreach raised $200M in 2021 at a $4.4B valuation and called itself "an iconic company poised to be a leader for years to come." Since then: five rounds of layoffs, the founder stepping down, no IPO, and secondary-market shares trading at roughly 7 cents on the dollar versus the price institutional investors paid in 2021.

So you have a challenger whose product reportedly didn't work, and an incumbent whose business is being repriced in slow motion. Here is the part that makes this a teardown instead of a eulogy: in 2026, both of them are selling the exact same thing. The robot. Outreach's new CEO now tells buyers, almost verbatim, the 2024 11x pitch: "You're no longer just licensing a tool for your reps. You're hiring agents." I want to pull both motions apart and show you why one decayed, why the other detonated, and what the wreckage says about the "AI replaces the rep" thesis a lot of people are still funding.

The number I keep staring at is 11x's: a claimed "$14 million in annual recurring revenue when in reality, the number of contracts that passed the three-month trial period totaled only about $3 million."

That is a 4.7x gap between the number on the board slide and the cash that survived the quarter. It is also the entire story of the AI SDR category in one ratio: the promise was huge, the durable revenue was a fraction of it, and the difference got papered over with a contract structure.

Outreach's curve is slower and sadder. No fraud, no scandal, just a category leader watching the market reprice it. From a $4.4B private valuation to secondary shares at roughly $1.68, a 17.6x forward multiple on a business growing 11%, and five years with no liquidity event after the CEO promised to "think about public life at our own rate." Own rate, it turns out, meant five layoff rounds.

Six dimensions, loosely graded, plus a composite for how AI-native each company's own go-to-market actually is. Not how AI-native its marketing claims to be. (Loose on purpose. The grade is the hook; the qualitative read underneath is the point.)

The interesting thing here, and the reason this teardown is different from the first two in the Index, is that nobody gets an A. 11x is AI-native by construction and still only lands a B-, because a beautiful agent architecture wrapped in a demo-gated, human-AE-backed motion that churns four of five buyers is not an AI-native go-to-market. It is a venture story with a chatbot attached. Outreach lands a C+, dragged up by the one thing it owns that 11x never built: a decade of first-party workflow data. Let me walk the rows that matter.

11x's growth model was the cleanest articulation of a thesis a lot of smart people believed in 2023. Benchmark's Sarah Tavel had written the canonical version of it months earlier: "Rather than sell software to improve an end-user's productivity, founders should consider what it would look like to sell the work itself," priced "relative to the cost of a human performing the work," which she argued could make the addressable market "10-50x larger" than a per-seat SaaS model implied. 11x's own head of partnerships put it bluntly at the Series A: "We don't target SaaS spend; we target hiring budgets. We're in the business of selling work."

That is a genuinely good idea. If you can sell the outcome of an SDR rather than a tool for an SDR, your contract value isn't capped by software budgets, it's capped by payroll, which is a much bigger number. The demand-gen flywheel matched the ambition: a manifesto engineered for LinkedIn, human-named agents (Alice for email, Julian for the phone) that generated press at every funding round, and a founder, Hasan Sukkar, posting lines like "We don't build software. We build digital workers. We don't sell tools. We sell work outcomes. We're turning code into labor." The early dogfooding story was real and good: Sukkar said the company used Alice as its own primary growth channel to reach its first $1M in revenue. Alice selling Alice. That's the kind of detail that makes a teardown writer sit up.

And then the product had to actually work, on someone else's pipeline, past the honeymoon. Here is where the flywheel threw a spoke.

The TechCrunch investigation in March 2025, sourced to nearly two dozen investors and current and former employees, plus a companion Sifted piece two days later, laid out a motion that was optimized for the appearance of growth rather than growth. Three mechanics did the damage.

First, the contract. 11x signed customers to one-year deals with a three-month break clause and booked the full annual value as "contracted ARR" on day one. When 70-80% of those customers walked at the three-month mark, the CARR number stayed on the slide while the cash evaporated. As one CFO commentator put it, "everything was treated as ARR."

Second, the logos. 11x listed ZoomInfo and Airtable as customers, on its website, in sales calls, and, per ZoomInfo, "now even on its AI dialer." ZoomInfo's response is the most quotable rejection I've read in a while: "We did not give them permission to use our logo in any manner, and we are not a customer. During the pilot, 11x's product performed significantly worse than our SDR employees." A company that builds sales-intelligence tooling for a living ran the AI SDR against its own humans and the humans won. Airtable said it was never a customer either.

Third, the product. A former engineer told TechCrunch "the products barely work." The reviews rhyme: a Trustpilot user reported spending $9,000, sending over 1,000 messages, and booking zero meetings, with Alice targeting a senior Google manager because he "used to work" at a smaller company. Inc. called the whole thing AI's "Theranos moment." Sukkar stepped down as CEO six weeks after the story ran.

The detail I can't stop thinking about is in the careers page. 11x, the company whose entire pitch is that you should stop hiring sales reps, is hiring sales reps. Human AEs, $120-130K base. And the tools they're asked to know? "Outreach, Apollo, Salesforce." Not Alice. The company selling the replacement staffs its own pipeline with the incumbents it's replacing.

Outreach is the textbook version of the motion 11x tried to leapfrog, and you can read it straight off the company's own surfaces. Manny Medina built Outreach by coining a category. First "sales engagement," then "sales execution platform," with the explicit theory that "you create a tribe and a tribe becomes a movement and the movement becomes something that raises you." The tribe gathered at Unleash, a conference Medina admits was "pretty much templated out of the Tony Robbins Unleash the Power Within show." Around it: Gartner Magic Quadrant leadership, IDC-commissioned reports, annual sales-data studies, and a sales-led demo motion with no public pricing.

That motion built a real asset, and it's the asset 11x never had. A decade of first-party engagement data: every sequence, every reply, every closed deal, sitting inside the workflow where reps actually work. The customer logos confirm the altitude. Snowflake, Databricks, Siemens, ServiceNow. These are enterprise SDR organizations, the exact buyers who pay a median of $45,350 a year, per seat, on annual contracts.

So why is it breaking? Because the same per-seat model that compounded on the way up compounds on the way down. When the thesis of the entire market becomes "you'll need fewer reps," a business that charges per rep is short its own product. Outreach grew 11% in 2023 and got marked to a 17.6x multiple it can't grow into. In February 2023 the CEO told staff the company had ["failed to meet annual recurring revenue targets" while cutting 7% of the team, the second of five rounds.

The AI retrofit tells its own story when you line up the dates. Outreach shipped conversation intelligence (Kaia) in 2020, autonomous prospecting agents in December 2024 (one month after 11x's a16z round), joined Anthropic's Model Context Protocol ecosystem in February 2026, and launched a ChatGPT app in June. That's a company sprinting to bolt agents onto a stack built in 2014. My favorite confirmation is in the DNS: Outreach's own domain verifies an Anthropic integration token and runs marketing on Marketo and billing on Zuora, the mature enterprise plumbing of an incumbent, not the clean-sheet architecture of a disruptor. The agents are real. The question is whether agents grafted onto a per-seat contract can outrun a pricing model that loses money every time a customer needs fewer seats.

The customer reviews tell the lock-in story the way ZoomInfo's did in Teardown #1. Outreach gets praised for sequence depth and Salesforce integration, and hammered for being "an expensive email scheduler" with "hidden costs", auto-renewal traps where missing the cancellation window by hours triggers a full-year bill, and support that goes "nonexistent post-sale." That's a motion monetizing inertia, not love. The irony is symmetrical with the challenger's: the company whose product exists to shrink the SDR headcount runs a large SDR army of its own, roughly a quarter of its go-to-market org. Both companies sell the end of the sales rep while employing rooms full of them.

This is the section nobody else writes, and with two private companies it runs off the venture narrative instead of an earnings call. The gap is just as wide.

Disconnect 1: "They actually work" vs "the products barely work."

a16z's investment announcement, November 2024, said of Alice and Julian: "most importantly: they actually work. Alice already has a 3x higher response rate than human SDRs." Four months later, the same customers told TechCrunch the AI "failed to generate effective leads" after a month. Either the efficacy evidence cited in diligence was the same CARR-inflated, logo-borrowed customer list, or the diligence missed what buyers were already saying out loud.

Disconnect 2: "79% retention" vs "20-30% retention," in the same window.

11x's official response to the investigation claimed its "retention rate is currently 79%." Sifted reviewed internal Slack messages from the summer of 2024 showing retention "sitting at around 20-30%." Both numbers cannot describe the same company.

Disconnect 3: the warning the lead investor wrote, then ignored.

Eight months before Benchmark led the Series A on the premise of selling "digital workers," Sarah Tavel published this: "I've seen a bunch of companies that talk about selling 'virtual employees'. My gut is this is a limiting mental model. First, the reality is that LLM technology, while seemingly magic, is still early. It still performs best when given a limited scope vs the more general variety of tasks an 'employee' typically handles." The thesis predicted its own failure mode. The check got written anyway.

Disconnect 4: the $4.4B "category leader poised to lead for years" vs the secondary tape.

Outreach's Series G called it "an iconic company poised to be a leader for years to come." The reality since: five layoff rounds, the founder out, no IPO in five years, and secondary shares at roughly 7 cents on the dollar versus what 2021 investors paid.

Disconnect 5: both ends of the barbell now sell the identical robot.

This is the one that matters for the category. 11x's 2024 pitch was "stop licensing tools, hire a digital worker." Here is Outreach's CEO Abhijit Mitra in 2026: "When you bring on Outreach, you're no longer just licensing a tool for your reps. You're hiring agents," and "the unit of value is shifting from seats to skills." The incumbent rebranded from outreach.io to outreach.ai, shipped autonomous agents, and adopted the disruptor's language wholesale. The structural problem is brutal: if the clean-sheet, $76M, AI-native disruptor couldn't keep four of five buyers past three months, the incumbent retrofitting a 2014 stack toward the same claim faces the same product question, plus a per-seat pricing model that punishes it for succeeding.

The fair version of this teardown has to acknowledge that 11x is not dead, the new CEO is saying smart things, and the underlying idea was not crazy. The fragility is specific.

The deepest one is that 11x's data layer is reportedly thin. Multiple reviewers describe Alice as "a wrapper for Amplemarket with rubbish GPT prompts," with stale contacts and sending domains burned inside 90 days. That is the opposite of Outreach's position. Outreach's moat is the first-party workflow data 11x has to rent. An agent is only as good as the signal it acts on, and 11x is acting on cold third-party data it doesn't own.

The second fragility is the one the company already half-admits. The marketing has softened from "replace your sales team" to "won't replace your sales team," and the CEO has recanted the category label. Good. But the pricing and the contract structure still belong to the old story. You cannot sell "outcomes" on a demo-gated annual contract with a CARR-counting break clause and expect anyone burned by the first version to trust the second.

The third is a credibility tax the whole category now pays. Jain is right that "most people don't care if it's human or not, they're trying to solve a problem." But after the Theranos comparisons, "trust us, it works now" is not a go-to-market. Proof is.

Three moves. Each one chained to a specific finding, because a plan without a trigger is a horoscope.

1. Kill the demo-gate and the break-clause, ship a transparent self-serve tier metered on surviving outcomes.

Trigger: 11x is fully demo-gated on a one-year contract with a three-month break clause, the exact structure that produced the $14M-vs-$3M gap and the Theranos coverage.

Why it moves the needle: pricing on meetings that actually book, not contracts that get signed, rebuilds the trust the controversy burned and finally aligns revenue with the only number a buyer cares about.

2. Publish a third-party-audited retention and meeting-yield benchmark, broken out by segment.

Trigger: the "79% retention" claim collided with internal "20-30%" Slack data, and reviews cluster on "zero meetings booked."

Why it moves the needle: the CEO already concedes AI SDRs "don't work in the current form." An audited number is the only credible way to say what does work, and for whom. It also turns the category's biggest liability (nobody believes the metrics) into 11x's differentiator.

3. Re-scope Alice from "replace the SDR" to "the senior rep's force-multiplier," and retire the human-replacement marketing the company is already walking back.

Trigger: 11x hires human AEs whose tool list is Outreach and Salesforce, the blog has softened to "won't replace your team," and the CEO recanted the category.

Why it moves the needle: the honest, defensible wedge is augmentation for teams that already run outbound, the exact segment the 70-80% churn proves it was over-promising to. Sell to the rep who wants leverage, not to the CFO who wants to fire the rep.

The AI SDR story was never really about whether an agent can write an email. It can. It was about whether "replace the human" is a go-to-market or just a headline. 11x proved that a perfect venture narrative and a product that churns four of five buyers is a press cycle, not a business. Outreach is proving that owning the workflow and the data is worth more than owning the buzzword, even when the buzzword is yours to lose.

Here's the transferable lesson, and it cuts against the manifesto. The durable advantage in AI go-to-market is not the agent. Everyone will have the agent. The advantage is the proprietary signal the agent acts on and the workflow it lives inside, which is exactly why the incumbent with a decade of first-party data and the challenger renting cold lists ended up closer on the scorecard than anyone predicted in 2024. Sell the work, sure. But you have to actually do the work first.

11x invented the AI SDR. Its CEO says the AI SDR doesn't work. Believe him, and build for the version that does.