AI Hospitality Group Takes the Whole Hotel P&L and Treats Token Cost as a Line Item
Kishan Dahya, CTO of AI Hospitality Group, explains how the AI-native hotel manager runs full property P&Ls, consolidates 13 to 15 systems into one agent layer, and moves workflows into cost-controlled production at about $3 per run.
Owner margins have eroded about 20% over five or six years, says Kishan Dahya, co-founder and CTO of AI Hospitality Group (AIHG), and his answer is to run the whole hotel rather than sell owners another tool. On the Hospitality Daily Podcast with Josiah Mackenzie, Dahya said roughly 34 cents of each revenue dollar goes to labor and 12 cents to franchise fees. AIHG takes over a property’s full P&L as an AI-native hotel manager, and 80 to 90% of its income comes from performance above a hurdle set before the management agreement is signed — so most of what it earns depends on results, not software seats.
The operating model starts with consolidation. AIHG pulls 13 to 15 systems — including the PMS (Stayntouch), ERP, POS and STAR data — into one agentic orchestration layer backed by a Microsoft Fabric data lake. Agents act as “digital employees” with their own instructions, skills, tools and scheduled automations, and they work inside email and Slack, where staff give approvals. Deterministic code handles routine steps; an LLM is used only where judgment is needed. Examples include a group-sales agent that parses lead emails and updates the CRM, and a daily deposit tracker. Automating housekeeping room assignment cut two to two and a half hours from the head housekeeper’s day at one Los Angeles County hotel.
Dahya is blunt about token economics. At about $3 per run, a GM spending $30 a day on prompts is not viable, so proven workflows move from chat experiments into cost-controlled production. Human approval stays in the loop elsewhere in the hotel back office too: Inn-Flow’s new AP Agent codes invoices to the general ledger, but every invoice still requires human review and approval.
For owners and asset managers, the useful test is Dahya’s: price what each automated workflow costs to run per day against the labor hours it returns, before scaling it across a portfolio. His view is that the bottleneck is now implementation, not model capability — which makes cost per workflow an operating metric, not an IT detail.
Source: Hospitality Daily — Kishan Dahya, CTO at AI Hospitality Group: Making AI Pay Off in Hotels Auto-generated brief — verified before publishing.