Trends Podcast

In three months, a startup called Viktor went from launch to a pace of fifteen million dollars a year.

About four months after that number was reported, Slack announced an AI agent of its own that does the same job.

Viktor is an AI coworker a team shares inside that chat app, the one many companies work in all day. Anyone on the team can ask it for a report or a draft.

It does the work where the rest of the team can see it.

AlphaSignal is an eight-person company that runs a technical AI newsletter with more than two hundred eighty thousand subscribers.

It sells sponsorships. Its business runs in Slack.

On February nineteenth, a salesperson named John Apolinar installed Viktor in AlphaSignal's Slack to help him research sales prospects. He didn't tell anyone.

On March second, the general manager, Kevin Curry, sent Viktor his first message. He asked it to pull up the numbers for a call with his cofounder.

Two weeks later, Kevin asked Viktor for a sponsorship proposal for a client.

Thirty minutes later, the proposal was ready.

Two days after that, he asked it to get the sales commission spreadsheet in order. By midnight, the sales team had a working app for it.

Kevin posted about it: "This is not a chatbot. It's a coworker."

By late March, three more salespeople were building proposals with Viktor.

It spread by example. Colleagues saw each other asking Viktor to do tasks and started asking too.

In April, Viktor caught a five thousand seven hundred dollar mistake in one client's contract. It worked out what that meant for each salesperson's commission, down to the cent.

Sixty-seven days after John's quiet install, the team was running eighteen automated workflows through Viktor.

Viktor's case study says a proposal that took ninety minutes now takes ten. It says the eight of them work like a team of fifteen.

Since Viktor charges one price for the team rather than per person, adding colleagues kept the bill the same.

Teams pick a plan by how much work they give it, from fifty dollars a month up to fifty thousand.

AlphaSignal was one of many. In May, the investment firm Accel led a seventy-five million dollar round in Viktor.

Two of Slack's cofounders invested too.

More than two thousand organizations used Viktor by then.

A run rate is a company's latest month of revenue multiplied by twelve. Viktor's was fifteen million dollars a year.

Slack tried this once before. In twenty sixteen, it set up an eighty million dollar fund for startups building chat bots. In its first seven months, the fund invested about two million dollars across fourteen startups.

Those bots posted alerts and took commands. The alerts were mostly noise, so teams switched the bots off.

Unlike those bots, Viktor finishes the task, like the proposal Kevin got back in thirty minutes.

Viktor's number was reported on May twenty-first, twenty twenty-six. On September fifteenth, Slack announced its built-in agent for teams, included in some of its paid plans, inside the app Viktor runs on.

A startup selling a shared AI inside Slack now competes with Slack.

A small company can win by selling to agencies.

An agency keeps a separate group chat with each client, where that client asks about their project. An agency with thirty clients has thirty of those chats.

Today, at eleven at night, a client messages their chat to ask whether the new campaign still launches on Monday.

The question sits there all night.

The next morning, the account lead opens thirty chats, finds the question, looks up the campaign and writes back. The client waited until morning for a yes.

With thirty clients, each morning starts with a stack of last night's questions.

Now picture the same chat with an AI that knows that client's project.

The client asks at eleven at night. The AI answers right away with that client's dates and numbers.

If the question touches something sensitive, like a price change or a missed deadline, the AI holds it and flags it for the account lead.

In the morning, the account lead opens the chat and reads what the AI said. They answer only the questions that needed a person, like a client asking to move the launch date.

The client got an answer in the night.

The account lead started the day with less to chase.

That's the opportunity. Agencies pay for an AI assistant that answers their clients in the chat they already use. Each client only ever sees their project.

A company called Pylon already sells this kind of AI to companies that talk with their customers in shared chats. Its existing customers spent fifty-eight percent more a year later.

A customer that paid Pylon a hundred dollars one year paid about a hundred fifty-eight the next.

Once the AI knows the accounts, leaving gets expensive. After a year with an agency, it knows thirty clients' projects.

An agency that switches has to teach a new AI all of that again.

The price can grow with the agency too. ClearFeed sells support software for chats shared with clients and can charge per chat. So an agency's bill grows with each new client.

To build it, start with one agency that has ten to fifty clients, big enough for those morning stacks to pile up.

Connect your AI to each client's project, so it answers with only that client's numbers.

For the first week, read each question the clients ask. Teach the AI the ones it missed.

Show the account lead what the AI answered, so they trust it before the clients do.

Send anything sensitive to the account lead, so the agency stays in charge of the relationship.

Viktor went from launch to fifteen million dollars a year in three months. Visit trends dot V C and join more than fifty thousand founders to get the next opportunity.