Spoiler: Groove is still running. We stopped signing up new customers and building new things on it, and everything new went into Helply. If you use Groove today, this is the story behind that decision.
“The only way Groove can overcome the NRR zombie phase is by dramatically changing our GTM strategy and rebooting the product and brand.”
I wrote that sentence at my office in Newport, Rhode Island, on May 29, 2024.
I reread it probably ten times before I sent it to a business partner, because I knew that once it was written down, I couldn’t pretend anymore.
Most founders will tell you not to admit this kind of thing publicly. That it makes you look weak, or that customers will get nervous, or that competitors will smell blood. But I’ve always shared the uncomfortable stuff, and this is about as uncomfortable as it gets for me.
Everything I’ve got
I started Groove at 29, in my parents’ basement.
By 2024 it was doing about $5M a year, profitable every single year, and we never raised a dime. Groove has been everything to me for the last 15 years. It’s becoming Helply now, and Helply is everything to me too.
It’s everything I’ve got.
And it was stuck.
Missing our mark
On paper, things looked fine. We had 92% net revenue retention and 85% logo retention, and customers who loved us stayed for years.
The only problem?
Nobody new was walking in the door. We were adding about 22 new customers a month, every month, and no matter what we tried, that number wouldn’t move. I kept pulling up Stripe over and over, and every time it showed the same flat line.
There’s a name for this: an NRR zombie. Your existing customers keep paying for years, but you can’t win new ones. You’re not dead…you’re just not alive either.
I hated that phrase the first time I heard it, mostly because it was accurate.
Then AI made it worse
AI shrinks support teams, and when your customers need fewer people, they need fewer seats.
We charged per seat.
In 2025, seat downgrades made up 59% of the revenue Groove lost, and they happened three times as often as cancellations. Our customers weren’t leaving. They were shrinking, and we charged for the exact thing that was shrinking.
On top of all that, every year since ChatGPT came out, downgrades took out more revenue than new customers brought in. That’s a hell of a thing to watch in your own dashboard.
Trying to save it
So we did what most founders would do. We tried to fix it without starting over.
In June 2024 we bolted AI onto Groove: drafts, rewrites, summaries, tagging. It made people faster at clicking buttons, but it didn’t change the actual work.
In January 2025 we launched Helply as a separate AI agent that plugged into Groove or any other help desk. It got to 252 customers, mostly our own, each paying 1.3x what they paid before. Then it hit its own ceiling. (More on that one in two weeks.)
In February 2025 we bought an AI knowledge base company. Great team, great product…and now we were building three products at once.
We were scrambling. And when you’re scrambling, you make irrational decisions.
We made a few.
The decision
Ultimately, I stopped trying to save it.
In September 2025 we decided to start over, and in March 2026 we closed signups on Groove at $4.7M ARR, with over 1,400 customers. Groove’s cash flow paid for the entire rebuild. No investors, no bridge round, nobody to ask for permission.
That’s the part I’m proudest of. The old product funded the new one.
Looking back, this wasn’t a hard decision because the answer was unclear. It was hard because I didn’t want it to be true. We hedged for over a year when we should have committed.
Takeaway: hedge until you’ve proven the bet, then go all in.
What surprised us
The model doesn’t matter much. Everyone has access to the same ones. What nobody else has is your customers’ tickets, docs and history, so we built the data layer first.
Pricing AI is a pain in the ass. Seats plus outcomes meant two bills, and outcomes alone meant arguing about what “resolved” means. We landed on credits for the work the AI does, simple enough for a customer to budget for.
And we ran our own support on it before we sold it to anyone. Our median first response went from 18 minutes to under 6, and support went from six people to two. Nobody lost their job; the other four moved into sales and success.
If you run a SaaS company
Before you bolt AI onto your product, ask yourself:
- Does your revenue go down when your customer’s team gets smaller?
- Can AI finish the work, or does a person still have to push it forward?
- Do you measure success by how many people use your software, or by how much work it gets done?
- If you were starting from scratch today, would you build the same product?
The fourth one is the hard one. It took me a year to answer it honestly.
If your answer is no, you already know what to do. I just hope you do it faster than we did.
One question to take with you: if you were starting from scratch today, would you build the same product?
Next Thursday: how much we spent pitching a product that didn’t exist yet. (It’s a lot.)
See you next Thursday,
Alex