Running a studio of one
The Hark Co. · Founder
The Hark Co. was a small studio building AI products for B2B operators. I ran it for a year, from May 2025 to May 2026, alongside college at IIT Patna.
Founders came with rough ideas and left with shipped products. The studio made $8K from five clients in the US, the UK and the Netherlands, with no outside funding. Two of the five came back, and the longest engagement ran six months.
Two problems took up most of that year: finding clients, and shipping for them with three contract specialists spread over three time zones.
Finding clients
A studio with no track record lives or dies on outreach. The usual tool is cold email, and the usual reply rate is 3 to 5%.
So I built an outbound agent. It enriched each prospect, scored them on fit and on buying signals, and drafted the outreach. Then it stopped and waited for me.
The agent did the research and the first draft. The send button stayed with me.
A founder reads your first email once. If it gets their company wrong, there's no second email. The agent was fast at the research and I was the check on the one step you can't take back.
It's the same split I worked on later at IBY: agents that ask before they act.
- Cold email norm3 to 5
- The Hark Co.58
400 prospects turned into 230 conversations.
Where it leaked
I also built a lead-scoring tool to see where the outreach leaked. Here is the whole path, from 400 prospects to 5 clients.
Tap a step.
The share on each arrow is how many made it to the next step.
Replies were the easy part. The drop is in the middle, where 230 conversations became 10 pitches. Once a founder wanted a proposal, half of them signed.
So the work that mattered wasn't sending more email. It was getting from a friendly reply to a founder who wanted to see a plan.
So I moved the pitch earlier. Every good first call ended with a one-page plan for the founder's product, written that day, so the second conversation was about their plan instead of about us.
Here's why that one step is worth the attention. Everything else stays as it was; only the middle changes.
Holding 230 conversations and half of pitches signing, as they were.
One engagement, start to finish
The six-month one started the way most did: a founder with a manual process, a spreadsheet and a deck. The first call ended with a one-page plan, and the first week was spent sitting in on the process itself, because the spreadsheet showed what they did but not why.
We shipped a first version in three weeks that did one step of the process end to end, and nothing else. The founder used it on real work that week, and the next five months came from what they asked for once it was in their hands, not from the original plan.
Shipping with three specialists
I shipped with three contract specialists across three time zones. I owned scope, schedule and delivery for every project, so a slip anywhere was mine to answer for.
The work ran from MVPs for founders who had only a deck to AI features added to products that already had users. The stack stayed small on purpose: Next.js and TypeScript for the product, Python for the agent work, and a hosted Postgres, so any of the four of us could pick up any project.
Tap a client.
The longest engagement ran six months.
Revenue is the number people ask about. The one I'd point to is that two of the five came back. For a studio this size, a returning client skips the whole funnel above, and the 4% step with it.
Clients who come back
- Reply rateconversations per 100 prospects
- Pitch ratepitches per conversation
- Close rateclients per pitch
- Time to shipfrom signing to the first thing live
What I kept
An agent that drafts and a person who approves turned out to be a good way to split the work, and I've built versions of that split since, at IBY and in DealDesk.
The other thing is the funnel. Before the lead-scoring tool I'd have guessed the problem was volume. The numbers said it was one step in the middle.