The episode in short.
Matt Swulinski is Head of Growth at Viktor, an AI employee that lives in Slack and Microsoft Teams, reads the context of a whole workspace and picks up work without waiting to be asked. Before Viktor he worked on growth at Superhuman and was marketing employee one at Wispr Flow. In episode 2 of Startup Discovery he explains why Viktor sells itself as a hire rather than a tool, why teams with five or more active users churn at under 1% a month, and why the pricing is linear. It is a close look at the growth mechanics of an AI product at 50m+ ARR, built mainly from Warsaw.
What we covered.
- 1:27 (watch from here on YouTube)
What does Viktor do, and who uses it?
Viktor exists to remove the admin work that sits between the tasks that matter in knowledge work. It is the team's third attempt at that idea, after browser automation and then an email agent: email produced drafts rather than finished work, so the AI agent moved to Slack, where a small company's communication actually happens. Setup takes a few minutes, after which Viktor reads the workspace and starts taking on work. Its core customers are businesses under 100 people.
- 5:14 (watch from here on YouTube)
How did Matt get from Superhuman to Viktor?
Matt taught himself growth marketing by running his own agencies after a short spell in corporate consulting. He then spent four years in SaaS: on Superhuman's early growth team, then as marketing employee one at Wispr Flow, where he ran most acquisition from before revenue to a $2B valuation. He joined Viktor full time in July and works from the founder's office across marketing, product vision and investor conversations.
- 6:55 (watch from here on YouTube)
How does Viktor manage to be proactive?
Matt credits the agent harness the CTO built, designed to keep working as frontier models change, and says much of the engineering goes into choosing the right moment and place to message someone. Most software has about 30 minutes to impress a new user; Viktor keeps watching tasks in the channels it is added to, like a background agent, which Matt calls an infinite time to value. His example: when a teammate who has never used Viktor pushes a high-priority due date back three times, Viktor researches the task, delivers part of the work with a spec for the rest, and offers to help.
- 11:01 (watch from here on YouTube)
Why does Viktor call itself a hire, not a tool?
Matt walks through the labels: an AI agent sounds like a bot you hand a task to, an AI coworker helps one person, and an AI teammate sounds like a junior colleague. Viktor chose AI employee because it speaks to the CEO, COO or head of people looking at the gaps in a hiring plan: an employee owns a role and does not wait for a to-do list. People judge tools on cost and hires on return (AI ROI), so the framing puts Viktor into a different buying decision.
- 14:08 (watch from here on YouTube)
Do generous free credits hurt retention?
Not when they bring the team in. Usage sticks once five or more people at a company use Viktor, and for teams with five or more active users churn is under 1% a month. The free credits are designed to get the admin who installed Viktor to invite colleagues, and when someone tries to cancel, Viktor shows what their colleagues have built and now depend on, which usually turns the cancellation into a conversation about using the credits better.
- 16:45 (watch from here on YouTube)
Why is Viktor's pricing linear?
Credits scale in a straight line: $100 buys 40,000 credits, the most popular tier, and $1,000 buys 400,000. Matt's reasoning is that the work Viktor produces does not depend on team size (closer to cost per task than per seat), so there is no reason to push anyone onto a bigger plan, and volume discounts would tell small teams they get less value. Pricing is usage based: most customers start on a small tier, turn on automatic top-ups with a ceiling, and spend according to the work they get back.
- 18:34 (watch from here on YouTube)
Can a team build its own Viktor?
It is the classic build vs buy question. Getting something that roughly works is easy; the hard parts are durability, reliability, finishing work without mistakes, and integrations. Viktor has about 3,300 integrations live from a catalog of 9,000 it is rolling out. It builds its own for the most important tools, uses aggregators such as Pipedream, Composio and Zapier for the rest, and fixes the aggregator integrations nobody maintains.
- 20:40 (watch from here on YouTube)
How does Viktor compare with Claude Tag?
In Matt's view Claude Tag is built for the enterprise: it scopes permissions and context to channels, and in his experience its setup is the hard part. Viktor bets on small businesses without silos, where one agent with whole-company context and memory makes better decisions. He praises Anthropic's models, which Viktor mainly runs on, says Viktor's numbers did not drop when Claude Tag launched, and reads the launch as validation of the market.
- 25:23 (watch from here on YouTube)
What comes after chat: Viktor Spaces and custom software?
Slack works as a company operating system but is a poor place to read and work with documents, which is where Viktor Spaces (apps and docs) come in. Viktor can build and maintain dashboards with their own database, buttons and actions, which Matt sees replacing a lot of BI tool work. He expects every business to end up with its own software, such as a CRM fitted to a narrow niche instead of a generic setup, and the Viktor team already uses it for decks, mockups, dashboards and a personal chief of staff.
- 30:12 (watch from here on YouTube)
Why build inside Slack and Teams instead of a new app?
Getting people to adopt a new surface is hard, while Slack already holds the messages, documents, decisions and unfinished tasks an agent needs. An agent that reads that context becomes part of the team; a separate desktop app would go unused. Simon adds from his own sales experience that companies resist new tools with high training and integration costs, and adding an agent to Slack or Teams costs almost nothing.
- 32:07 (watch from here on YouTube)
One agent for everything, or many?
Matt believes one agent on frontier models produces better work than a collection of specialised bots, the multi-agent approach. Picking which bot handles what adds cognitive load, while he can run 11 Viktor threads at once without re-checking context or integrations. He takes the competing agents from xAI and OpenAI seriously but sees them built around personal use and quick replies, and says no customer has named one as a reason for leaving.
- 34:56 (watch from here on YouTube)
What does a day as Head of Growth look like?
Matt owns acquisition: affiliates, partner programs, performance marketing, landing pages and the website. About 60% of his time goes to moving work forward in Viktor threads, much of the rest to hiring with a high bar, and the remainder to whiteboarding with engineering, product and design.
- 37:26 (watch from here on YouTube)
Can you build a generational AI company from Warsaw?
Warsaw is Viktor's main office, with smaller teams in Munich and Dublin and a New York office opening. Matt points to talent arriving from Ukraine and Germany, returning Polish expats, and quality of life. Companies like Lovable and ElevenLabs show that Europe works if you plan for global growth from day one, and after weeks in San Francisco in September he found his team's culture and shipping pace held up well against most teams there, at lower cost and with fewer distractions.
- 41:44 (watch from here on YouTube)
What did Superhuman and Wispr Flow teach Matt about growth?
At Superhuman he learned product-led growth and referral loops, and built its sponsorship, referral and affiliate programs once word of mouth hit a ceiling; his lesson is that affiliates and sponsorships deserve a place beside paid and organic. At Wispr Flow he built acquisition from zero. Both products make one person more productive, whereas Viktor gets more valuable as more people add their context, which he sees as a much steeper value curve.
- 44:51 (watch from here on YouTube)
What is Matt's advice for young founders?
Polish the experience, because vibe coding something is easy and making it enjoyable is hard. Start paid acquisition early, since it forces you to build conversion tracking, attribution and analytics, beginning with Meta, Google search on brand and intent terms, and lifecycle email for incomplete signups. Invest in creative that explains the product in plain language, rewrite generic AI copy yourself, and give agents feedback the way you would a direct report instead of quietly fixing their output (human in the loop).
- 49:04 (watch from here on YouTube)
Why does taste matter when AI output is everywhere?
No number of agents replaces taste, least of all against workslop: the judgment that something is good and the sense of how to make it better. Matt expects companies to become boards of about three people with agents doing the rest, and what will separate them is taste, vision, drive and the choice of what to work on. Taste stays subjective, so founders should train it on purpose.