A 12-Person Office Runs Six Texas Hotels — the Fix Wasn't AI, It Was Centralizing the PMS

Alexa MGMT is centralizing six Texas hotel properties onto one PMS with a 12-person home office team, betting that back-office automation — not a flashy AI feature — is what keeps front-desk staff focused on guests.

Alexa MGMT’s COO, Cole Di Carlo, framed the move explicitly around staffing constraints and guest experience: “you need to automate as much of the back-end, administrative work that bogs down front-of-house staff,” in order “to bring the human aspect of hospitality back.” The rollout is staged rather than a single cutover — it starts with the Holland Hotel in Alpine, Texas going live on Mews in August 2026, with the remaining properties migrating over a roughly six-month window. The announcement doubles as a scale marker for Mews itself, which the release states now serves approximately 15,000 customers across 85 countries.

Di Carlo’s framing lines up with RobosizeME’s Sean Anderson, who argued weeks earlier that not all hotel automation should be AI: for defined, repeatable back-office workflows like loyalty reconciliation and OTA invoice matching, he contends RPA is cheaper, more reliable, and easier to audit than an AI-based approach, and ties that directly to data sovereignty and PCI compliance requirements. Alexa MGMT’s own language never claims AI as the driver either — it’s a staffing-capacity and centralization story, the same category as Eccobell’s Kenilworth Hotel deployment, where a QR-code web app moved routine admin requests off the front desk so staff time went back to genuine guest interaction. Three different vendors, three different tools, the same underlying diagnosis: a lean team can’t absorb administrative load, so the fix is moving that load off human hands — regardless of whether “AI” is the label on the tool doing it.

For operators sizing their own back-office automation, the Alexa MGMT case is a useful data point precisely because it isn’t built around a model. It’s a staged migration with a named property, a real go-live date, and a stated reason — a more testable claim than most AI vendor pitches make.