A restaurant kitchen and a food cart both sell dinner, but only one of them needs a liquor license, a lease, and a line of credit before it serves the first plate. For most of the last decade, AI startups were restaurants: expensive build-outs, investors on the hook, a long wait before anyone tasted anything. Agent startups are looking more like food carts. Small footprint, fast to open, and the money you need to start is closer to a good laptop than a Series A.
Which is why a number that would normally get scrolled past caught my attention. In 2026, Bangladeshi brothers Sabik and Shafi Sultan raised more than $30,000 for AgenticHire, their AI-based hiring startup. The money came from accelerator programmes, competitions, grants, and institutional support rather than a venture round.
Thirty thousand dollars is not a headline figure by Silicon Valley standards. It is roughly what a single mid-tier enterprise software contract costs. But I spend my days looking at AI agents that actually do work for people, and the interesting part of this story is not the size of the check. It is the shape of it.
Non-dilutive money is a strategy, not a consolation prize
Grant and competition funding tends to get framed as what you settle for when real investors say no. That framing is getting stale. Accelerators, grants, and institutional programmes hand you capital plus a deadline plus a room full of people asking uncomfortable questions about your customer. What they do not take is a slice of your company.
For an agent product, that trade is often the right one. The cost curve has changed. You are not training a model from scratch. You are composing one — orchestrating existing models, wiring up tools, handling state, and doing the unglamorous work of making the thing reliable enough that a hiring manager trusts it with a candidate pipeline. That work is mostly time and judgment, not infrastructure spend.
Compare it to Octolane, another startup with Bangladeshi founders, which raised $2.6 million in seed funding for a self-driving AI CRM out of San Francisco. Both are agent companies. Both are attacking a workflow that people currently do by hand. The capital gap between them is nearly two orders of magnitude, and yet both are plausible businesses. That spread tells you something about how wide the entry ramp has gotten.
Why hiring is a good place to point an agent
AgenticHire is working on AI-based hiring solutions, and of all the workflows an early team could pick, recruiting is one of the more sensible ones. Hiring is a pile of repetitive, high-volume, low-joy tasks stacked on top of a handful of genuinely human decisions:
- Reading and sorting applications that mostly say the same thing
- Scheduling across time zones and calendars that refuse to cooperate
- Chasing candidates and interviewers for updates
- Writing the same rejection and follow-up notes over and over
- Keeping records straight enough to survive an audit
That first layer is agent-shaped. Clear inputs, clear outputs, tolerance for a human check at the end. The second layer — who to actually hire — is not, and any hiring product that pretends otherwise is asking for trouble. The line between those layers is where these products will be judged.
The part that deserves scrutiny
Hiring agents inherit hiring’s oldest problem. Automated screening reflects whatever patterns it learned from, and those patterns are not neutral. Any team building here has to treat fairness auditing and human review as product requirements, not a compliance checkbox added before a customer’s legal team asks. I have no information about how AgenticHire handles this, and I am not going to guess. But it is the question I would ask first, and any buyer should ask it too.
What I take from a $30,000 round
The Sultan brothers did not raise enough to hire a team or buy a market. They raised enough to build, ship, and find out whether anyone wants it. For an agent product, that might be sufficient to reach the only milestone that matters early on: a real user who would be annoyed if you turned the thing off.
What I would watch next is not the next funding announcement. It is whether the product handles the boring failure cases — the malformed résumé, the candidate who replies at 2am, the interviewer who reschedules four times. Agent demos are easy. Agent reliability is the whole job.
Two brothers, one workflow, thirty thousand dollars, and no investors on the cap table. Somewhere between that and a $2.6 million seed round is where a lot of the useful agent software of the next few years is going to come from. The food carts are open. Now we find out who can cook.
🕒 Published: