A Y Combinator startup thinks so. Whether retailers agree—and whether regulators will let them—is the harder question.
Lexius emerged from stealth this winter with a pitch that sounds almost too convenient: plug our software into whatever security cameras you already own, and we'll flag shoplifters in real time. No hardware swap. No installation crews. Just AI watching your existing feeds, then sending your staff a push notification—complete with video clip—when it thinks someone's pocketing merchandise.
The company isn't subtle about its timing. Retail shrink has become something of a moral panic, with the National Retail Federation logging an 18% spike in reported shoplifting incidents last year alone. Stack that onto a 93% climb between 2019 and 2023, and you have an industry willing to try just about anything. Violence tied to theft incidents jumped 17% year-over-year, per NRF's October survey.
Whether those figures represent genuine crime waves or convenient scapegoats for inventory write-offs—analysts remain split—matters less than the pressure retailers feel to do something. Lexius is betting that something involves software, not ripped-out cameras.
The Mechanics, Briefly
The system is straightforward, at least on paper. Lexius integrates with existing video management platforms—Lorex, Hikvision, UNV among them—and begins analyzing footage for what it considers suspicious behavior. Staff receive alerts on their phones the moment the algorithm thinks it's spotted theft in progress. They can review the clip, verify, and intervene.
Beyond the core detection, Lexius advertises traceback search across multiple feeds, "known offender recognition with chain-wide intelligence" (more on that landmine shortly), fall detection, and custom safety alerts. Pricing is tiered by camera count: Standard handles up to 10, Advanced up to 20, Enterprise offers unlimited cameras plus white-glove support. Actual dollar figures? Request only.
Installation supposedly takes minutes. The founders, David Elskamp and Liam Webster, are still in Y Combinator's Winter 2026 cohort—Demo Day lands March 24. Elskamp handles the CEO role with an AI/ML pedigree; Webster, the CTO, came out of UC Berkeley's EECS program with prior research stints in AI and privacy at ICSI. The team is three people. The Delaware incorporation and California foreign filing both date to June 2024.
The Numbers Lexius Wants You to See
The company has published a case study centered on Erewhon, the Los Angeles grocer that's become something of a cultural punchline for $20 smoothies and aspirational wellness signaling. According to Lexius's own write-up, the Beverly Hills location cut shrink by 60% within six weeks, prevented over $3,000 monthly in losses, halved manual footage review time, and delivered triple the return on investment.
There's an anecdote about catching checkout under-scanning the same day the system went live. All very tidy. All hosted on Lexius's site. No third-party auditor has weighed in.
A LinkedIn post claims similar wins at a Nisa convenience store: $2,350 in prevented theft, 90% less time spent reviewing video, 9x ROI. One complication—Nisa locations have been documented deploying competitor Veesion to reduce losses. Whether Nisa actually runs Lexius, or whether this is one franchisee experimenting, remains unclear.
These are vendor-supplied metrics. Treat accordingly.
Standing Room Only in This Aisle
Lexius is hardly alone. Veesion, probably the category leader, claims north of 5,000 store deployments and just closed a $43 million Series B. Business Insider has profiled them. Panoptyc says it's live in more than 20,000 locations with 30% shrink cuts inside six months. Scylla AI does behavior detection with optional watchlists. Aivilon runs on local network appliances and pings alerts via app or Telegram. 3DiVi layers facial recognition onto shoplifting detection for offender blacklists. XVeillance sells a plug-and-play "station" that hooks into existing CCTV and pushes mobile alerts alongside enterprise dashboards.
Every single one of them touts no new hardware. Every single one pushes mobile-first alerts. Lexius's edge, if there is one, appears to rest on integration simplicity, the traceback search functionality, and some emphatic privacy language on the website. Whether that's enough to crack a saturated market is anybody's guess.
The Privacy Problem Nobody Wants to Talk About
Now we get to the messy part.
Lexius advertises "known offender recognition with chain-wide intelligence." That phrasing suggests either facial recognition or person re-identification tech sophisticated enough to track individuals across multiple stores. At the same time, the marketing copy promises "zero racial profiling," insists the system "does not analyze skin color/clothing/race," and claims footage never leaves the retailer—Lexius doesn't access it. The company calls itself "fully compliant."
Compliant with what, exactly? The regulatory map for retail facial recognition looks like a patchwork quilt sewn by committees that never spoke to each other.
Portland, Oregon banned private use of facial recognition in places of public accommodation—retail included—back in January 2021. Violators face a private right of action and statutory damages. Illinois's Biometric Information Privacy Act has already spawned class actions; Target is currently defending a lawsuit alleging its anti-theft camera system violated BIPA by collecting biometric data without informed consent.
Then there's the FTC's 2023 settlement banning Rite Aid from deploying facial recognition in stores for five years, stemming from misidentifications and inadequate safeguards. The message there was unambiguous: retail FR carries reputational and regulatory risk even where it's technically legal.
Lexius hasn't disclosed whether "known offender recognition" actually generates facial templates. Or how long data persists. Or how the system would function in jurisdictions where biometric surveillance is restricted or banned outright. For a startup positioning itself as "privacy-first," those questions aren't optional extras. They're the ballgame.
What Happens Next
The iOS app (version 1.1.5 as of early February) is live. The website is operational. Y Combinator visibility should generate some early traction—Demo Day tends to do that. But Lexius will need more than ease of setup to win accounts at scale.
Retailers want false-positive rates. They want precision metrics separating actual theft from innocent browsing. They want to know how much staff time this actually saves versus creates. Those numbers aren't public yet. Competitors like Veesion have media validation and thousands of deployed sites. Lexius has vendor-hosted case studies and a YC badge.
The timing, admittedly, is right. Retailers are spending to stem losses. Nobody wants to replace functioning cameras. Whether Lexius can carve out defensible territory in an already-crowded market, navigate the regulatory thicket without stepping on a legal landmine, and deliver results that hold up under independent scrutiny—that's the real test.
For now, it's another AI pitch in a very crowded aisle. Just one with better integration promises and a privacy problem it hasn't fully explained yet.
