Direct answer
Inventory repeated information work, establish its volume and economic importance, score each workflow for value, feasibility, risk, and reuse, then validate the strongest candidate through direct observation. Prioritize a measurable workflow with an accountable owner over a vague company-wide idea.
Opportunity scorecard
Score one workflow before you chase a tool
Opportunity score
15 / 25Worth discovery
Map the workflow and find the weakest factor before committing to a build.
Discovery
Look where information moves but judgment stalls
AI is especially useful when people spend substantial time finding information, transforming it into a familiar format, applying repeatable criteria, or moving it to the next person. The system can handle more of the mechanical work while preserving the human decision.
Interviewing leaders produces strategic ideas. Shadowing employees reveals operational ones. Use both. Ask what customers wait for, what employees recreate, what managers compile, what experts answer repeatedly, and where work falls into an inbox without context.
- Queues and backlogs
- Repeated customer or employee questions
- Manual research and document review
- Copying between systems
- Reports assembled from several sources
- Leads or requests waiting for qualification
- Expert knowledge that does not scale
- Quality checks dependent on memory
Prioritization
Score value, feasibility, risk, and leverage separately
| Dimension | High score means | Evidence |
|---|---|---|
| Frequency | The unit of work repeats often | Weekly or monthly volume |
| Capacity | Meaningful time or delay can be reduced | Handling time, waiting time, backlog |
| Economic impact | The workflow affects revenue, margin, cost, quality, or risk | Financial or operational baseline |
| Information readiness | Required sources are accessible and reliable enough | Source inventory and sample review |
| Process clarity | Inputs, decisions, outputs, and exceptions can be mapped | Observed workflow and owner agreement |
| Risk | Failure is detectable, reversible, and controllable | Risk review and escalation design |
| Adoption | Users feel the pain and want improvement | User interviews and owner commitment |
| Reuse | Knowledge or infrastructure can enable other systems | Shared sources, integrations, or evaluation |
Portfolio
Use a portfolio matrix instead of a ranked list
Plot expected value against implementation difficulty, then mark risk and strategic reuse. Quick wins are useful for learning. Strategic foundations may deserve investment even when they are harder. High-effort, low-value ideas should leave the roadmap.
Do not let a high score hide a fatal dependency. A workflow may look attractive until the team learns that its data cannot be accessed, the decision cannot be evaluated, or the process owner cannot participate.
Questions to ask
Interview for evidence, not enthusiasm
- Show me the last three times this work happened.
- What arrives, who touches it, and what leaves?
- Where do you search, copy, wait, check, or start over?
- Which exceptions consume most of the effort?
- What would become possible if this took half the time?
- What mistakes matter, and how are they detected?
- Which information is sensitive or restricted?
- Who would own the improved workflow?
- What number would prove the change was worthwhile?
Before building
Validate the workflow with a one-page opportunity brief
Problem
Describe the unit of work and current friction in plain language.
Baseline
Record volume, handling time, delay, error, quality, conversion, and cost where relevant.
System hypothesis
State which work AI, automation, software, and people would each perform.
Dependencies
List sources, integrations, permissions, reviewers, and process changes.
Risk
Describe sensitive information, failure consequences, and escalation.
Evidence
Define the threshold that would justify scaling, and the result that would stop the project.
The value point
After this page, you should be able to decide:
Which workflow has the strongest combination of value, feasibility, control, adoption, and reuse.Your working output should be a workflow score, portfolio matrix, interview guide, and one-page opportunity brief.
Questions business leaders ask
Frequently asked questions
Which business processes are best for AI?+
Frequent information-heavy processes with measurable outcomes, accessible data, understandable decisions, and manageable risk are strong candidates. Searching, summarizing, classifying, comparing, drafting, routing, and follow-up often contain useful opportunities.
Should we prioritize time savings or revenue growth?+
Prioritize the strongest evidence. Time savings are easier to baseline but only create value when capacity is recaptured. Revenue opportunities can be larger but harder to attribute. A portfolio can include both.
How many opportunities should an assessment identify?+
Capture broadly, then narrow. A useful assessment may document dozens of ideas but should recommend a small first portfolio with one primary implementation and a short backlog of next candidates.
What makes an AI opportunity too risky?+
Risk rises when data is highly sensitive, decisions are consequential, failures are hard to detect or reverse, affected people cannot appeal, or accountability is unclear. High risk does not always mean impossible, but it requires stronger evidence and controls.
Research anchors
Primary and authoritative sources
Examples and planning ranges are clearly labeled. Source terms, provider behavior, and regulations can change; verify current requirements for your organization and jurisdiction.