AI-native CX intelligence that tells you what's breaking, what it's costing, and what to fix.
Every company that sells anything is sitting on a pile of customer feedback it barely reads. The complaints in the support queue. The one-star review left at 11pm. The phone call where someone spent four minutes explaining why the delivery slot didn't work. Individually, each is noise. In aggregate, it's the truth about the business - and almost nobody has time to read all of it. SentiSum's entire reason for existing is that gap.
The London company builds what it calls AI-native CX intelligence. In plain terms: software that reads every customer conversation across every channel and answers three questions a customer-experience leader actually cares about. What's breaking? What is it costing us? What do we fix first? The homepage doesn't hedge about it. The line it leads with is "Your CX is leaking six figures a month." It is the kind of sentence that either annoys you or makes you open a spreadsheet.
For most of the last two decades, "listening to the customer" meant sending a survey. The problem is that surveys mostly reach the people who already love you or already hate you, at a response rate that rounds to a rounding error. The 96% who don't reply are the ones you most need to hear. SentiSum's founding bet was that the real data is already there, buried in the conversations customers have with you every day - support tickets, live chats, call transcripts, app-store reviews, NPS free-text - and that the job is reading it, not collecting more of it.
That bet has aged well. CEO Sharad Khandelwal, who came up through a machine-learning background before moving into customer experience, has spent years making a fairly contrarian argument in interviews: that the survey, as a primary tool, is on its way out. The pitch isn't anti-feedback. It's that feedback you have to beg for is worse than feedback you already own.
Tell me what's breaking, what it's costing, and what to fix.— The three-question test SentiSum built the product around
There's a specific kind of frustration this targets. A support director can tell you their team closed ten thousand tickets last month. What they usually can't tell you, without a week of manual tagging, is that a fifth of those tickets traced back to one broken checkout step, or that returns complaints quietly doubled after a courier change. The answer exists inside the tickets. It just takes reading all ten thousand of them, which is exactly the task humans quietly give up on and software is good at.
At the centre is an engine SentiSum calls Kyo. It ingests the messy, multi-channel stream of customer conversations and turns it into structure - topics, sentiment, and, crucially, root causes rather than surface symptoms. On top of that sit a handful of agents, each doing one job. The Insights Agent explains the root cause of an issue and attaches a pound figure to it in plain English. The Early Warning Agent watches for anomalies, so a spike becomes visible before it hardens into a trend. The Quality Agent scores conversations and predicts satisfaction without anyone filling in a survey.
The last box in that chain is the part that quietly matters most. SentiSum's stance is that dashboards are where insights go to be ignored. So instead of building another screen nobody opens on a Monday, it pushes what it finds into the places teams already live - Slack, email, and increasingly AI assistants. The insight comes to you; you don't go hunting for it.
The customer list reads like a receipt from a week of British consumer life. The recipe-box company Gousto uses it to get at the voice of the customer behind its support tickets. Ticketmaster, the shoe retailer Schuh, the parking app JustPark, the fintech Cleo, the dog-food subscription Butternut Box, the Spanish marketplace Wallapop, and the consumer-goods group BAT all appear among its users. Khandelwal's own back catalogue of CX work includes names like British Airways and Just Eat. The through-line is volume: these are businesses drowning in customer contact, where reading everything by hand stopped being possible years ago.
Voice-of-customer analytics is not an empty field. Enterprise incumbents like Medallia and Qualtrics own the survey-led end of the market. A cohort of newer text-analytics players - Chattermill, Thematic, Idiomatic, Enterpret, Wonderflow - crowd the middle. SentiSum's differentiation is less about having a topic model nobody else has and more about the framing. Most tools sell "insight." SentiSum sells a number: the cost of the problem, tied to a specific fix, delivered where the team works. Quantifying the pounds is what gets a CX finding taken seriously in a room full of people who think in revenue.
SentiSum runs as a B2B SaaS company, selling subscription access to its platform and plugging into the helpdesks its customers already run - Zendesk, Freshdesk, Dixa, Gorgias. It has a partnership with the outsourced contact-centre group Ascensos, which folds SentiSum's analysis into the CX work it does for its own clients. The funding story is deliberately modest: an early angel round, then a seed round of roughly $700K in 2017 with backers including Newable, 500 Global and the London Co-Investment Fund. No nine-figure war chest, no blitz.
What's notable is the shape that produced. By 2025 the company was reported to be at around $2.5M in revenue with a team of roughly 23 people. That's a lean ratio, and an increasingly fashionable one - proof that an AI-era software company doesn't need to raise enormous sums to build something real. In 2024 the founding bench got deeper when Vivek Ganotra, an enterprise veteran with time at Salesforce, THG, BAT and Accenture, joined as a co-founder to help scale the go-to-market side.
Unify feedback. Find the "Why". Act fast.— SentiSum's own three-step summary of the loop
Khandelwal is the kind of founder who talks about customer experience in the language of data rather than empathy posters. His route in ran through finance and machine learning before it ran through CX, and that shows in the product's instincts: treat feedback as a dataset with a signal to be extracted, not a mood to be managed. He's spent a decade making the same argument on podcasts and stages - that companies over-trust surveys and under-read the conversations they already have - and SentiSum is essentially that argument turned into software. The 2024 addition of Vivek Ganotra, who spent years selling into large enterprises, pairs that product conviction with someone who knows how to get it bought.
Culturally, the company runs remote-first with periodic in-person gatherings, and describes a flat, everyone's-idea-counts way of working - the kind of setup a 23-person team can sustain and a 300-person one usually can't. It's a deliberate size. Staying small keeps the ratio of revenue to headcount healthy and the product decisions close to the people who use it.
Step back and SentiSum is a small company standing on a large trend. For years, "understand your customer" was a slogan attached to tools that mostly produced charts. The generative-AI wave changed the economics of reading unstructured text at scale, and companies that were built around that assumption - read everything, structure it, price it, route it - are the ones now able to make the slogan literal. SentiSum's job is the unglamorous middle of that story: not the model, but the plumbing and the framing that turn a firehose of complaints into a short, ranked list somebody can act on before Friday.
It's certified where enterprise buyers need it to be - SOC 2 Type 2, GDPR - and it reads feedback across more than 100 languages, which matters when your customers are spread across markets. None of that is the headline. The headline is the reframe: bad customer experience isn't a soft problem to be surveyed. It's a number, and SentiSum wants to be the company that reads it back to you.