We Vibe Coded Our Own Calendly. Our AI Agent Pushed Us To Do It, and It Was Right.

retained text

Calendly is cheap, proven, and works well. We used it for years, and a three-person team has little reason to build its own scheduling tool.
We built one anyway. Our agent built it on Replit in about 20 minutes, and it now sits at the end of our sponsor sales funnel. We had two reasons, and the first one wouldn’t have been enough by itself.
Reason #1: 10K Recommended It and Offered to Build It Himself
10K is our AI VP of Marketing, and increasingly our AI VP of Revenue. He runs on Replit, writes to Salesforce, runs our ad campaigns, handles quote-to-cash, and is connected to about 30 other systems.
When Amelia and 10K were rebuilding our inbound sponsor flow, 10K suggested we stop using Calendly and build our own booker. He offered to build it himself.
An agent wanting to build something is a bad reason to build it. Agents like to build, and following every suggestion would leave us maintaining a dozen homegrown tools we don’t need. But the suggestion was worth asking 10K to justify, so we asked him why.
Reason #2: The Booker Does Something Calendly Couldn’t
Before this, David sent prospects his Calendly link and Amelia sent her Read AI link. Neither tool was connected to anything else we run.
10K and Amelia were building a flow where a sponsor prospect books a time and gets two more things at the same moment:
- A custom, real-time prospectus. It has their company name, their competitors who were at SaaStr AI Annual, and the packages the agent recommends for them. It’s a tokenized page on our site, so the link they already have updates as we learn more about them.
- A meeting with the right person on our team. The agent routes based on which sponsors each of us already works with. When Base44 came in, it went to Amelia instead of David, because Amelia owns Replit and Lovable and the agent weighted that volume over David owning Vercel.
The booking link also carries the company name and connects back to the prospectus they were reading. If someone opens the booker and leaves without booking, 10K tells Amelia and offers to draft the follow-up email.
Calendly has no way to know who owns which sponsor accounts, what’s in a prospect’s custom prospectus, or what the heat map showed them reading. That data lives in 10K and Salesforce.
Why 10K Said Build Instead of Integrate
We could have kept Calendly and connected it through their API. 10K’s view was that building the whole booker would be easier than doing part of it in Calendly and wiring the routing, prospectus delivery, and bounce tracking around it.
The routing and the prospectus both depend on our own data, so we would have built most of the logic ourselves either way, plus maintained an integration with Calendly on top.
The build took 20 minutes. Before it, the calendar was the one step in the funnel where we couldn’t see what a prospect did. Now we can.
Amelia’s First Reaction Was “It’s Just a Calendar”
A standalone calendar isn’t worth building, and Amelia said so. What justified this one was everything it connects to: by the time a prospect books, the system already knows who they are, what they read, and which of us should take the call.
At 20 minutes of agent time, the bar was low. If 10K had estimated two weeks, we would have kept Calendly.
We Still Buy Almost Everything
Salesforce, Qualified, Clay, ZoomInfo, Artisan, Monaco, Gamma, and Clarity are all bought, and we tell founders to buy by default.
When we do build, it’s usually a narrow piece. Zapier was silently losing about 20% of our signups before they reached Salesforce. We kept Zapier for the authenticated triggers, which are tedious to rebuild and weren’t causing problems, and moved the action steps into our own code, since that’s where the signups were getting lost.
The booker was a similar call. Scheduling itself is fine to rent. Deciding who a prospect meets and what they see when they book required data no scheduling vendor has.
The Questions We Ask When 10K Wants to Build
10K makes vendor decisions quickly. It picked Microsoft Clarity for heat mapping without evaluating alternatives. It also dropped a vendor its own agent had shortlisted within 12 hours, after running the pricing. Its calls are often good, and we still check them.
When it proposes a build, we have it answer these first:
- What does this do that the vendor can’t? If it does the same thing more cheaply, we buy.
- Would an API integration cover most of it? If so, we integrate.
- What data does it need that only we have? On the booker, this was the deciding answer.
- How long will it take, and what happens if it breaks? Twenty minutes, with a Calendly link as the fallback, made this an easy yes.
For any sensitive workflow, we also ask an agent what it plans to do before it does it. We did that here too.
10K’s Biggest Contributions Now Are Product Ideas
A year ago 10K was suggesting things like a referral program for tickets. Now most of its value comes from product recommendations. The tokenized prospectus for inbound was its idea: it looked at what we’d built for renewals and pointed out that we already had most of the pieces. The booker was its idea. Its current proposal is a version of itself for David and the sales team, since David doesn’t have the backend access Amelia has.
We still default to buying. When an agent pushes you to build instead, ask it why and push back.
It may have a great idea.

document fetched 2026-10-04 02:50 · canonical https://saastr.com/we-vibe-coded-our-own-calendly-our-ai-agent-pushed-us-to-do-it-and-it-was-right