Most organizations begin with the question “Where can we use AI?” That question produces long lists of use cases and very little clarity. A more useful starting point is the outcome you are trying to improve, the friction in the journey that produces it, and the point where human judgment and AI capability together change the result.
Use-case workshops usually start from the technology. People are asked what could be automated, summarized or generated, and the answers reflect what the tools happen to do well today. The result is a catalogue of interesting experiments with no shared logic for which ones matter.
Interesting and valuable are not the same thing. A use case can work beautifully in a demo, save a real person real minutes, and still leave the surrounding journey exactly as slow, inconsistent or frustrating as before. The improvement is absorbed by the next handoff, the next approval, the next wait.
Making one task faster is not the same as improving a journey, and improving a journey is not automatically the same as creating business value. Each of those is a separate step, and discovery work that skips them ends up prioritizing by enthusiasm rather than by outcome.
AI opportunity discovery works better when it begins where value is actually created or lost: in the customer, employee or business journey. You look at what that journey is supposed to achieve, where it breaks down, and only then ask what people, AI and automation should each contribute.
The discovery logic
Every discovery conversation should be able to name the outcome it is improving: a decision, an experience, a cycle time, a quality level, a cost, a risk. Without that, prioritization has nothing to prioritize against.
Rework, waiting, duplicated effort, repeated clarification. Friction is usually visible to the people inside the journey long before it appears in any dashboard.
Context that exists at one step and is missing at the next is one of the most reliable indicators of an AI opportunity, because it is a problem of retrieval, synthesis and continuity rather than speed.
Difficult decisions are slow, inconsistent, or made without the evidence people need. AI can often improve the inputs to a decision even where it should not make the decision.
Some of the strongest opportunities are not efficiencies at all. They are things the organization would like to offer customers or employees but currently cannot do at scale.
Mapping where judgment, relationship and accountability are essential is part of discovery, not a constraint on it. It tells you which opportunities are about augmentation rather than automation.
The question is not “what can AI do here?” but “what should this journey achieve, and what would have to change for it to achieve it?”
In the places where your journeys already lose value: where information does not travel, where decisions are slow or inconsistent, where people spend time reconstructing context, and where customer or employee needs go unmet because meeting them at scale was previously impossible. Those points are found by examining journeys, not by surveying available tools.
Identify them from the journey and the outcome it should produce. Prioritize them on the value of the outcome, the strength of the evidence that the friction is real, the feasibility of changing the work around it, and whether the organization can actually absorb the change. An opportunity that no team can adopt is not a priority, however elegant the technology.
With business problems, expressed as outcomes. Use cases are a useful vocabulary later, when you are describing how a specific opportunity will be addressed. Started with, they anchor the whole program to the current capabilities of current tools, which is the shortest path to a portfolio that ages badly.
Ask what changes outside the task. A valuable opportunity changes a decision, an experience, a cycle time or a risk that someone outside the immediate team can recognize. An interesting one improves a step whose gains are reabsorbed by the rest of the journey. Both can be worth learning from, but only one belongs in a value case.
Because they show the whole chain rather than isolated tasks. A journey view makes visible where value is created, where it leaks, and which moments actually determine the outcome. It also reveals opportunities that no individual team would raise, since the friction lives between functions rather than inside one.
Automation suits work that is stable, rule-based and repeatable. AI suits work that involves language, synthesis, pattern recognition and drafting under ambiguity. Humans should keep the judgment calls, the relationships, the accountability and the decisions where being wrong is costly. Discovery should make those boundaries explicit rather than leaving them to be discovered in production.
The shift is not a new method for its own sake. It changes what gets discussed, what gets funded, and what leaders can defend when the AI budget comes under scrutiny.
The pipeline gets shorter and more defensible, because every item is attached to an outcome someone owns.
Teams stop arguing about which tool is more impressive and start comparing opportunities on the same terms.
Journey-level discovery surfaces friction that sits between departments, where individual teams have no mandate to act.
Decisions about what stays human are made deliberately at design time, not implicitly by whatever the tool defaults to.
Because the claimed benefit is traceable to a journey and an outcome rather than to an estimate of minutes saved.
Discovery output feeds directly into work design, adoption and value realization instead of stopping at a slide.
If you are shaping an AI portfolio, a discovery process, or a value case that has to hold up in front of a board, I am glad to think it through with you. This work also runs as a facilitated discovery workshop, and I speak on the topic.
Or write directly to sara@bermudez-t.com.