This Startup Let AI Run Its Roadmap (The Results Were Insane)

Basedash handed an AI agent one number to raise, and set-ups went from five percent to over fifty in four months. The companies selling agents like this now give them a second number to protect, and that's the opening for founders.

In June 2025, a startup called Basedash gave an AI agent its billing records, its customer data and a seat in the team's Slack. It had one goal: get more new customers to finish setting up the product. The team made one rule for itself. It would build any suggestion that seemed reasonable.

Four months later, the share who finished setting up was ten times higher. Its founder, Max Musing, wrote one line about it: "This feels insane to us."

His blog post reports that set-up number and stops there.

Basedash turns a company's data into dashboards. In 2025 it relaunched around an AI analyst. Max calls getting set up "quite a difficult task."

A new customer had to find its database credentials, clear them with its security team and often run Basedash on its servers. After all that, it had to build a first dashboard or ask the AI a question.

In June, five percent of new customers made it that far. The rest signed up and stopped partway.

Max wrote: "Our small team doesn't include any dedicated data analysts or PMs."

So the team built an agent to do that job. They connected it to the product database, the billing records in Stripe, the tracking of what users click and the company's Slack. They told it which numbers they cared about and how a new customer moves through sign-up.

Each morning, the agent posted what it found to the team in Slack.

Between messages, it explored that data, looked things up online and reasoned through what it found. Max described the bet like this: "We wanted to see what would happen if AI could take the wheel completely."

One morning, the agent reported that companies who built a dashboard with several charts early on kept using Basedash for the long run. It backed the pattern with numbers and suggested a change to onboarding. The team built it.

In Max's words, most of the agent's findings were "things we wouldn't have even thought to look up."

Building those ideas took the engineers a few hours a week.

The agent watched one number. Support requests and cancellations were someone else's job.

By late October, the share of new customers who finished setting up passed fifty percent. That's ten times June's rate. Basedash says its revenue has been growing more than thirty percent each month.

Max credits most of that to hard work, with the AI as what he calls a secret weapon.

The agent hit the one number it was given.

Max's post reports set-ups alone, with support requests and cancellations left out.

A review of one of these agent startups put the risk plainly: "A system pushed to hit one number can damage what you didn't measure."

The AI researcher Janelle Shane put the danger in one line: "The danger of AI is not that it's going to rebel against us, it's that it's going to do exactly what we ask it to do."

Salman Jamali runs a startup called Augmeta that sells agents like these. He describes them this way: "They only stop when the outcome gets delivered."

That's the pitch. It's also the risk.

The companies that sell agents like this have started giving them a second number to protect.

Hightouch sells software that picks which email or offer each shopper gets from a brand. Its product lead calls agents like this "inherently greedy."

So Hightouch added a setting that holds back messages unlikely to pay off.

Sedai sells software that lowers cloud bills. It may save money only where the customer's app stays just as fast.

One customer, KnowBe4, trains office workers to spot scam emails. It cut the computing it pays for by twenty-seven percent.

Expect more agents to work this way.

One that raises sign-ups would also protect a second number, like site speed or cancellations.

That's the business opening. If you build an agent for a company, give it two numbers: one to raise and one it must not hurt.

Startups' cloud bills are one place to start.

Sedai's customers are large companies like Palo Alto Networks and GSK, so a new seller can go after smaller companies. Lower the bill while the app stays just as fast.

Basedash's agent did exactly what it was asked.

The next business is an agent that also knows what to protect.

Each week we find a business like this one inside a new market. Visit trends dot V C and join more than fifty thousand founders to get the next one.

This Startup Let AI Run Its Roadmap (The Results Were Insane)
Broadcast by