I've been quiet on the YC blog because we've been heads-down with our first wave of customers. But I want to share the thesis we're building on, because I think it matters for any founder watching the AI-meets-enterprise wave.
I'll keep it tight.
The thesis in one paragraph Salesforce has more than 150,000 customers. It runs the revenue operations of a meaningful slice of the global economy. And the actual day-to-day experience of using Salesforce is, by every honest account, miserable. AI doesn't make Salesforce incrementally better. AI replaces the entire human-and-spreadsheet layer that's wrapped around Salesforce for the last fifteen years. The category that delivers that replacement is, in my honest opinion, one of the largest still-uncaptured surfaces in enterprise software.
Why nobody's built this yet The honest answer: until very recently, you couldn't.
The "AI for CRM" wave that hit between 2018 and 2023 produced a lot of glorified text-summarizers and a couple of decent forecasting tools. None of it was the actual layer — the system that sits between humans and the database and does meaningful work without supervision.
Three things changed in the last eighteen months that make the actual layer possible. Model reliability crossed a threshold where you can hand off real workflows, not just suggestions. Function-calling matured enough that an agent can interact with structured systems without falling apart. And — quietly — a new architectural pattern emerged for letting LLMs operate on systems of record without corrupting them. (We wrote about that pattern on DEV; I won't rehash here.)
What this means in practice: an AI layer for Salesforce can now keep account hygiene clean autonomously. It can write meeting notes back to the correct opportunity with the correct context. It can reconcile pipeline against signal. It can tell a rep which deal to work on right now by reading the same data their manager is reading and reasoning about it the same way.
Why this is a first-mover game (and why I keep saying this on every call)
Most enterprise AI companies build an AI feature and try to sell it on top of the existing stack. We took the opposite bet.
That structural choice matters because every AI layer compounds with usage. The orgs that adopt early train the system on their actual workflows. Their data shapes the product. Their feedback shapes the roadmap.
For any founder reading this whose company runs on Salesforce: the question isn't whether you'll have an AI layer in the next two years. The question is whether you'll be on the inside of someone else's adoption curve, paying list price after the design-partner cohort is closed, or whether you'll be one of the teams whose go-to-market motion is in the foundation of the system everyone else ends up using.
I don't think that framing is hyperbole. I think it's how most platform shifts have actually played out. The orgs that adopted Salesforce in 2003 won the next decade of revenue operations against orgs that waited until 2008. The orgs that took cloud seriously in 2010 ate the lunch of orgs that didn't until 2015. AI in the revenue stack is the same shape of bet, on a faster clock.
What we're looking for
If you're running on Salesforce and you've been waiting for the AI layer that actually does the work — not summarizes it, not gestures at it, does it — we should talk.
(Aside, because the same question keeps coming up on every intro call: yes, we're being aggressive on pricing for the first wave of design partners. No, it's not on the website. Email me before that changes — taylor@revkit.ai.)
About the author
Taylor is the founder of revkit.ai, the AI layer for revenue teams running on Salesforce. Before revkit, Taylor spent ten years leading RevOps and GTM inside B2B SaaS companies — more than a dozen of them, collectively over $250M in ARR. Find Taylor at taylor@revkit.ai or Linkedin @TGreen7163.
1 comments