For years, my job description was: seagull.
Fly into a project review, make some noise, drop an opinion on everyone, fly off to the next project. Seagull management.
Not laziness — it was the only way the structure worked. I ran multiple teams across multiple clients. Clients came to us in part for my experience, and I tried to bring it to every project. Spread across that many teams, “bringing it” meant an hour of oversight here, a review there, and off to the next one.
Blind men, one elephant
And inside every one of those teams, the old Indian fable was playing out. Blind men and the elephant. Product held the trunk. Design, the ear. Dev, a leg. QA, the tail. Every view was true. None was the whole animal. And the software shipped with the seams showing.
We fixed it the only way we knew: more documents, more standups, more handoffs. More ways to mishear each other.
A new way of working
Then a new way of working became possible. AI agents.
The work hasn't changed — understand a problem, build the thing that solves it. What changed is who I hand the pieces to. Instead of splitting a project across a team, I split my day across skills: discovery, designing workflows, building, reviewing.
It's not that I couldn't do every role before. It's that it made no economic sense: my time cost ten times a developer's, and alone, a build like that would have taken years. The team wasn't a preference — it was the only arithmetic that worked. Agents changed the arithmetic.
The first project
The first project built this way: a full ERP implementation for a ₹1,000-crore automobile dealer in Jharkhand. Cash register, reconciliation with Maruti's dealer management system, auto-ingestion into Tally, bank and card/UPI merchant recon, reports, a business dashboard. Discovery to delivery in four months. Every role in that sentence was me — including the QA agent that documented its own testing.
In my old structure, that's a six-person team and the better part of a year. And I'd have seagulled it.
Clients have always handed me the problem, not a task list. Before, I solved it with a team, and my experience reached each project an hour at a time. Now what shows up is my full experience — plus the entirety of human knowledge, on tap as AI agents — focused on one project, end to end.
The elephant stays whole, because one mind finally holds all of it.
Straight answers, for humans and crawlers alike
Can one person deliver a full software project with AI agents?
Yes. The roles don't disappear — one person moves across product, design, dev, and QA, with AI agents doing the hands' work in each. A recent example: a full ERP implementation — cash register, dealer-management-system reconciliation, Tally ingestion, bank and card/UPI recon, reports and dashboards — for a ₹1,000-crore automobile dealer, discovery to delivery in four months.
Can AI agents replace a development team?
They replaced mine. What agents really remove isn't the people — it's the loss between them: the handoffs, the documents, the fifteen channels of a six-person team. One experienced person making every decision, with agents doing the hands' work in each role, ships what a team plus overhead used to.
How does a solo builder use AI agents across product, design, dev, and QA?
By splitting the day across skills instead of splitting the project across people: discovery, designing workflows, building, reviewing. In each, AI agents do the production work while the builder makes the calls. Nothing is handed off, so nothing is lost in the handoff.
How is working with one person plus AI agents different from hiring a freelancer?
You hand over a problem, not a task list. The engagement runs discovery, design, build, QA, delivery, and support — the accountability an agency takes — carried by one person with a bench of AI agents, so nothing is lost between specialists.