![[Un]Churned by Gainsight](https://substackcdn.com/image/fetch/$s_!7AoO!,w_120,h_120,c_fill,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffe2167ac-0bcf-4575-9712-8d5ef3588851_300x300.png)
![[Un]Churned by Gainsight](https://substackcdn.com/image/fetch/$s_!hKlf!,e_trim:10:white/e_trim:10:transparent/h_72,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe14b36dd-52b9-48a3-9f93-3f6a459d55ff_1344x256.png)
The Best AI Product Is the One Only You Can Build
A builder's POV on going from zero code to deployed agent in a month, and why the thing worth building is already inside your company.
![[Un]Churned's avatar](https://substackcdn.com/image/fetch/$s_!vkJ0!,w_108,h_108,c_fill,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0464ad30-26c2-4f32-b429-ae4283dd5586_200x200.png)
From the day your SaaS solution launches, it’s already drifting from what your customer actually needs. Sometimes their leaders know it. Sometimes nobody does. But it’s in the data and a strategy shift is coming, and your onboarding cycle wasn’t built for it.
You do your best, you train the customer admins, you train your own CSMs, but eventually the instance drifts. Sometimes it’s minor stuff: tech debt, automations running in the dark and going nowhere. Sometimes it’s major: the exec sponsor left and you’re unmoored. How do you get it back? Do you even know how big a problem it is for you?
I Hadn’t Written Code Since 2006
I came to this question from an unusual angle. My background is in GTM, not engineering and definitely not software development. When I stepped into a new role leading Gainsight’s GTM org’s AI strategy, I hadn’t written a line of code since a Java Programming class in 2006. I got to this work not because I was the obvious person to build something, but because I was deep enough in the business to know what was worth building.
When I started the role, we were torn between surveying the org and trying to sort and solve all the problems at once, and just jumping in and solving one problem end-to-end. We’d heard that this was the right approach, but we didn’t know which use case was going to have the biggest bottom-line impact, nor which we could deliver quickly. Luckily I was not the only person trying to solve this. Josh Schachter had just joined Gainsight and had a mandate to do exactly that: find an AI use case for the CS team and solve it.
Asking the Right Questions
We asked our partner, Method Garage,and they simply asked the CSMs. Actual conversations with your internal customers! Embarrassed to say we hadn’t started there. I was focused on scraping use cases out of Slack posts and surveys, but the real answer was in our top CSMs heads. Of course, the problem was our pre-AI answer to the problem I described above: customer instances drift from their goals, and CSMs were logging in and evaluating configurations one feature at a time, taking 8 to 12 human hours to get from there to a reasonable presentation on what was right, what was wrong, and where to start on bringing the instance into alignment with the customer’s goals.
The decision to focus on Instance Reviews as the solvable use case came out of a simple question our partners posed early in the process: if we could build software to help our own CSMs, what would it do? The answer was immediate and unanimous. I joined the project a bit later as the SME on the configurations the tool would need to analyze, and this turned out to be exactly the right entry point.
The Foundation Already Exists
Method Garage proposed a design, and then the crucial question landed: who is managing these internal builds today? Who will build it, host it, run AppSec, authenticate users?
We didn’t know. We’d heard rumors that some teams had built things like this, but no real answer. Then we found it: Anoosh, who leads operations for our Solutions Consultants, had already done the hard work.
In July 2025, well before I started building anything, Anoosh had worked with engineers, developers, and IT across Gainsight to stand up a secure, compliant software stack. He built it originally for Solutions Consultant to generate bespoke data for community demo orgs. He’d also built an RFP response generator that matched dozens of questions to a library of hundreds of answers. In doing so, he’d gotten IT and engineering approval, established real web hosting (internal only), and created something that almost no other team at Gainsight had: infrastructure that was actually ready to build on.
He hadn’t built it for us, but it was there. Without it, our time to value would have stretched into quarters.
If you’re reading this and wondering whether you already have something like this inside your company, my advice: find your Anoosh before you find your use case. Look for the person who has already fought the AppSec and hosting battles on behalf of a different team. That foundation is usually closer than you think, and it’s what lets a non-engineer like me ship production software in 26 days.
Vibe Coding as a Subject Matter Expert (SME)
Josh asked me to drop our careful analysis of use cases and infrastructure and, starting February 1st, build the instance review agent on Anoosh’s stack. I had a Product Requirements Document (PRD) and an existing repository. I kid you not, I hadn’t written a line of code since 2006. I’d been using Claude for about two months, but never Claude Code. I didn’t have the first clue what an IDE was.
From there, I climbed my way up the learning curve. I spent the first week orienting myself by learning enough vocabulary to ask the right questions, making mistakes that Claude Code helped me walk back, figuring out the difference between understanding what needed to be built and being able to describe it clearly enough for a model to build it. That gap is where the SME really makes an impact.
Claude Chat was my PM. Claude Code was my engineer. I was the SME who knew what “correct” looked like.
By February 26th, the Instance Review agent was deployed on production hardware and ready for use with customers. I built, tested, got it vetted by CSMs, and deployed in time to demo at our Revenue Kickoff, in less than a month from the first time I had opened VS Code. We delivered the first instance review from this internally built software to a very kind and open-minded customer’s ops team only three weeks after I started building.
Did I learn React or Python? No. But I shipped production software with both.
If You’re Thinking About Building
Today 27 CSMs have shared instance reviews with more than 50 customers. The effort has paid for itself already in time savings alone.
Our Professional Services team is now building on this; the Instance Review agent is becoming part of how we systematically detect and close that drift before it becomes a renewal conversation. You can’t buy this as a product. You could hire someone to build it, but they’d need your SMEs in the room anyway. When the expertise already lives inside your company, AI closes the gap between knowing what to build and being able to build it.
If you’re an SME thinking about taking the same leap, here’s what I’d tell you:
Find the infrastructure first. Don’t start with the use case. Start by asking whether someone in your org has already done the AppSec, hosting, and IT approval work. That stack is your launchpad.
The SME work is the real work. I didn’t need to become an engineer. I needed to be specific enough about what “good” looked like that Claude Code could build toward it. Your domain expertise is the input that makes the output worth anything.
Ship something, then improve it. I had a working agent in three weeks. While it wasn’t perfect, feedback from real CSMs on real customer instances is what made the next version better. You won’t get that from a PRD.
What did you decide to build instead of buying? Or anything you regretted building?
Steven Doty is a Senior Program Manager of Gainsight Systems at Gainsight, where he leads AI strategy and builds the tools his GTM team actually uses. He came to this work from ops, not engineering: eleven years at Rosetta Stone, where he was one of Gainsight’s earliest customers, owned the company’s NRR reporting process, and won employee of the year for it.
Kendra McClanahan, now his manager, did his onboarding session in 2015. He picked up Gainsight that day and hasn’t been able to put it down since. He lives in the Shenandoah Valley of Virginia, where he restores native plant habitat and wages a losing war against deer, rabbits, and his own Claude token usage.