AI agents at work: Replit’s results, $28k teams, and more
Headlines and trends
Replit’s self-driving company experiment
At Replit, agents have absorbed much of the coordination overhead that normally eats company time. Designers now ship product changes directly instead of handing specs to engineers. Support teams use agents to investigate customer problems. Engineers wrote nearly three times as much code in the first half of 2026, with review times, rollbacks, and incidents holding flat. The gap between Replit and most companies: those companies still run agents in personal chat windows with borrowed credentials and no audit trail.
Payments moats in an AI world
As AI lowers the barrier to copying software, defensibility is shifting toward regulatory approvals, banking relationships, and accumulated operational knowledge. The argument here is that records of agent decisions and human corrections over time become a competitive asset that no one can replicate by cloning your product.
Strategies and tactics
Start your agent deployment at the contact form
Inbound qualification is a high-ROI entry point for agent deployments because the task is bounded: answer questions, qualify the lead, route it, and book a meeting. The downside is minimal since the alternative is a form sitting in a queue. One case study cited 614 meetings booked and $1 million in closed revenue.
A direct report changed how one product head uses AI
A head of product at an AI startup tried Claude Cowork in January and found it couldn’t do the job. A month later, a newly hired product manager showed how he had Claude Code agents handling his work. Watching a direct report operate that far ahead moved the needle more than any product demo. By April, the head of product had Cowork running nine jobs. The practical takeaway: keep a list of what AI can’t do yet, and retry the top items every couple of months.
⚒️ Tools and resources
Full Claude Platform now on Google Cloud
Startups with a Google Cloud commitment can now spend it on the full Claude Platform, which Anthropic runs and Google bills. Code execution, skills, and web search are included. New features arrive the same day they land on Anthropic’s own platform. Managed Agents and the MCP connector are currently in beta.
Replit Monitoring Agent
Replit’s Monitoring Agent acts as an on-call engineer for your published app. It watches for errors and slow responses, investigates each alert, tests a fix, and emails you a summary. It can also push the fix live on its own if you give it permission.
Vercel Connect now searchable from the CLI
Vercel Connect CLI now lets you search over 1,000 services directly from the terminal using filters like category or connection method. JSON formatting is supported for scripts and agents. You need CLI version 64.1.0 or later to access this.
Miscellaneous
Tokens over hires: $28k/year, 657 pull requests in month one
Two founders are buying engineering capacity in tokens rather than headcount, spending about $28,000 a year on AI total. In their first month, they merged 657 pull requests. Their biggest single bill is roughly $1,500 a month for an AI code reviewer so neither founder has to read every first draft. Their real constraint turned out to be attention: while one founder was occupied elsewhere, a small database problem grew until it took the product down. Their conclusion is that their next hire will own one piece of the product end to end, agents included.
The Square Deal framework for AI founders
George F. Johnson’s Square Deal framework suggests sharing productivity gains among employees, customers, and the business. That means meaningful employee ownership, passing savings to customers, and reinvesting in growth. Lincoln Electric is cited as a proof point, with stock returns more than twice the S&P 500’s since its 1995 IPO.
⚡ Quick links
- Let chaos reign, then rein it in: Some of the first office PCs were expensed as desks or typewriters to get past IT.
- Advanced AI adopters now spend less: Companies furthest along with AI are starting to cut spend as cost controls tighten and models get cheaper.
- When anecdotes and metrics disagree: When customers complain but the dashboard looks fine, trust the customers and fix what the dashboard measures.
- Loving ugly babies: Most companies die in the long stretch before anyone cares, and that stretch is where most of the value gets built.
I’m founder of BizStack at Cagri Sarigoz LLC and a passionate advocate for entrepreneurs.
With over 14 years in tech, marketing, and AI, including my role as Head of SEO at CitizenShipper and co-founder of TaleBot at Intale AI, I’m dedicated to sharing genuine, useful product insights and tips.
At BizStack, I aim to cut through the digital noise to provide clear, actionable advice.
And more than all else, I’m a father to a (always) little girl and a husband.
Contact me at [email protected] for assistance.
