90 Days to $1.5M ARR: A Company Selling a Problem That Doesn't Exist Yet
Deep thoughts on AI and aspiration —— ByteThink Circle
A company called Gushwork recently raised a $9 million seed round led by Susquehanna, with Lightspeed and B Capital participating. Here are the numbers: 90 days to $1.5 million in annual recurring revenue, 50% to 80% monthly growth, over 300 paying customers, generating more than 1,000 sales leads per month for clients. Average customer subscription runs $800 to $900 monthly, 95% are in the United States, mainly the most traditional B2B businesses—industrial distributors, contract manufacturers, professional services firms.
What they’re selling is simple to describe: getting these companies to appear in AI search results.
Buyers Have Moved, Sellers Are Still at the Old Address
The starting point for B2B buyer supplier research is shifting from Google to ChatGPT, Claude, and Perplexity. And the questions they ask are highly specific: technical specifications for particular use cases, industry-specific compliance requirements, side-by-side comparisons of multiple vendors. All of this happens before filling out contact forms or scheduling sales calls. By the time you receive an inquiry, their research is already complete.
The problem is on the other side. According to Gushwork, 90% of companies are completely invisible in these AI answers. When a potential customer asks “what are the best precision bearing industrial distributors in the Midwest,” you’re probably not in that answer, not even in the conversation. This isn’t about ranking higher or lower—it’s about existence versus non-existence. The data confirms the scale of migration: OpenAI disclosed in July 2025 that ChatGPT receives approximately 2.5 billion prompts daily.
My assessment is that the most valuable characteristic of this market right now is that buyers and sellers are not on the same timeline. Buyers have migrated, sellers haven’t awakened yet. This kind of misalignment has happened once before in history: in the early days of search engines, the first companies to do optimization captured a decade-plus dividend. By the time everyone acknowledged SEO was important, the dividend had been distributed. The length of the window depends on how quickly the majority recognizes the problem, and recognition is precisely the slowest step.
A side note on data reported by the shovel seller: according to Gushwork’s statistics, traffic from AI search accounts for only 20% of their clients’ total website traffic, yet contributes 40% of lead conversions. Logically this makes sense—users who arrive via AI search have already completed deep research, have stronger purchase intent, and shorter decision cycles. But this is the company’s own data. How much discount to apply—I’ll get to that.
Not Selling Tools, Packaging an Agency Into Software
Gushwork’s product is called AI Feeds, which can be broken into two parts.
One part is seven specialized AI agents: Memory, which maintains company profiles ensuring information consistency; Research, which scans thousands of AI search queries to identify content gaps; Strategy, which clusters questions into content blueprints; Content, which generates landing pages and guides; Publishing, which handles distribution; Backlinking, which acquires backlinks from a network of partner websites; and Auto-Update, which automatically refreshes content as models change. The other part is a content library mounted under the client’s own domain, specifically architected for AI crawler ingestion habits. Plus a backlink network containing approximately 200 to 300 partner websites, from which each client receives 10 to 20 backlinks.
From this system, I draw two judgments.
First: the real asset of companies like this is not agent orchestration. Orchestration capability will commoditize—model providers could turn similar functionality into APIs at any time. What’s genuinely hard to replicate are two things: the backlink network requires years of accumulation and can’t be bought with an API; the operational intuition for “what content gets cited by AI” has to be developed through working with client after client. The standard for identifying which parts are moats is simple: which parts can’t model companies provide—those are the ones that matter.
Second: the very existence of the Auto-Update component acknowledges the instability of the game rules. Once enough companies batch-optimize content, AI engines will inevitably tighten citation and ranking rules. Methods effective today will likely fail in two years. So when enterprises buy this kind of service, they’re essentially buying a subscription for continuous adaptation, not buying a one-time optimization project. Anyone who treats it as a project will discover they’ve bought an expired can when rules change.
Metrics Anchored on “Leads,” Not Traffic
Every page generated by AI Feeds includes tracking and lead capture. Whether prospects arrive from Google or ChatGPT, as long as they take action on the content, leads flow directly into the client’s inbox and dashboard. Which keywords, which content, which channels are driving the sales pipeline—the data uses a unified framework.
This design deserves attention from everyone doing B2B services. Half the arguments in the marketing industry stem from inconsistent metric frameworks: when you report impressions, rankings, or traffic, buyers and sellers always talk past each other, and three teams fight over credit for the same conversion. Align metrics to the layer the client can directly convert to money, and negotiations become immediately simpler. This wave of AI service companies generally needs to learn this lesson—those who learn it first get the orders first.
Three Places Requiring Discount
This company’s story has three gaps that investors have questioned, and service buyers should question even more.
| Question | Specifics |
|---|---|
| All data is company-sourced | 20% traffic contributing 40% leads, a client closing $200K-$350K contracts—all from company statements, no third-party verification |
| Education costs underestimated | 90% invisibility corresponds to 90% lack of awareness. Scaling from 300 to several thousand customers requires explaining “you don’t exist in AI” to people who don’t acknowledge the problem exists. This kind of sales cost curve is often steeper than expected |
| Platform risk | All optimization methods rely on AI engine ranking mechanisms. When mechanisms change, promised results get entirely reshuffled |
The first gap means data direction is credible, magnitude is questionable. The second gap determines the ceiling of this business: self-aware customers are always a small fraction. The rest have to be educated one by one through sales, and educational sales is the most expensive and hardest to scale. The third gap no one can avoid—you can only hedge it through continuous adaptation, which is also the real bet of their entire product form.
An Action You Can Take This Week
Whether you buy this kind of service or not, there’s one zero-cost action worth doing immediately: take the ten most core buyer questions in your business, ask ChatGPT, Claude, and Perplexity, and see if you’re in the answers.
If you are, study why, solidify the reasons into a content strategy, and protect that position. If you’re not, congratulations—at least you know where your customer acquisition budget should go next. The judgment logic needn’t be complex: wherever your customers make decisions, that’s where your visibility should be built. And right now, that location is migrating in batches.
As for entrepreneurs looking for direction, this kind of misalignment—“buyers have migrated, sellers haven’t awakened”—is worth pursuing more than any single-point technology. Because it comes with a countdown timer. The day sellers collectively awaken is when first-mover advantage zeroes out. The countdown is the true source of first-mover advantage.