When a shopper lands on Puffy's website at 1 a.m. wondering whether the King mattress will fit through a 30-inch doorway, somebody has to answer. Increasingly, that somebody is software built by Alhena.ai, a Pleasanton company that says its agent resolved 63% of Puffy's customer inquiries without a human stepping in [Alhena.ai]. The pitch is unfashionably narrow: skip the general-purpose chatbot dream, and instead sell a guided-selling layer purpose-built for e-commerce categories where product complexity drives cart abandonment and returns [Alhena.ai].
Alhena, founded in 2022 and previously known as Gleen, rebranded earlier in its life to signal a sharper focus on revenue rather than deflection [Alhena.ai]. Co-founders Ashu Dubey (CEO, ex-LinkedIn growth) and Surya Krishnan are betting that the interesting money in customer experience is no longer in shaving a few cents off a support ticket. It is in turning the support surface into a storefront. Their product line now includes shopping assistants, support chatbots, and voice agents, all marketed under a "hallucination-free" banner [Alhena.ai].
The bet
The wedge is integration depth. Alhena's guided-selling layer connects into ERP for inventory and pricing, CPQ for configuration rules, and CMS, PIM, and CRM systems for everything else [Alhena.ai]. That plumbing matters because the categories Alhena is chasing, eyewear, mattresses, B2B industrial parts, are categories where a wrong answer about lens prescription, mattress firmness, or contract pricing means a return, a chargeback, or a lost account. The company says deployment takes under two hours and slots into existing help-desk tools [Prospeo.io], which is the kind of claim that matters more to a VP of CX than any benchmark score.
The customer list is eclectic in a way that suggests product-market fit is still being triangulated. Puffy (mattresses) and Tatcha (skincare) are pure DTC. SKF is a Swedish industrial bearings maker. LeanData sells revenue operations software. DIMO and Matter Labs (zkSync) are crypto infrastructure companies using Alhena to support developers rather than shoppers [Alhena.ai]. The common thread is products that are hard to explain, which is exactly the wedge the company claims.
Why it could be big
The back-of-envelope math on customer-experience automation is genuinely interesting. A mid-size DTC brand fielding 50,000 monthly inquiries at a fully loaded cost of roughly $6 per human-handled ticket spends about $3.6M a year on support. Resolving 60% with software, on the order of what Puffy reported [Alhena.ai], removes about $2.2M in annual cost before you count any incremental revenue from agents that actually close sales. Even if the realized savings are half that in practice, the unit economics are the kind that make a CFO take the meeting.
The cap table reflects investors who tend to back contrarian product bets rather than crowded categories: South Park Commons, Slow Ventures, The Spartan Group, and 6th Man Ventures, with roughly $4.9M disclosed in seed funding [Crunchbase, 2026]. That is a deliberately lean number for a company reporting $4M in revenue with 36 people [GetLatka]. If those figures hold, Alhena is operating at roughly $111K in revenue per employee, which is respectable for a seed-stage AI company and unusual for one that has not raised a priced growth round.
Disclosed seed funding | 4.9 | $M
Reported revenue | 4.0 | $M
The team and the traction
Dubey has spent the last two years on the podcast and interview circuit talking about generative AI for customer experience [CXChronicles Podcast; eCommerce Fastlane Podcast; Authority Magazine, 2026], which is roughly the right amount of founder visibility for a company at this stage. The broader team carries resumes from LinkedIn, Microsoft, Uber, Facebook, Siemens, Accenture, and McKinsey [Prospeo.io], and the company is currently hiring a founding full-stack engineer in Pleasanton [Alhena.ai], a title that suggests the engineering org is still small enough that the next hire shapes the roadmap.
The customer case studies are the strongest external signal. Puffy's 63% resolution rate is a specific, falsifiable number tied to a named brand [Alhena.ai]. The zkSync developer-support deployment is a different shape of proof: it shows the agent can ingest dense technical documentation and answer questions from engineers, not just shoppers asking about return windows [Alhena.ai].
The honest counterfactual
The bear case is competitive. Ada and Forethought have spent years and substantially more capital building general-purpose AI customer support, and both have enterprise sales motions Alhena has not yet demonstrated at scale. What bulls answer is that the incumbents were built for ticket deflection in the pre-LLM era, and their architectures show it. Alhena is wagering that a vertical, e-commerce-native agent that touches the ERP and the CPQ, not just the help desk, is a different product category than what Ada sells, and that complex-product retailers will pay for the difference. The Puffy and SKF deployments are early evidence for that thesis. Whether it generalizes to a hundred more accounts is the open question.
What to watch
The next twelve months should clarify two things. First, whether Alhena raises a priced Series A on the back of the reported revenue trajectory, and at what multiple, which will tell us how growth investors view vertical CX agents versus horizontal ones. Second, whether the customer mix consolidates around one or two verticals (mattresses, eyewear, industrial parts) or stays scattered. Focus tends to compound in this category. The voice-agent product, mentioned alongside the chat and shopping agents [Alhena.ai], is the other watch item: voice is where the next round of CX budget is moving, and shipping a credible voice agent on top of the existing integration layer would meaningfully widen the moat.
Alhena's job, in one sentence: beat Ada at the specific game of selling, not just supporting, complex products online. The incumbent has the logos and the balance sheet. The challenger has a sharper wedge and a shorter integration timeline. That is usually a more interesting fight than the funding gap suggests.