Flint is the first self-improving system for performance marketing. The vision is a machine any business can put a dollar into and get ten dollars out. We start with paid search on Google, and the core product is the Google Ads agent.
The agent has two parts. The first is a natural-language interface over the Google Ads API, built with Google's approval so accounts stay in good standing. Anything you can do in the Google Ads UI, you can do by asking: "what's my lowest-converting ad right now?", "which campaigns had 1,500+ clicks?", "update these three campaigns." Customers increasingly use it instead of the Google Ads UI. The second part is a recommendations engine: actions the agent wants to take on your behalf, like adding new keywords it researched, filing negative keywords, improving quality scores, editing ad copy, and running A/B tests until conversion rates hold.
Flint’s landing-page agent serves that loop: every campaign gets its own corresponding landing page, so spend never dead-ends on a generic page. Buy the right keywords, ship the matching page, measure conversions, feed what worked back into the next iteration. That loop is the self-improving part. (One mix-up worth clearing early: Flint is not an AI search or SEO company.)
Sales-led. Demand is mostly inbound. Michelle posts a lot on LinkedIn, the brand is known, and our first AE has been overwhelmed with calls; we signed a second salesperson in August 2026. We qualify hard on ad spend and disqualify fast, so sales time goes to the companies where Flint can make the biggest difference. We're also testing conferences as a channel for the first time.
The heuristic: high ad spend, low engineering density. The more engineers a company has relative to its ad budget, the more likely it builds something in-house and churns. The fewer it has, the more Flint changes how they operate.
That points at consumer brands, which we began selling to in July 2026. These teams have no engineers, no one checking Google Ads every day, and production work scattered across four different agencies. When we show them one system that lets them do the job without managing all of that, the reaction has been "where can I sign?" We still serve the B2B SaaS companies we grew up with, and agencies are a second vector: they always need more clients and better margins, and they reach businesses that would never find Flint directly. The buyer is the growth or marketing lead, or the founder.
The real incumbent in the very long run is agencies. Most Fortune 500 ad spend runs through them, and smaller companies default to them because a bad in-house performance hire is costly and good ones are rare. The in-house alternative is that rare great performance marketer, plus Google's own recommendations, which always steer you toward spending more with Google. On the landing-page side we still displace builders like Unbounce, Leadpages, and Webflow, and we tend to win those head-to-head. What nobody else has is the full loop in one system: campaign execution, the landing page for every campaign, and measurement of what the changes did to conversions. Long term we'd rather make agencies more effective than fight them; they handle the last mile (the "why isn't our ad showing?" investigations) while Flint does the work underneath.
Yes, and it keeps getting bigger. Version one was human-in-the-loop: a marketer asks, Flint generates a page. As models improved, the strategy shifted to building a system and a data moat around the customer, deciding on their behalf what to build next. The latest step is that the harness became the product: the Google Ads agent runs the channel itself, and pages are one of the things it ships along the way. Two things stayed constant. Customers connecting their ad accounts is critical, because that feedback loop is what compounds. And we care less about being the one that generates the code than about owning the system that decides what to build, ship, and change. Same mission, bigger surface area.
A fair risk we actively think about. The ad platforms keep encroaching on adjacent territory, and one of them may ship campaign automation of their own. Our bet is that whoever builds it can't be neutral: Google will always steer you toward more Google spend, Meta toward Meta, and neither can advise reallocating budget across channels. Marketers will want an independent layer that sits above the ad platforms rather than inside one, the same way engineering teams resist locking into a single model provider.
Intervention-keyed, advertiser-aligned outcome data.
Volume is not the claim. Google has orders of magnitude more ad-performance data than Flint ever will. What Flint accumulates is different in kind: not just what a campaign did, but what change we made, why we made it, what we chose against, and what happened next. That deliberation happens inside our system, so the unit of our dataset is a decision and its consequence, which is exactly what you need to make the next decision better. Google sees outcomes; it never sees the reasoning that produced them.
The data is also advertiser-aligned, and that's the structural point. Google's incentive is for you to spend more; ours is for you to spend well. A Google-native optimizer can never credibly optimize against Google's own revenue, and that conflict is permanent. It's the reason an independent optimization layer has a right to exist at all.
The most defensible piece is cross-account learning. Any single customer's history, they could take with them. But Flint sees patterns across hundreds of advertisers that nobody can see from inside their own account and no new entrant can reproduce on day one. Every customer cycle deepens that dataset, the dataset sharpens the product's defaults, and better defaults win more customers. Rising model capability makes the loop cheaper to run and more effective, which is the test we hold ourselves to: the thing getting cheap should be fuel, not foundation.