When most AI startups are still thinking about how to “help users work faster,” one company has chosen a different path: do the work for them.
That company is Outset. Its product is an AI-moderated user research platform. Not an AI that simply transcribes meeting notes, but an AI researcher that can act as an interview moderator, ask follow-up questions, analyze conversations in real time, and generate insight reports.
The results are striking. HubSpot used it to complete more than 100 in-depth user interviews in a matter of days. Microsoft used it to improve Copilot retention by 5%. At Away, one researcher completed 75 interviews overnight.
And you may never have heard of it.
While Chinese media was busy covering ChatGPT, Claude, and Midjourney, Outset quietly built a surprising commercialization path in the niche market of enterprise user research.
This article breaks down why.
Why “Replacement” Is More Valuable Than “Assistance”
Start with the pain of traditional user research.
Imagine you are a researcher at a SaaS company. A product manager asks you to understand real user feedback on a new AI feature by next week. You need to:
- write an interview guide
- recruit participants
- schedule sessions
- run 20 to 30 one-on-one video interviews, each lasting 45 to 60 minutes
- transcribe recordings
- code and analyze the data
- write a report
For one person, that takes at least two to three weeks. Often it takes one to two months.
The most expensive part is not the analysis. It is the interviews themselves. Even an experienced researcher can only conduct four or five deep interviews in a day before cognitive overload sets in.
Outset’s product logic is simple: let AI run the interviews.
The user defines the research goal and drafts the question framework. Outset’s AI moderator invites participants, follows the planned guide, and can probe or ask spontaneous follow-ups based on the respondent’s answers, much like an experienced human moderator.
After the interviews, AI performs cross-interview theme clustering, sentiment analysis, quote tagging, and report generation.
That is replacement, not assistance.
Most AI products operate in assistance mode: the human does 90% of the work, and AI speeds up 10%. Outset operates in replacement mode: the human does 10%, such as designing the question framework and interpreting the results, while AI handles the remaining 90%.
Why is replacement more commercially valuable? Because customers are not paying Outset for “saving a bit of time.” They are paying because it lets them do something they could not previously do.
HubSpot’s research team used to run at most four rounds of deep research per year. With Outset, it can run 12 continuous rounds of research and increase sample size in each round. That is not an efficiency gain. It is a capability jump.
The Commercial Signals Hidden in Its Model
Outset follows an enterprise PLG path: product-led growth, but enterprise buyers pay.
On its website, you will not find a simple Free versus Pro pricing table. The pricing page points to custom pricing. The funnel is: enter email, book a demo, sales follow-up, enterprise contract.
That is not an accident. Its customer list, including HubSpot, Microsoft, WeightWatchers, Glassdoor, and Away, suggests an enterprise SaaS product with annual contract values likely in the $50,000 to $500,000 range.
This model offers direct lessons for B2B founders.
First, the product must be thick enough. If your AI product only helps someone rewrite copy, enterprises will not want to go through a custom pricing process. Only when AI can take over a complete workflow, not just one step, will enterprise buyers sit down to discuss a contract.
Second, build a quantifiable ROI narrative. Outset’s website is full of numbers such as “100 interviews completed in days” and “Copilot retention improved by 5%.” These are not vague claims about efficiency. They are concrete before-and-after metrics.
Third, use flagship customers to pull the long tail. If Outset can impress research teams at HubSpot and Microsoft, every mid-sized company researcher starts asking: if Microsoft is using this, should we try it too?
Why Users Do Not Just Use ChatGPT
A natural question follows: if GPT-5 or Claude is so powerful, why not use a general-purpose model for research?
The answer is Outset’s core moat.
The first layer is workflow packaging. A general model can answer a single prompt. Outset wraps the entire loop of design, recruitment, execution, analysis, and reporting. ChatGPT cannot recruit participants, schedule sessions, and send reminder emails by itself. That is product work.
The second layer is domain knowledge. Good user research is far more than asking questions. Researchers need to know when to probe, when to stay quiet, and how to explore contradictory answers. Outset’s AI moderator is tuned for this specific craft, moving between friendly conversation and deep exploratory questioning.
The third layer is a data network effect. Every interview conducted in Outset improves the system. The more interviews a customer runs, the more the system understands that customer’s user segments, industry terminology, and research framework. Switching costs grow with use.
The fourth layer is compliance and enterprise trust. Outset has enterprise-grade security credentials and data-processing agreements with large customers. That is not something a weekend SaaS wrapper can replicate.
Three Moves AI Founders Can Copy
Move One: Find Replacement Workflows, Not Assistance Workflows
The test is simple: if you remove AI from this workflow, would the company need another full-time person to do the work?
If the answer is yes, that is a replacement opportunity.
User research fits. So do bookkeeping tools such as Booke AI, legal document review products such as EvenUp, and insurance claims processing. The same replacement shift is happening across professional workflows.
Move Two: Use 10x Time Compression as the Anchor
Outset’s strongest message is not “our AI interviews are high quality.” It is “100 interviews completed in days.” Any executive can understand that.
Find your product’s X-to-Y story. Ideally, it should involve a 10x compression in time, cost, or throughput.
Move Three: Start With Enterprise PLG
Do not rush to build a free plan. If your product solves a clear enterprise-grade problem, either saving money or making money, go directly through demo, custom pricing, and enterprise contracts.
That increases ACV. More importantly, enterprise contracts have much higher renewal rates than individual subscriptions.
The First-Mover Advantages That Are Hard to Copy
Outset also benefited from timing and relationships.
It entered the market around 2022 and 2023, when AI capabilities were exploding but enterprise trust in AI remained cautious. “AI-moderated interviews” was a relatively safe use case. Outset used that safe entry point to build enterprise trust, then sold more ambitious products on top of it.
The founding team’s background also matters. Public information suggests the team came from product and research roles at Google and Facebook, with deep networks in user research. A research leader at HubSpot may already have known them. That kind of relationship capital is difficult for followers to copy.
But every successful startup has some luck and network advantage. The point is that when the luck arrived, the product was ready.
Closing Thought
Every major upgrade in AI capability creates a new replacement wave.
In 2023, the story was AI-assisted writing: Jasper, Copy.ai, and similar products.
In 2024, it was AI coding agents: Devin, Cursor, and others.
In 2025 and 2026, we are seeing AI replace professional workflows: Outset replaces parts of the researcher workflow, Booke AI replaces bookkeeping work, EvenUp replaces legal assistant work.
Their shared trait is that they are not satisfied with helping humans work faster. They actually finish the job.
What professional workflow will be next?
It depends on whether you can find the task humans do not want to do, still must do, and cannot afford to skip, then let AI take it over.
That is what Outset teaches.
Disclosure: Outset’s revenue and user numbers are not publicly disclosed. This analysis is based on public information from the company website, case studies, and third-party rankings. It is not investment advice.