Starting with the help users need
I self-initiated this work to explore what an AI-first product experience could mean beyond a prompt interface.
Rather than starting with a chatbot or a list of AI features, I worked backwards from the help users might need at different points in a workflow: understanding what changed, explaining why, anticipating what might happen next, deciding what to do and, in some cases, preparing an action for review.
That became a capability model for exploring AI inside structured enterprise workflows.
A capability model for AI assistance
| Capability | User question |
|---|---|
| Descriptive | What happened? |
| Diagnostic | Why did it happen? |
| Predictive | What might happen next? |
| Prescriptive | What could we do? |
| Agentic (bounded) | Shall I prepare this for you? |
| Generative | What would you like to create or explore? |
What happened?
Why did it happen?
What might happen next?
What could we do?
Shall I prepare this for you?
What would you like to create or explore?
The model is not a maturity ladder. More AI is not automatically better. Each level changes the requirements for data quality, transparency, trust and user control. The right level depends on the problem and the consequence of getting it wrong.
Capability is not the interface
A capability describes what the AI does. It does not prescribe how the user has to experience it.
A predictive capability, for example, could appear as a proactive insight, guided decision support, narrative interpretation or conversational query. The interaction pattern should follow the workflow rather than forcing every AI experience into a prompt.
- Proactive insightSurface something meaningful when the product detects it.
- Guided decision supportHelp people compare options and consequences inside a structured workflow.
- Narrative interpretationExplain what changed, why it matters and what may require attention.
- Conversational queryAsk an open-ended question about likely outcomes.
From framework to product experience
I used the model to explore AI across project and portfolio-management workflows, turning the strategy into concrete Diagnostic, Predictive, Prescriptive, Agentic and Generative product concepts and applying the same model across a broader opportunity map.
I developed the work into a 16-slide product and experience strategy and presented it to Altus leadership.
The protected case shows the original product concepts and strategy artefacts behind this work.
Continue with access
For a deeper look at the work, including additional screens, decisions and supporting evidence, enter your passcode or request access.