In 2012, “online qualitative research” sounded like a slightly awkward compromise. The useful bit of a focus group was the room: a moderator who could notice hesitation, ask the second question and watch somebody handle a product. The expensive bit was also the room: travel, facilities, recruiting, recording, interpreters, snacks, schedules and the mysterious back chamber where brand managers whispered behind glass. Discuss.io began by putting that room on a webcam.
Founders Zach Simmons and Jim Longo built an on-demand platform that connected brands and agencies with targeted participants for live interviews and focus groups. Simmons brought software experience, including time at Amazon Web Services. Longo brought decades in market research. The pairing mattered. Discuss was never merely video conferencing in a lab coat; it wrapped recruiting, research roles and project logistics around the call.
Today the company is called Discuss. Its platform runs human-led interviews, AI-moderated sessions, asynchronous video tasks, uploaded-research analysis, surveys, translation, highlight reels and global participant recruitment. Its customers include Danone, Mastercard, HelloFresh, Unilever, PepsiCo, HP, Dexcom and research agencies. In November 2025, Discuss combined with survey vendor Voxco and text-analytics product Ascribe. The new organization said it had more than 200 employees in 16 countries and 900 customers worldwide. Daniel Graff-Radford, previously Voxco’s CEO, took the chief executive job; outgoing Discuss CEO Simon Glass became an adviser.
That global footprint also shapes the workplace. Discuss describes itself as fully remote, with colleagues in 26 countries speaking 21 languages. Its published operating values are unusually direct for corporate wall art: build what matters, own it together, put integrity first and tell the truth early. Half of its leaders are women, according to the careers page. The details fit the product. A company selling better listening across cultures has little room to run as a single-office monoculture. They also explain why services remain part of the offer. Enterprise research is not solved by shipping software alone; somebody still has to recruit the right person, check the connection, manage consent and rescue a session when the camera refuses to cooperate.
The first thing that failed was the ordinary video call
Glass has described the early category without nostalgia: video interviews were clunky, broadband was unreliable, and online qual felt like a side project. Generic meeting tools could connect faces, but researchers needed consent, screened participants, stimuli, mobile screen sharing, an invisible observer room, session support and a fast route from a long recording to a usable clip. A Zoom link and a spreadsheet could imitate the surface. They could not make the workflow pleasant.
Then the pandemic changed the argument. Remote research stopped being an efficiency experiment and became one of the few ways to keep talking to consumers. Discuss had spent eight years living through the uncool part. When demand arrived, the product already understood that a moderator, participant and brand observer need different controls. The market’s objection changed from “is online research real?” to “how much of this can we run?”
“We’ve gained more in-depth understanding of our consumers in the past 6 months than the past 6 years.”Elizabeth Ackerly · Danone Strategic Insights & Analytics
Danone’s “Consumer Love” program shows what the software looks like when it becomes an operating system rather than a booking tool. More than 800 employees engaged with over 1,000 consumers in 25 countries, according to the case study. Teams saved more than 5,000 video moments and put them into presentations and decisions. They also asked Discuss’s Insights Agent more than 600 business questions. The result Danone emphasizes is organizational: consumer contact spread beyond the insights department to marketing, sales and senior leadership.
Three agents and one very human caveat
Discuss’s AI turn started with Genie, a generative-AI synthesis product launched in 2023. In February 2025 it introduced three purpose-built agents. Interview Agent conducts self-paced conversations and follows researcher-defined probes. Translate Agent handles multilingual material. Insights Agent searches across responses, finds themes and compares segments. Later releases added Virtual Personas grounded in a customer’s own research, quant-style visualization and controls for an interviewer’s voice, tone and probing depth.
The distinction from an AI-only interview startup is optionality. Mastercard uses AI moderation for message testing or product feedback, where speed and scale matter, while preferring human-led conversations for financial stress and other sensitive topics. Discuss can analyze both in the same environment. That is less theatrical than claiming a bot can replace a seasoned moderator. It is also more credible.
The company’s own practical guidance contains the important warning: a vague activity goal produces a vague study. One question should carry one idea. Probing instructions need to define what a useful answer looks like. Automation can produce more conversations, faster. It cannot rescue a confused brief, repair a biased sample or decide whether a customer’s polite answer is the truth.
Where this does not work
Skip the AI shortcut when the topic is highly sensitive, the sample is weak, the question needs expert improvisation, or the organization lacks people who can challenge the output. A fast summary of badly designed research is simply an error that reached the meeting early.
What it costs - and what customers claim to save
Discuss is enterprise software, and the pricing page behaves accordingly. Essentials, Elevate, Expert and Enterprise plans all say “request pricing.” Each includes unlimited users, live sessions, AI interviews, asynchronous research, uploaded files, AI setup and summaries, Virtual Personas and highlight reels. Higher plans bundle more support. Recruitment, project coordination, simultaneous translation, screener writing and reports can be included or purchased as add-ons. There is no honest public dollar figure for a typical contract because the scope changes with markets, services and volume.
The measurable pitch is avoiding per-project drag. Quadrant Strategies says AI moderation cut both its project turnaround and cost per project in half. Discuss cites conventional live translation at roughly $200 to $440 an hour; its in-room English translation aims to remove that interpreter layer for observers. Mastercard describes an unlimited-research model that lets teams run more qualitative work without a fresh internal budget fight for every conversation. These are customer and company claims, not a universal calculator. Recruiting a narrow B2B audience in five markets will never cost the same as interviewing ordinary shoppers in one.
The new product is memory
Most research dies in a deck. A study answers one launch question, gets presented once, then disappears into a drive named FINAL_v7. Six months later, a new team commissions a cousin of the same work. Discuss Intelligence, released in 2026, attacks that waste. A user asks a question in ordinary language; the system searches across past projects, scopes the relevant studies and returns an answer tied back to participants, segments, verbatim quotes and video clips.
Traceability is the product. A generic summary bot can sound decisive while sanding away contradictions. A useful research system must let a skeptical brand manager click from a claim to the interview, notice who said it and understand which market was represented. If Discuss keeps that chain intact, old research begins to compound. If teams upload poor studies or stop checking the evidence, the library becomes a beautifully organized rumor mill.
The Voxco combination makes the ambition larger. Qualitative research is good at the “why”; surveys are good at measuring the “what.” Ascribe adds open-ended text analytics. Discuss wants those modes to live in one continuous flow, from exploration to validation to retrieval. Competitors span Qualtrics and Forsta at the enterprise end, UserTesting and dscout in experience research, research-community platforms such as Recollective and Fuel Cycle, and AI-native interview vendors such as Outset and Listen Labs. The stubborn alternative is still an agency, a video tool and a heroic spreadsheet.
The part worth stealing
Productize the logistics
Find the skilled hour trapped inside weeks of coordination. Remove scheduling, handoffs and retrieval before trying to automate judgment.
Let customers choose the method
Use AI where consistency and volume win. Keep a human option where sensitivity, ambiguity or improvisation matters.
Make output reusable
A clip linked to a claim travels farther than a 70-page transcript. A searchable archive can answer the next question before a new project starts.
Measure decision latency
Hours saved are pleasant. The sharper metric is whether an answer arrives while a product, campaign or strategy can still change.
Discuss’s story is not that AI discovered the customer. It is that a company spent years digitizing the messy machinery around customer conversations, survived long enough for remote work to validate the first thesis, and then used AI to attack the next bottleneck: analysis and memory. The 2026 Forrester Wave named it one of two Leaders in experience research platforms, and Discuss announced an Inc. 5000 appearance in August. Awards are lagging indicators. The live test is simpler: can a team ask a better question, hear an honest answer and find it again when the next decision arrives?
Keep looking