TCPG: Token Cost Per Guest

The Buzz Is Tokens

The buzz around AI is no longer only about what it can do. Increasingly, it is about what it costs to run — and that cost is measured in tokens.

Hospitality is adopting AI at lightning speed. Guest messaging, concierge, translation, upselling, review responses, sentiment analysis, sales prospecting, RFP handling, revenue commentary, diagnostics, and owner reporting are all being routed through generative AI — often faster than finance and technology leaders can keep up with. Every one of those interactions consumes tokens, and every token has a price.

The industry has seen this pattern before. OTA commission was once the unremarkable price of incremental demand — until it grew into one of the most scrutinized costs on the hotel P&L. Tokens will repeat that story, only faster. Hospitality must now measure the cost of intelligence the same way it learned to measure the cost of distribution: per guest.

That measure has a name — TCPG, Token Cost Per Guest © — and the time to define it is now, before it defines us.

Executive Summary

For decades, hotel performance has been measured through RevPAR, ADR, Occupancy, GOPPAR, CPOR, payroll ratios, guest satisfaction, and reputation scores. These remain essential, but they were built for a world in which service was delivered primarily through people, rooms, physical assets, and departmental labor.

AI introduces a new layer of cost and capability. Every AI-generated response, translation, upsell, recommendation, review reply, summary, or report consumes a digital unit of work: tokens. Individually trivial; at enterprise scale, a measurable operating expense — exactly as OTA commission once was. Pertlink proposes a new metric:

TCPG = Total AI Token Cost ÷ Number of Guests Served

TCPG is more than a cost figure. It is a lens into the efficiency, discipline, governance, and maturity of AI-enabled hospitality. Just as OTA cost became the board-level measure of distribution, TCPG is set to become the measure of intelligence — and as AI embeds into operations, it will soon need to be incorporated into USALI, the Uniform System of Accounts for the Lodging Industry, whose 12th Revised Edition had an adoption date of January 1, 2026. ¹

The strategic question is no longer “Can AI improve hospitality?” but “What does AI-enabled service cost per guest, and what value does it create?” — a conversation the industry should begin at HITEC 2026.

1. The New Cost Layer in Hospitality

AI does not behave like payroll, a fixed software license, utilities, or marketing spend. It consumes compute, and in generative AI, that compute is priced in tokens — the units models use to process and generate language. Every prompt, reply, summary, translation, and recommendation has a token footprint.

The more AI is used across the guest journey, the more token consumption rises.

At a small scale, this is immaterial. But a resort, casino hotel, or international brand using AI across messaging, sales, loyalty, revenue, engineering, housekeeping, HR, procurement, marketing, and owner reporting can consume millions, even billions, of tokens. Leaders may know they are “using AI” without knowing how much is consumed, by which departments, through which vendors, on which use cases, or whether any of it improves the guest experience rather than simply creating digital noise.

Without measurement, AI becomes invisible. TCPG makes it visible.

2. Tokens Are the New Commission: The OTA Lesson

Hospitality has navigated a cost transformation like this before. When online travel agencies arrived, they were welcomed as a powerful new source of demand. Few operators initially treated commission as a strategic cost line — it was buried in distribution, absorbed as the price of incremental business, and rarely measured on a guest-by-guest basis.

Then the math caught up. As OTA dependence grew, commission became one of the highest and most scrutinized costs on the P&L, and the industry built a whole discipline around it: cost of acquisition, channel mix, net ADR, and the long campaign to win back the direct relationship. Commission moved from an invisible deduction to a board-level metric.

Artificial intelligence is about to repeat that pattern — only faster.

Tokens are to AI what commission is to distribution: the unit cost of doing business in a new channel.

Today, tokens look small and easy to ignore, scattered across subscriptions and vendor bundles. But as AI spreads at the speed of light, token consumption will compound, and a major variable cost will again be hiding in plain sight. The lesson of the OTA era is simple: what you do not measure, you cannot manage — and what you cannot manage eventually manages you. That is why Pertlink proposes asking every guest:

What did this guest cost to serve with intelligence?

3. Defining TCPG

Token Cost Per Guest is a hospitality AI operating metric that measures the direct cost of generative AI usage attributable to each guest served. If a hotel spends US$120 on tokens in a month and serves 2,000 guests, TCPG is US$0.06 per guest.

On its own, the number seems small; its value is in what it reveals. A low TCPG may signal efficient usage. A rising TCPG may reflect deeper engagement, expanding adoption, poor prompt design, or fragmented deployment. A high TCPG can be entirely acceptable if it drives revenue, speed, loyalty, or labor-productivity gains. TCPG is therefore a value-efficiency metric, not a cost-cutting metric: the goal is not to minimize it but to optimize it relative to measurable outcomes.

4. Why Hospitality Needs a New AI Metric

RevPAR, ADR, GOPPAR, CPOR, and NPS remain essential, but none measure the cost of digital intelligence. AI is becoming an operating layer that touches nearly every function — and without a metric like TCPG, its costs are scattered across IT subscriptions, messaging platforms, CRM, marketing, revenue systems, departmental SaaS budgets, vendor bundles, and shadow AI tools. That creates five risks:

1. Cost leakage — spend grows quietly across platforms without ownership.

2. Poor ROI visibility — management cannot connect AI spend to outcomes.

3. Vendor opacity — token usage is hidden inside bundled pricing.

4. Weak governance — departments experiment without a common framework.

5. Guest-value misalignment — tokens are spent on content that adds no value.

5. Applying TCPG Across the Business

TCPG can be measured at several levels, each answering a different question:

• Property — total token cost ÷ guests served; the headline operating benchmark.

• Department — front office, sales, marketing, revenue, engineering, HR, finance.

• Use case — per concierge chat, translation, review reply, RFP, upsell, recovery case.

• Journey stage — pre-arrival, in-stay, departure, post-stay, loyalty re-engagement.

• Segment — leisure, business, VIP, loyalty, group, MICE, direct vs. OTA guests.

Use-case TCPG is the most powerful view: not all AI is equal — some create revenue, some reduce friction, some are merely interesting. It tells leadership what to scale, refine, pause, or retire, and lets hotels align AI intensity with guest lifetime value.

6. TCPG, ROI, and Vendor Choice

TCPG should never be read in isolation. The essential question is what value is created per dollar spent per token — measured against upsell revenue, direct conversion, response and resolution times, review scores, productivity, and reduced escalation.

A hotel spending US$0.08 per guest on pre-arrival upselling that generates US$4.00 per guest in ancillary revenue does not have a cost problem — it has an investment. A hotel spending US$0.25 per guest on elaborate concierge content that few guests read should redesign that workflow. And let’s not forget the TCPG of a booking made via an LLM like ChatGPT.

Low TCPG is efficient. Productive TCPG is a strategy. Profitable TCPG is a transformation.

TCPG also reframes procurement. Instead of asking only “What is the monthly subscription?”, hotels should ask “What is the effective AI cost per guest served — and per useful outcome?” That moves vendor selection from feature comparison to economic comparison, which matters as the AI marketplace becomes crowded and over-promised.

7. TCPG as a Governance and Board Tool

AI governance usually centers on privacy, security, bias, compliance, and human oversight. It must also include financial discipline. TCPG creates a shared language: the CIO can explain AI usage operationally, the CFO financially, the GM in service terms, and the owner in terms of profitability and asset value. It belongs not only on the IT dashboard but on the executive one.

Boards do not need AI hype. They need AI economics. TCPG provides one of the first simple measures of it.

8. TCPG and USALI

For TCPG to be useful at the industry level, it will soon need to be incorporated into USALI. The framework is not static — the 12th Revised Edition, published by HFTP with AHLA and the Global Finance Committee, emphasizes greater transparency, broader data sets, sustainability, and new schedules.² AI token consumption is the next operational cost category that requires clearer treatment, rather than being buried within software, systems, marketing, or brand-technology charges — the very mistake the industry once made and corrected with OTA commission.

A possible USALI-aligned concept: AI Token Consumption Expense — a dedicated subcategory for variable generative AI usage costs across guest-facing, commercial, operational, and analytical workflows.

Tracked this way, TCPG would sit alongside RevPAR, ADR, GOPPAR, and CPOR, letting hotels compare AI operating efficiency across properties, brands, ownership groups, and asset classes — extending USALI’s common language for benchmarking to the cost of intelligence.

9. From Metric to Method

Adopting TCPG follows a clear, practical path that any property or group can begin now:

1. Register every AI tool in use — official systems, vendor modules, and shadow AI.

2. Classify use cases as guest-facing, commercial, operational, analytical, or experimental.

3. Track token costs, models, usage limits, and overage with vendors and internal platforms.

4. Establish a baseline TCPG, then break it down by department, journey stage, and use case.

5. Map TCPG to value — revenue lift, response time, review scores, productivity, escalation.

6. Build a dashboard showing cost, trend, and impact by vendor, department, and use case.

7. Optimize prompts and workflows to reduce waste without reducing guest value.

8. Benchmark across properties, brands, and markets to surface best practices.

As that discipline matures, hotels move up a five-stage ladder — from 

• Level 1 Invisible (AI used, cost untracked) to

• Level 2 Token-aware (variable cost understood),

• Level 3 Departmental (cost allocated by function),

• Level 4 Value-linked (cost set against outcomes), to

• Level 5 USALI-informed (structured reporting alongside core KPIs).

At Level 5, the hotel has moved from AI experimentation to an AI operating discipline.

10. The Human Value Stack

The purpose of AI in hospitality is not to remove humanity from the experience but to remove friction. AI should help staff see more clearly, respond faster, personalize more intelligently, and spend more time on moments that matter. Used poorly, it produces generic communication, hallucinated recommendations, weakened accountability, and service that feels efficient but soulless. TCPG helps leadership ask whether token consumption strengthens or weakens the human experience. The real question is not how many tokens were used, but:

Did those tokens help us deliver better hospitality?

11. The Time to Act Is Now: A HITEC 2026 Agenda

The window to get ahead of this is open, but narrow. AI adoption is accelerating at the speed of light; every month, more workflows are routed through generative AI, and more token costs are absorbed into budgets that no one is yet measuring. The longer the industry waits, the harder it becomes to retrofit discipline onto a cost layer that has already sprawled across departments, vendors, and properties. The OTA era showed what happens when a new cost is left unmeasured for too long. Tokens should not become the next lesson learned in hindsight.

There is no better venue to begin than HITEC. HITEC 2026 takes place June 15–18 at the Henry B. Gonzalez Convention Center in San Antonio, Texas — the world’s largest gathering of hospitality finance and technology professionals, produced by HFTP, the same body behind USALI. It is precisely the room where a new operating metric should be debated, pressure-tested, and refined.³

1. Agree on a common industry definition of Token Cost Per Guest and its variants.

2. Engage HFTP, AHLA, and the Global Finance Committee regarding the USALI treatment of AI token costs.

3. Establish early benchmarks across brands, markets, and asset classes.

4. Set vendor-transparency standards for token usage and pass-through pricing.

5. Frame TCPG as a value-efficiency metric anchored in the human experience.

The time to define TCPG is before it defines us.

12. Conclusion

AI will increasingly shape how hotels communicate, personalize, sell, operate, govern, and serve. As it grows more powerful, it must also grow more measurable. TCPG gives hospitality a simple, practical way to manage the economics of AI-enabled service — bringing visibility to the token layer, supporting governance, sharpening vendor choice, linking AI usage to guest value, and opening a path to USALI-informed reporting.

Just as the industry learned to measure the cost of distribution, it must now learn to measure the cost of intelligence. The future hotel will not be judged by how much AI it uses, but by how intelligently, responsibly, and humanely it uses AI to serve guests.

Token Cost Per Guest is not just a cost measure. It is a lens into the efficiency, discipline, and maturity of AI-enabled hospitality. The intelligence may be artificial, but the experience is human.

The concept of TCPG is envisioned by Pertlink.

Made with the help of AI tools, but with a HITL.

References

1. HFTP, AHLA and the Global Finance Committee, 12th Revised Edition of the Uniform System of Accounts for the Lodging Industry, adoption date January 1, 2026. Source: ahla.com.

2. 12th Revised Edition of USALI, HFTP / AHLA / Global Finance Committee. Source: ahla.com.

3. HITEC 2026 (HITEC North America), produced by HFTP, June 15–18, 2026, Henry B. Gonzalez Convention Center, San Antonio, Texas, USA. Source: hitec.org.