The idea behind the book
What is vertical AI?
Vertical AI applies artificial intelligence to the knowledge, rules and workflows of a particular industry. Its value comes from helping a customer complete a task in that industry, with clear limits on what the system can do.
Start with the outcome
A customer rarely wants to operate your software. They want to change a booking, arrange a service or resolve a problem. An effective vertical AI workflow understands that request, checks the relevant business rules and works with the systems that hold the actual records.
In Vertical AI: Changing the Game, Clive Mayhew calls this an operating layer: the part of the business that coordinates work between a customer’s intent and a completed outcome. An answer alone does not complete the transaction.
How is it different from a general chatbot?
| Question | General chat experience | Vertical AI workflow |
|---|---|---|
| What does it know? | A broad range of subjects. | The industry’s terminology, current business rules and relevant customer context. |
| What can it access? | Information made available in the conversation or connected tools. | The specific booking, service or record systems needed for the task, within approved permissions. |
| What counts as success? | A useful response. | A verified outcome, or a clear handoff when the task cannot be completed. |
| Who handles exceptions? | The user may need to find another service. | A named person or team receives the context and takes responsibility. |
These are design choices, not labels that guarantee performance. A general model can power a specialised workflow. A chatbot with an industry name may still have no authority to act.
A golf booking example
Imagine a member says: “Add my guest to Saturday’s booking and reserve another cart.” The workflow needs to identify the right booking, check guest eligibility, check space and cart availability, explain any charge and obtain any required confirmation. Only then should it update the connected records and confirm the result.
If the cart is unavailable or a guest restriction applies, the system must explain that limit and offer an allowed alternative or a staff handoff. Saying “done” before the booking system confirms the change creates another service problem.
This is an illustrative workflow, adapted from The Pro Shop Moment. It is not a claim about a particular club’s live integration.
Four things leaders need to build
- Domain knowledge: the vocabulary, policies and exceptions that determine a correct answer.
- Connections: access to current records and the systems where actions happen.
- Authority: clear rules for what can be read, proposed, approved and changed.
- Evidence: confirmation of the completed action, an audit trail and measures of customer outcomes.
The same discipline applies beyond golf. A hotel guest request has different rules, but it still needs a trustworthy path from intent to outcome.
Where should you begin?
Observe a repeated customer request before choosing a tool. Record where staff copy information, explain an exception or repair a failed self-service journey. Then choose one bounded workflow to build and test yourself.
The Front Door Audit is a practical starting exercise. The goal is to understand the work closely enough to make better decisions about what AI should carry and where people should remain in control.