HITEC 2026 Was the Show Where Agentic Governance Reached Centerstage

While the vendors are marathoning from scrum to scrum to innovate their products, HITEC is a sprint. Everyone knew that ‘AI’ would be the buzz word, but for this iteration in San Antonio, the mission was to get a glimpse of which companies actually had the goods. 

The question is no more about chatbots, virtual concierge and one-off tools, but how hotels manage, govern, orchestrate and extract value from dozens and eventually hundreds of AI-powered workflows operating simultaneously across the enterprise. 

In short, AI agents are infrastructure, and fast movers can reap big rewards. I’m keeping this one brand-agnostic but needless to say, some vendors are definitely better than others.

Governance Via the Agent Management Platforms (AMP)

I’m calling this new software category the Agent Management Platform (AMP). With so many agents spread across disparate systems, tracking them all, your token spend and your fallback pathways becomes central to operations.

Which models are being used? What data can these agents access? What decisions are they authorized to make? What is the audit trail? Can you build in a PII scrub? And perhaps most importantly, what is the cost or, more specifically, how can we prevent ‘token creep’?

Whether in an orchestration framework or not, hoteliers are deploying specialized agents across revenue management, marketing, reservations, maintenance, finance, guest messaging and operations. What happens when one workflow uses GPT-5, another uses Claude, a third uses an industry-specific small language model, and all of them are consuming tokens at different rates?

Token Spend Optimization Requires Agentic Budgeting

Here’s the macro: the architects behind OpenAI and Claude seem to be acting a bit like heroin dealers. We’re all getting hooked with cheap subscriptions and freemium pricing, and yet both of these monoliths have promised IPOs later this year or not. Once they’re public, the goal shifts to profits above all. Set aside energy costs and sourcing rare earth metals to make more GPUs; that alone should set off alarm bells. This is where that token creep becomes a reality.

Today, many organizations treat AI costs as negligible. In three years, finance teams will be scrutinizing token consumption the same way they currently analyze labor costs or OTA commissions. Hotels will increasingly arbitrage tasks amongst different models based on performance, speed and cost. 

Simple guest requests may use inexpensive SLMs (small language models) while high-value commercial decisions are escalated to more sophisticated systems.

The AMP becomes the governance layer that determines which agent does what, what permissions it has and how token-efficient it operates.

The Disciplines of Harness Engineering and Loop Engineering

Another emerging concept visible throughout HITEC was the shift toward what can best be described as loop engineering and harness engineering. Much like how social media and paid search reinvented marketing, these two will soon be required business knowledge by senior managers. 

Loop engineering is the practice of designing operational workflows where AI continuously observes, recommends, acts and learns from outcomes, creating a self-improving cycle. Harness engineering is the complementary skill of knowing how to direct, constrain and govern those AI systems so they remain aligned with business objectives rather than simply generating more activity.

Rather than personally executing every task or reviewing every report, leaders increasingly become architects and validators of decision loops: defining objectives, establishing guardrails and determining where human intervention creates the most value. Success will depend less on managing individual processes and more on designing systems that can cycle through themselves while keeping the accountability for outcomes within the executive’s hands.

There’s tremendous value in automating BOH tasks in order to liberate team’s time and minds for guest-facing enhancements, but the real first step is the willingness or courage to change the way one works. The managers who thrive will be those who learn to build then trust well-designed loops without surrendering oversight, using AI not as a replacement for judgment but as a force multiplier for it.

The PMS Is Shifting to an Operating System

With the dawn of an AI-native PMS that can fulfill the tasks for an independent boutique property at the starting cost of $200 per month, one would be reasonable to predict that the heavy-hitting PMS vendors are destined to be usurped, with their only real moat being how they safeguard the payment gateway and the chain of custody around that.

Instead, there’s a feature-rich rebranding at play towards the PMS as an OS. The philosophy is no longer about ‘best in breed’ as that makes horrible business sense in the face of what unified data and agentic workflows above it can do. Wholly connected ecosystems are thus not just about the ‘single pane of glass’ for operations (that’s still important), but in getting even more predictive. 

Imagine restaurant pacing that combines reservations by segment, weather forecasts, group arrivals and local events. If you had that, you would know that a big golf group is going to be hitting the 19th hole at a specific time and can adjust accordingly by opening a new section. Or maybe, it’s that fewer people frequent the spa on a sunny day, but as a storm rolls through they’ll all venture indoors, allowing you to throttle spa treatments to only the high-margin sellers. Or think about itinerary generation that automatically incorporates spa availability, dining inventory, activity schedules and guest preferences. Or it can be real-time profitability optimization across every revenue center. 

Or, or, or…the use cases are endless! And if a specific niche capability is still missing? Increasingly, hotels may choose to build lightweight applications themselves using low-code and AI-assisted development tools rather than creating another long-term data silo. The future tech stack may be surprisingly (and finally) small.

Another Category with the Guest Success Management Systems (GSMS)

“It’s in the notes.” Such are the painful words of how experiential hotels have operated to date insofar as getting a clear view on who a guest is and their context for a given stay. Such notes can end up being dozens of pages long, making it impossible to read in real-time and also difficult for even the best CDPs to structure. 

This is where the emerging concept of a Guest Success Management System (GSMS) enters the picture. Think of it as sitting above the PMS, CRM and guest profile database, with an underlying logic and AI-native platform to rapidly exchange data then integrate into a UI that front desk associates or housekeepers can interpret. 

The GSMS is not about the post-stay RFM or RGR goals but about on-the-spot contextual intelligence so that an associate knows exactly what to do in the moment they need to do it. This is the layer that executes the wow moments. Is this guest in front of me celebrating an anniversary? Is this a returning guest who previously complained about room location? Is this guest allergic to bananas but love mangos? 

For luxury resorts, boutique hotels and independent properties, this category could is invaluable for making a great impression amidst a sea change of so many new luxury competitors looking to capitalize on the K-shaped economy. While the large chains will continue competing on scale, independents can compete on relevance. In this sense, the GSMS becomes a great tool to keep a lean team yet still deliver an anticipatory service level that wins better CSAT, loyalty, more direct bookings, ancillary spend and ADR growth.

The Real Story from HITEC 2026

Looking back on the week, the biggest takeaway is that hospitality technology is entering its second phase of AI adoption centered around accessible orchestration tools that can manage an entire ecosystem of agents, harmonized data and workflows. 

The winners, though, will be those who both move fast and break things, but also those with the best governance, the cleanest data foundations and the clearest understanding of where human judgment still creates value.