Horizontal AI vs. Vertical AI
Most organisations begin with horizontal AI.
That means general-purpose systems such as ChatGPT, Claude, Gemini, Copilot, and similar tools. They are powerful because they can serve many users across many domains. They are accessible, flexible, and useful enough to change the way people work almost immediately.
But the same quality that makes horizontal AI useful also limits it.
It is built for everyone. That means it is not built around one specific business.
The missing business context
A general-purpose model does not naturally know how your organisation thinks. It does not know your commercial priorities, governance rules, brand standards, risk thresholds, service model, client expectations, internal decision logic, or history of what has and has not worked.
It can infer. It can approximate. It can respond fluently. But unless those rules and patterns are specified, the model is operating from general knowledge rather than institutional knowledge.
That is acceptable for low-risk tasks. It is not enough for production workflows where consistency, auditability, and business specificity matter.
What vertical AI changes
Vertical AI is built for a specific business, in a specific domain, under specific constraints. It does not rely on the model guessing what good means. It operates from the organisation’s knowledge, policies, workflows, decisions, and rules.
This is where the Intelligent Business Operating Model becomes important.
An IBOM is the structured operating layer that sits between a business and its AI. It turns business knowledge into machine-readable assets that agents can use. It defines what agents know, what they can do, what they cannot do, when they must escalate, and how their outputs are checked.
Horizontal AI supplies capability. The operating model supplies specificity.
Why prompts are not enough
A prompt can guide one interaction. A specification can govern repeatable work.
That difference matters. A prompt may produce a useful answer today, but it rarely captures the complete reasoning system needed for consistent operation. It does not reliably preserve context across sessions, people, clients, workflows, or changing business conditions.
A specification is different. It becomes a reusable business asset. It can be reviewed, versioned, validated, tested, improved, and transferred. It gives the organisation a stable foundation for agentic work.
The question is not whether the model is impressive. The question is whether the business has taught the system how to operate.
The practical route
The path from horizontal AI to vertical AI is not to abandon general-purpose models. It is to use them deliberately.
General-purpose AI can help surface knowledge, structure requirements, identify missing controls, test specifications, and accelerate construction. But the destination is not another chat interface. The destination is a governed AI capability built from the organisation itself.
That is the architectural shift: from tools that answer prompts to systems that execute governed business intent.
Read the white paper
This post is adapted from From NASA to AI Agents: The Evolution of Spec-Driven Development. You can also download the PDF.