A while back I wrote about a LinkedIn post I didn't write — planned, written, branded and published by an AI marketing agent we built on our own platform. It was a good story. But a good story about automation means nothing until you come back a quarter later and answer the only question that matters: is it still running, and did anything actually change?
It is. And it did. Here is the honest version, in percentages — because for a small team the shape of the curve matters far more than the size of the number.
What a quarter of marketing on autopilot changed
Since the agent took over the calendar, our publishing output is up 390% — close to five times the volume, across LinkedIn, Facebook and a third channel that came online mid-quarter. Not one of those posts was written, scheduled or chased by a human. The agent plans the slot, writes from our own knowledge base, renders the branded image, runs it through a quality gate, and publishes.
In the same window, sign-ups grew 72% and paying customers grew 45%. People are not just arriving — they are staying, building, and reaching for a card.
We did not get more disciplined about marketing. We built something that is disciplined for us — and then went back to building the product.
What I am not going to claim
I could tell you the agent caused that growth. I do not know that, and neither does anyone selling you a similar story. The agent ran through the quarter; the product also got better, users told other users, and our customer stories started doing their own work. What I can say with certainty is narrower and, I think, more useful: showing up consistently stopped being something we had to remember. Everything else got a chance to compound on top of that.
I am also not going to quote reach or impressions. We do not have reliable audience data from those channels, so any number I gave you would be decoration. Publishing is what we measured, so publishing is what I am reporting.
Then the government backed us
Partway through the quarter, NeoBuild was selected onto a government-supported accelerator — three months of intensive work on funding readiness, followed by 15 months of continued startup support. For a bootstrapped team in Chennai that has spent most of its energy on the product, being chosen by a programme with public backing is a different kind of validation from a customer signing up. It is people whose job is to look hard at whether this becomes a company.
An accelerator place is not revenue and it is not product-market fit. What it buys is the one thing small teams are worst at giving themselves: structured pressure. Three months of being asked the questions we have been too busy to answer — and then fifteen months of support through the stretch where most startups quietly drift.
Why these belong in one story
Because they are the same argument seen from two sides. The agent is proof that we run our business on the thing we sell: our go-to-market is an application built with NeoBuild, using the same describe-the-workflow method our customers use for approvals, inventory, field service and client portals. The accelerator is a signal from outside that the approach survives scrutiny.
Customers paying is the signal that matters most. The other two make it repeatable.
What we are doing with the next quarter
- Funding readiness, done properly. Metrics, unit economics, and the story that connects them — exactly what a three-month intensive is good for.
- Measuring outcomes, not output. A 390% jump in publishing is a volume number. The next honest post needs a number about what that publishing produced.
- More channels, only if they earn it. The agent already drafts for channels we have not switched on. Writing them was never the hard part; deciding they deserve attention is.
- Staying in the product. The agent exists so that marketing is never again the reason we stop building.
If you are the small team in this story
The lesson keeps arriving in the same unglamorous shape: the work your team keeps dropping is not a discipline problem, it is a piece of software you have not built yet. Ours was marketing. Yours might be approvals, onboarding, scheduling, or the report someone rebuilds by hand every month.
You do not need a bigger team to fix it. You need the workflow described clearly once — see what NeoBuild generates from a single description, and what it costs to find out.
The workflow your team keeps dropping is the one worth building.
Describe it in one sentence at Neobuild.dev and watch a working application take shape — the same way we built the agent that ran our marketing while we stayed in the code.