“Agent Washing” Why Most AI for Business Isn’t What It Claims to Be

The AI hype cycle is doing what hype cycles do: collapsing meaningful distinctions in the rush to sound relevant. The word "agent" which has a specific technical meaning is now being applied to rule-based chatbots, simple automations, LLM-powered FAQ tools and, occasionally, actual autonomous systems that can plan and execute multi-step tasks without human intervention.

For wellness operators, spa owners, hospitality businesses and studio managers trying to make smart technology decisions, this creates a real problem. You can't evaluate something you can't accurately name. And you can't protect your business from bad investments if the vocabulary being used to sell you something has been deliberately blurred.

This piece is a plain-language guide to what the spectrum actually looks like, what is genuinely viable for operators right now, and how to separate signal from noise when vendors pitch you "AI agents."

What is a chatbot?

A chatbot is a rule-based or intent-driven system that simulates conversation by matching user inputs to pre-written responses. Early versions followed strict decision trees. More recent "AI-powered" chatbots use large language models to understand natural language more flexibly, but the fundamental model remains reactive: a user sends a message, the chatbot retrieves or generates a response, and the loop ends there.

Chatbots cannot leave the conversation window. They cannot take action in external systems unless specifically hard-coded to do so via a webhook. They do not learn from interactions in real time. They cannot plan across multiple steps or adapt based on outcomes.

What is an automation?

An automation (as in Klaviyo flows, Make scenarios, Zapier zaps) is a triggered workflow a sequence of pre-defined steps that executes when a condition is met. A guest checks out → a post-visit email is sent. A membership lapses → a re-engagement sequence begins. An appointment is booked → a reminder fires 24 hours prior.

Automations are deterministic. Every input produces the same output. They are not AI in any meaningful sense, though AI tools can be integrated into them. They are enormously valuable arguably the single highest-ROI technology available to small wellness operators but calling them "agents" is simply wrong.

What is a true AI agent?

A true AI agent is an autonomous system that can perceive a goal, plan the steps required to achieve it, take action across systems, evaluate outcomes and adapt its approach. According to MIT Sloan's 2025 analysis, agents are "autonomous software systems that perceive, reason, and act in digital environments to achieve goals on behalf of human principals, with capabilities for tool use, economic transactions, and strategic interaction."

A true agent can cross system boundaries without being hard-coded to do so. It can use APIs, read and write data, make conditional decisions in real time, and continue a multi-step task without a human in the loop at each step.

Capability	Chatbot	Automation	AI Agent
Responds to user input	✓	—	✓
Executes pre-defined workflows	—	✓	✓
Plans multi-step tasks	—	—	✓
Acts across external systems	—	Limited	✓
Adapts based on outcomes	—	—	✓
Operates without human prompts	—	✓	✓
Learns from interactions	—	—	✓

What "agent washing" looks like in practice

Agent washing is when a product usually a chatbot or a simple automation with a conversational interface is marketed as an "AI agent" to capitalise on the term's growing appeal. Typical patterns include:

• A FAQ chatbot described as an "enquiry agent" that "handles customer requests autonomously"
• A Zapier workflow with a GPT step called an "AI agent for bookings"
• A booking widget with a chat interface marketed as a "concierge AI agent"
• A sequence of pre-written email templates triggered by time called an "AI-driven guest journey"

None of these are wrong to use several are genuinely useful. The problem is misrepresentation, because it leads operators to buy things at the wrong price, with the wrong expectations, and to miss the systems that would actually deliver the outcomes they were promised.

What is actually viable for wellness operators right now?

As of 2026, the technology that delivers the best return for most wellness operators is not agentic AI it is well-configured automation combined with well-designed communication sequences. The following are real and deployable today:

• Post-visit email and SMS sequences triggered by booking system events
• Rebooking windows with personalised prompts based on treatment history
• Lapsed-client re-engagement flows with conditional branching
• Enquiry handling via a well-prompted LLM chatbot integrated with your booking tool
• Review response agents that draft responses for human approval
• CRM enrichment from social and booking data via Make or Zapier

True agentic systems ones that can autonomously handle a guest enquiry end-to-end, cross-check availability, make a provisional booking, send a confirmation and follow up are becoming viable, but they require proper data infrastructure, governance frameworks, and human oversight protocols. Deploying them without those foundations is high-risk.

The governance question nobody is asking

One of the underrated risks of agent washing is that it encourages operators to deploy systems without asking the questions that matter: What happens when the agent makes a mistake? Who is accountable? How does the business catch errors before they reach guests?

MIT Sloan's 2025 research found that in enterprise AI agent deployments, 80% of implementation work was consumed by data engineering, stakeholder alignment, governance and workflow integration not by the AI model itself. For smaller operators without dedicated technical teams, the governance burden is proportionally even higher.

The right framing for wellness operators

The opportunity in AI for wellness is real. It is just not the opportunity most vendors are pitching. The question to ask is not "how do I deploy an AI agent?" it is "where in my guest journey is there a decision or action that is currently manual, time-sensitive and high-impact enough to justify automation or agentic design?"

That question leads to very different and very useful answers. And it starts with understanding exactly what kind of system you actually need.