Emily Kucharski has heard the pitch before: AI can unlock insights from mountains of unstructured text. Survey responses, call transcripts, customer reviews—feed them into a model, and out comes intelligence.
The problem, she learned the hard way, is that enterprises can't always trust what comes out the other end.
So Kucharski, who spent a decade in market research at WPP and Publicis agencies, built WholeSum. The London-based startup just closed a $335,000 pre-seed extension on April 7, 2026, led by Love Ventures and Beamline, with participation from angels, bringing its total pre-seed haul to roughly £980,000, according to Tech.eu and Research Live. It's a modest sum in an era of frothy AI valuations, but WholeSum isn't chasing scale at all costs. It's chasing something harder: auditability.
The company's pitch centers on a specific frustration in regulated industries. Banks can't afford hallucinations. Pharmaceutical companies need to show their work. And yet most AI-powered text analysis tools, Kucharski argues, produce outputs that are neither reproducible nor statistically defensible. WholeSum's platform attempts to thread a narrow needle—combining large language models with statistical inference to deliver what it calls "uncertainty-aware" insights that can survive scrutiny from auditors and regulators.
Whether that's enough to carve out a defensible niche in the crowded AI analytics market remains to be seen. But early traction suggests some buyers are willing to pay for transparency.
The Genesis: When the Models Started Making Things Up
Kucharski encountered the trust problem firsthand at a prior venture, where she watched large language models hallucinate numbers and struggle with reproducibility. Her co-founder, Dr. Adam Kucharski—a mathematician, epidemiologist, and author—brought the statistical rigor needed to build a different kind of system. The platform deliberately avoids generative steps for numerical outputs or quote compilation, instead focusing on uncertainty-aware analysis that enterprises can trace back to source.
It's a narrow design philosophy, and perhaps a bet that some portion of the enterprise market values auditability over speed or flash.
The initial pre-seed came together in January 2026: roughly £730,000 total led by Twin Path Ventures, with participation from SFC Capital and angels via Ventures Together, plus a non-dilutive Women TechEU grant from the European Union. Twin Path, a UK AI-first pre-seed investor backed by British Business Investments, had spotted Kucharski's positioning early. The Women TechEU program, which targets women-led deep tech startups, added validation—and capital without dilution.
Then came the top-up. Love Ventures, a London EIS fund that invests from pre-seed to Series A with a productivity focus, led the $335,000 extension. Beamline, a deeptech accelerator, joined as well. The structure reflects WholeSum's dual identity: a commercial product aiming for revenue, but also a research-driven platform trying to solve a genuinely hard technical problem.
Early Clients: Banks, Pharma, and Private Equity
WholeSum launched its API in mid-March 2026. Within weeks, it had signed a major UK bank, a large pharmaceutical company enriching CRM data from field team notes, and a private equity group analyzing reviews. Other early customers include Nottingham Business School and Female Founders Rise, supported by Barclays UK.
That's a respectable start for a company barely three months into commercial availability, though it's worth noting WholeSum isn't disclosing deal sizes or usage metrics yet. The company, which incorporated in January 2023 as Wish I'd Known Ltd before rebranding to WholeSum Tech in December 2024, lists between two and ten employees on LinkedIn—a lean team, even by startup standards.
Kucharski told Research Live in March that WholeSum is hiring a founding engineer and an applied scientist to expand the platform's capabilities and support API pilots with market research agencies and enterprises. Growing the technical team seems urgent if the company hopes to move beyond initial pilots into production-scale deployments.
The Pitch: Reproducibility as a Feature, Not a Bug
WholeSum's architecture is built around control and transparency. The platform offers API-first deployment, with options for enterprise or local hosting. Data is encrypted at rest and in transit. The company says it doesn't train models on customer data—a table stakes promise in enterprise AI sales, but one that matters acutely in regulated industries.
The target market is clear: sectors where roughly 80 to 90 percent of data is unstructured, according to recent industry estimates, and where errors carry real consequences. Healthcare. Financial services. Market research. Anywhere that reproducibility isn't optional.
In an April statement reported by Research Live, Kucharski framed the problem as "AI's trust problem" in text analytics. It's a positioning that assumes buyers are increasingly skeptical of black-box outputs—that enterprises, burned by hallucinations or compliance failures, will pay a premium for systems that can explain themselves.
Maybe they will. The regulatory tailwinds are real. The European Union's AI Act, now in phased implementation, imposes transparency requirements on high-risk AI systems. Similar frameworks are emerging in the UK and US. If auditability becomes a regulatory mandate rather than a nice-to-have, WholeSum's early focus on statistical rigor could pay off.
Or perhaps the market will decide that "good enough" AI, delivered faster and cheaper, is preferable to the painstaking work of building auditable systems. That's the tension WholeSum will need to navigate as it scales.
For now, the company is betting that in some corners of the enterprise world, trust isn't negotiable. The next twelve months will reveal whether enough buyers agree to turn a modest pre-seed into a sustainable business.
