The AI that read your customer feedback so you didn't have to - one of the first companies to turn GPT-3 into a working analyst.
In 2020, a couple of brothers in Oakland got their hands on something most people had only read about: early access to OpenAI's GPT-3. Plenty of engineers used that access to make the model write poems or fake press releases. Daniel and Jeffrey Erickson used it to solve the least glamorous problem in software - the pile of customer feedback nobody has time to read. That was Viable, and for a moment it was doing "ask your data anything" before the phrase existed.
The premise was almost boring, which is usually the sign of a good one. Every company sits on a mountain of unstructured text: support tickets, survey responses, app-store reviews, sales-call transcripts, the internal notes nobody files. It is the most honest data a business owns and the hardest to use. Numbers go in a chart. Sentences go in a drawer. Viable's bet was that a language model could finally open the drawer.
The product could be reduced to a single interaction. You typed a question in plain English - "why are people canceling this month?" or "what do enterprise customers complain about most?" - and Viable read across every connected feedback source and wrote back a paragraph. Not a dashboard you had to interpret. Not a chart you had to build. An answer, in sentences, with the data behind it.
Founder Daniel Erickson put the pitch plainly in an interview: "There are no SQL queries or pivot tables, it's just asking questions in plain language and getting plain language, yet data-backed insights in return." The interface was the innovation. The model did the reading; the product decided that reading should feel like a conversation, not a query builder.
Viable was built for the people who own the feedback but rarely have time to sit with it: product managers, customer-experience leads, researchers, support teams. At the start the company aimed at scrappy early-stage startups still hunting for product-market fit - founders who needed to know what users actually wanted, fast.
Then something happened that founders dream about. Bigger companies started showing up on their own, carrying far larger piles of feedback and asking for help at scale. Erickson described the shift in modest terms: "It's a bit like we moved into customer feedback more broadly by popular demand." The company reported roughly a dozen paying customers in its early days, run by a team of about nine people, and steadily moved upmarket toward enterprise accounts.
The math on customer feedback never worked. A company might get thousands of support tickets a week. Reading them by hand does not scale, so teams sampled a few, guessed at the rest, and moved on. Tagging systems helped a little but demanded that someone build and maintain a taxonomy - a job that quietly rots the moment the person who owned it leaves.
Viable's argument was that this was never a reading problem. It was a translation problem. The feedback was already a database; it just happened to be written in English instead of columns and rows. A language model could sit between the two, and the research bottleneck - the analyst everyone waited on - could become something you queried directly.
The voice-of-customer market was not empty. Tools like Thematic, Chattermill, Idiomatic and the text-analytics features inside Medallia and Qualtrics were all working on the same corpus. What set Viable apart was less the model and more the refusal to hand you homework. Most tools gave you charts of themes and left the interpreting to you. Viable gave you the interpretation.
| The task | Older approach | Viable's approach |
|---|---|---|
| Getting an answer | Build a query or dashboard | Ask in plain English |
| Organizing feedback | Maintain a tag taxonomy | Auto-structured by AI |
| Who can use it | Analysts & BI teams | Any team, no SQL |
| Output | Charts to interpret | Written, sourced insight |
Erickson, a self-taught coder who had been CTO of Getable and VP of Engineering at Eaze, reportedly spent about a month experimenting with GPT-3 and the company's own data before the shape of the product clicked into place. The realization was that a question-and-answer system should be the core, not a side feature. OpenAI later featured Viable as an example of GPT-3 powering a new generation of applications.
As the platform matured, it stopped leaning on a single model. Later versions blended GPT-4, GPT-3 and roughly nine proprietary models built in-house - the off-the-shelf model for language, the custom ones for accuracy and for tailoring insights to what a specific business cared about. It was an early, practical version of a lesson the whole industry would learn: the foundation model is the engine, not the product.
Investors liked the wedge. An early seed round of about $4M was backed by Craft Ventures - the firm with Tesla, Airbnb and Bird in its history - alongside Javelin Venture Partners. In May 2022, Streamlined Ventures led a further $5M, with Craft and Javelin returning and new names including Merus Capital, GTMFund, Stratminds and Tempo Ventures joining. That brought the total to roughly $9M.
The business model matched the customer. Viable sold tiered SaaS subscriptions priced by data volume - a Startup plan around $600 a month for 2,000 datapoints, a Growth plan near $1,000 for 5,000, and custom enterprise pricing above that. Simple, legible, and pointed at the teams generating the most feedback.
Viable's timing was its best asset and, eventually, its hardest test. Being among the first to ship a real GPT-3 product meant it understood the "talk to your data" pattern years before it became a default. But that same pattern grew into one of the most contested ideas in software - every analytics vendor, every foundation-model lab, every new startup eventually offered some version of "ask your data anything."
Around mid-2024, third-party AI directories began marking Viable's product inactive, and the askviable.com site later went offline. No public statement explained whether the company was wound down, absorbed, or simply moved on. What is not in doubt is where it stood in the story: Viable scouted a path in 2020 that a large part of the industry is now walking.