Cobra's founder bets AI agents will capture services spend

1. Customer as first investor — ▶ 1:50

   With no prior job, Fort cold-messaged the CEO of a French company who had written about sales compensation. Over a weekend the CEO and his CFO built a rough comp app, then offered to be Cobra's first investors and first customers. A 100K pre-seed round from seven business angels followed. A second customer came from a LinkedIn demo request when there was almost nothing to demo, just a promise. Fort's main channel was outbound calls from his own phone. Editorial inference: a prospect trying to build the product himself is strong evidence of pain.

2. Per-rep pricing, free admins — ▶ 3:31

   Cobra charges per seat, but seats are the sales reps being paid commissions. Admin licenses for finance and revenue staff are free, so 100 commissioned reps means 100 licenses regardless of back-office headcount. The first 10 customers paid $10,000 a year, set by taking comparable, mostly US, market prices and splitting them in two. ACV today is about $30,000. The host's arithmetic of 250 customers times $30,000 gives roughly $7M in ARR, which Fort said was slightly above the real figure.

3. B2C playbook, weak attribution — ▶ 4:46

   Cobra now invests in trade shows, dinners and ads, applying B2C-style marketing to B2B. At a trade show it took the lowest, silver booth package and ran its ads on a screen as an icebreaker to draw passers-by. It also runs LinkedIn and YouTube ads and credits a meme ad with helping close an 80K-per-year deal. Fort concedes that tying revenue to LinkedIn is very hard because it is top of funnel. He relies on higher-level signals, such as more website traffic after ads run.

4. Skipping the valuation game — ▶ 8:10

   Cobra raised a 100K pre-seed, a 5 million seed in January 2022, and a 10 million Series A at the end of 2023, plus $2 million of debt from HSBC and two other banks. Fort says he declined the era's valuation game, which the host described as 40-50x ARR multiples, because it could become a trap. He believes markets collapsed soon after. The 2023 valuation was below 100 million. He recommends debt alongside a raise but warns it must be repaid and is not invested money.

5. Where AI replaces reps — ▶ 11:39

   Cobra's data shows its top earner took $1.5 million in variable compensation in 2025, about 75x the average, in a dataset Fort described as tens of thousands of reps. He would not say what that person sold. The host cited One Mind's Amanda, who says firms pay $100,000-$400,000 per AI agent versus roughly 300K for a rep hitting a 1 million quota. Fort's view: AI will replace reps for products with very fast, one-call sales cycles, but is far from handling complex sales.

6. Agents that do the job — ▶ 13:20

   Cobra plans to release three agents in June: analyst, architect and coach. The architect builds and deploys comp plans in minutes from a PDF or PowerPoint, versus the weeks typical implementations take because data must be sent and rules built across many plans. Fort frames the shift as moving from SaaS that helps people to AI that does the job. He doubts cannibalization, citing a framing of $6 of services for every $1 of software, which AI turns into software. This is a forecast, not demonstrated revenue.

7. G2 feeds AI answers — ▶ 15:36

   G2, with about 1,000 Cobra reviews, matters less for direct clicks than as a source AI assistants consult when asked for the best sales compensation software. Fort says that drives traffic, but he could not quantify it. Cobra's glossary-style programmatic SEO was intentional, a team member's idea, and took two to three years to show results. He says Cobra has under 1% of the market, draws a significant share of revenue from the US, and is relocating him to New York.

Antoine Fort's central bet is that the next phase of software is not better tools for people but products that perform the work themselves. Cobra, the sales compensation platform he co-founded in 2020, plans to release three agents in June: an analyst, an architect and a coach. Fort describes the shift as turning a SaaS product that helps people do their job into an AI product that does the job. He rejects the obvious objection that this will cannibalize existing subscriptions. His claim is the opposite: agents will let Cobra capture money that customers currently spend on people and consultants, and revenue should rise sharply as a result.

The argument rests on a specific mechanism. Commission software is slow to deploy. Customers must send data, then implement plan rules, and large organizations typically run many plans, each with its own quirks. Fort says a typical implementation takes weeks, and that the architect agent compresses it to minutes: a customer supplies a PDF or PowerPoint of the comp plan and asks the agent to build and deploy it. He frames the opportunity with a ratio he borrows from others, that every $1 spent on software is accompanied by $6 spent on services, and that AI converts the $6 into software. The conversation never tests that ratio, and it describes the size of a pool rather than Cobra's ability to price against it.

Cobra's existing business shows what is being extended. It serves more than 250 customers at roughly $30,000 in ACV, which the host's arithmetic puts at about $7M in ARR. Fort said the real figure is slightly above that. Pricing is per seat, but the seat is a commissioned sales rep, not a software user: admin licenses for finance and revenue operations staff are free. Editorially, that structure has an implication Fort did not address. Agents that remove administrative and implementation labor would not reduce seat counts, which track the reps being paid. But they would not obviously raise them either, so capturing services spend probably requires a new pricing logic, and the transcript does not say what it will be.

The company reached this position with modest capital and a stated refusal to chase peak-market valuations. After a 100K pre-seed round from seven angels, Cobra raised a 5 million seed in January 2022 and a 10 million Series A at the end of 2023, followed by $2 million of debt from HSBC and two other banks. Fort said he declined the valuation game of that era, when the host noted ARR multiples of 40-50x were common, because it could become a trap. He believes markets collapsed soon after. He would not disclose the 2023 valuation but confirmed it was below 100 million. His warning on debt is that it must be repaid and should not be treated as invested money.

Growth came from unglamorous channels used with discipline. Fort began with outbound calls from his own phone and a little network. The first customer was a French company whose CEO had written about sales compensation. He and his CFO built a rough comp app over a weekend, then offered to be Cobra's first investor and first customer. A second customer signed after a demo request, when there was almost nothing to demo beyond a promise. Today the mix includes trade-show booths, LinkedIn and YouTube ads in a B2C style, and G2 reviews, which Fort says matter because AI assistants draw on such sources when recommending software. He concedes attribution is weak and could not quantify G2's contribution.

Cobra's market also depends on a premise the host challenged: that commissioned reps survive AI. Cobra's data shows its top earner took $1.5 million in variable compensation in 2025, about 75x the average, across a dataset Fort described as tens of thousands of reps. The host cited One Mind's Amanda, who says firms pay $100,000-$400,000 per AI agent against roughly 300K for a rep who hits a 1 million quota. Fort's answer was that AI will replace reps for products with fast, one-call sales cycles, but is far from handling complexity. That position fits a company whose customers pay commissions to people. He offered judgment rather than evidence.

Both halves of Cobra's outlook therefore rest on assertion rather than data shown on air: that complex sales will keep humans in commissioned roles, and that agents will pull services budgets into software. If the first holds, the seat base stays intact. If the second holds, Cobra must find a way to charge for completed work rather than access. Fort aims to pass 10 million in ARR this year. The host's figures put the company at roughly 7-8 million, up from between 4 and 5 a year earlier, presumably in millions. Whether the June agents change that curve, or merely shorten implementations, is the number to watch.