Gartner: Over 40% of Agentic AI Projects Will Be Canceled by 2027

More than 40% of agentic AI projects will be canceled by the end of 2027, according to Gartner, which also estimates that only about 130 of the thousands of agentic AI vendors in the market are real. MIT studied 300 deployments and found that 95% of pilots delivered no measurable P&L impact. Those numbers frame a growing fight inside hotel management agreements over who owns AI agents, who pays for them, and who keeps the savings.

The hesitation is earned, but the cost of waiting is steep. In the hotel industry specifically, 98% of owners and developers have started incorporating AI, yet only 32% report AI embedded across most operational areas, and 73% want to do more but are unsure where to start.

What does history say about going second?

When Marriott's SVP of sales first pitched Bill Marriott on revenue management in 1985, Marriott went first anyway. By the mid-1990s, revenue management was adding $150 to $200 million a year to Marriott's top line. Owners who waited for RM to be "proven" waited a decade — and nobody remembers who was second.

The brands aren't waiting this time either. Marriott confirmed corporate job cuts on October 28. On Hilton's Q3 call, CEO Chris Nassetta's first technology priority was system efficiency: using AI, technology and process change to cut corporate costs and the cost of running the system for owners, with RISE 2.0 arriving in the first quarter of 2027. The brands are going first on AI for their own P&L. Whether any of it reaches an owner's GOP line depends on what the operator does in the middle of the P&L.

Who holds the risk?

In a standard HMA, the owner pays a base fee on revenue whether the new technology works or not. The author's argument: "nobody wants to be first" really means "nobody wants to pay to be first" — so make the operator go first. His firm has offered owners deals where 100% of its compensation comes from GOP improvement over trailing twelve months. His test: any operator pitching AI that won't make that offer has told you what it thinks of its own agents.

Who employs your agents?

Hotels spend about 1.4% of total operating revenue on IT — and 35% on labor. Agents are not a software line; they're a labor line, coming for the 35%, not the 1.4%. Microsoft calls this "digital labor," and 82% of leaders are confident they'll use it to expand workforce capacity within 12-18 months. Its phrase is "agent boss" — someone who builds, delegates to and manages agents — and 28% of managers are considering hiring AI workforce managers for hybrid teams of people and agents. The author's company created an Agent Deployment Manager role in October.

Here's the structural problem. In 80-90% of current HMAs, the manager is legally the employer of the hotel's employees. If an operator's agents are rented from a software vendor, the operator isn't the employer of its digital workforce — it's a client of a staffing agency. Five consequences follow.

• You can't redesign a process you don't own. BCG puts AI value at 10% algorithms, 20% tech and data, 70% people and processes. McKinsey's AI winners are three times more likely to fundamentally redesign workflows; everyone else sprinkled AI on old processes, which is why only 39% see any EBIT impact, mostly under 5%. MIT found over half of GenAI budgets go to sales and marketing tools, yet the biggest ROI sits in back-office automation.
• The markup is extreme. Staffing agencies mark up labor 25-40%, up to 75%. Rented agents mark it up roughly 100x: tokens cost about $0.002 for a six-turn support conversation on a small model, or about $0.21 for a long sales conversation on a frontier model — yet Salesforce Agentforce charges $2 per conversation and Intercom's Fin charges $0.99 per resolution. Median SaaS gross margin is 77%; public SaaS companies spend a median 39% of revenue on sales and marketing. The author's sales agent runs on about $1,000 a month in tokens, billed to the owner at cost.
• Tokens deflate; seat prices don't. LLM inference cost is falling 10x a year — Stanford measured GPT-3.5-level inference falling 280-fold between November 2022 and October 2024. Whoever owns the agent captures that decline. A rented agent gets a renewal notice.
• The Uber question. Rideshare prices rose 92% between January 2018 and July 2021 once venture subsidies ended. AI pricing is already tightening: xAI raised top-tier pricing 140% in August, and Gartner predicts inference cost per agentic workflow will rise more than fivefold through 2028, because agents burn 5 to 30 times the tokens per task of a chatbot.
• Termination rights. Agents, prompts, learned workflows and the data lake are the next version of the employee-transfer fight — and nobody has written the clause. The author's fix: a canonical data lake the owner can take, owner-held trace logs and attribution telemetry, and agent portability negotiated with key money and termination fees, not after. Only 11% of operators have a fully integrated tech stack, and 91% still rely on manual reporting — an agent can't learn from data nobody can find.

The prediction: within three years, agent ownership and data portability will be standard clauses in every institutional HMA, the way key money and termination fees are today. Owners who ask first get the best terms; owners who wait inherit whatever the first movers negotiated.