Market research has long sacrificed speed for rigor: weeks of recruiting and hours of interviews, not to mention fat invoices for decks that land after the moment has passed. The cash-advance company Keplar is built upon the premise that a conversational agent can turn this entire storyline on its head — with a voice AI to hunt down the full set of information in something around 30 seconds or less compared with several days.
The company says its AI programmatically carries out and analyzes customer interviews at scale, converting call recordings into synthesized insights, themes and executive-ready presentations. Keplar has raised a $3.4 million seed round led by Kleiner Perkins, with SV Angel, Common Metal and South Park Commons participating. Clorox and Intercom are among early customers.
Co-founders Dhruv Guliani, a former speech and voice AI engineer, and machine learning engineer William Wen started Keplar after talking to research leaders who were annoyed with slow cycles and superficial survey data. Their rationale: improvements in large language and speech models have made it possible for AI moderators to carry on complex back-and-forth interviews that sound natural and deliver stronger signal.
How Keplar’s voice researcher works to accelerate insights
Keplar integrates into a client’s CRM or list of customers, recruits the appropriate participants and then an AI interviewer is sent out by phone. The system calibrates questions on the fly according to responses, probes for detail and records paralinguistic cues that get lost in text surveys, like hesitations and emphasis.
Afterward, the platform auto-transcribes, codes themes, clusters sentiments and surfaces verbatims hinged to KPIs like churn risk or willingness to pay. The output is essentially what internal teams would receive from human-led qual: summaries, segment breakouts, highlight reels and slide-ready visuals. People respond to us as though we were human, addressed the way any agent speaks to them and sometimes using the moderator’s name — “Ellie,” “Andrew” or “Ryan,” Guliani said.
The point isn’t that AI makes research free; it’s that it makes research continuous. Rather than quarterly questionnaires, teams can have iterative conversations after every feature launch, campaign or pricing change and start to see trends developing within hours.
Why the market is ready for disruption in research now
The global insights industry is a huge, lumbering target. ESOMAR puts that at more than $100 billion if you include analytics, though much of the spend still goes toward manual interviews and static surveys. Yet response rates are plummeting: the Pew Research Center has reported single-digit participation in many phone surveys, a structural headwind for traditional approaches.
Voice AI has a different calculus. Completion rates of such conversational interfaces are higher than long-form questionnaires, and with branching dialogue, it is even possible to reclaim depth without inflating respondent burden. It’s time-to-insight for product teams: minutes to deploy, hours to patterns, not weeks, and a request loop.
There is a budgetary shift, too. With CX, growth and product analytics budgets combining, leaders are asking themselves why they should pay multiple times a year for custom studies to capture the same signals that always-on pipelines could continuously collect at far less expensive rates.
Quality, bias and compliance aren’t optional
No copy-pasting will ever defeat good old sloppy speed. The difficult challenges here are known to research pros: sampling bias, leading prompts and overfitting to the loudest segments. Keplar says they mitigate this with stratified outreach from CRM cohorts, thoughtful prompt design and clear confidence scores on findings. Independent validation will count; some clients will want to parallel-run human moderators to benchmark results.
Privacy and consent are also at the heart. If you record calls and produce transcripts, handling them must be in line with frameworks such as GDPR and CCPA, and opt-ins have to be crystal clear. Brand owners will also want to examine guardrails against hallucinations and the lineage of training data. Anticipate that procurement will ask whether each participant is aware they are talking to AI and how participants can get their data deleted or corrected.
Crowded field, clear angles in automated research
Keplar is hardly the only company pursuing automated qualitative research. Outset and Listen Labs have raised substantive rounds to attack similar pain points. Older players like Ipsos, Kantar, NIQ and survey platforms including Qualtrics are starting to push AI into their stacks, while voice model improvements from the biggest labs in town raise the technical bar each quarter.
Keplar’s value proposition is based on owning the entire interview end-to-end: natural-sounding voices, adaptive probing and packaging that lands directly in executive workflows. If it reliably turns raw calls into actionable, prioritizable work — not just transcripts! — it can sit next to product analytics — not just within the research function.
What adoption might look like in early enterprise use
Early enterprise use cases are simple: churn interviews within 24 hours of cancellation; immediate pricing and packaging probes after a roadmap change; post-campaign brand lift conversations that reveal creative fatigue; and empathy studies that translate pain points into backlog items. In both cases, throughput wins — dozens or hundreds of meaningful conversations per day — and the ability to trend qualitative themes just like any other metric.
If Keplar can deliver quality at scale, the result is a new rhythm of operations: always-on, voice-led insight pipelines that render quarterly research cycles antiquated. The ceiling is high but so is the scrutiny. Backed by heavyweight investors like Kleiner Perkins and logos to match, including Clorox and Intercom, the company has a real shot at proving that a well-designed AI moderator can do more than mimic human researchers; it can offer every team an ongoing conversation with its market.