Most companies do not have an AI-idea problem.
They have too many ideas and no reliable way to choose the first one.
Ask a leadership team where AI could help and the list grows quickly:
Customer service.
Sales.
Finance.
Marketing.
Operations.
Reporting.
Knowledge management.
Contracts.
Forecasting.
Every department can identify tasks that are repetitive, slow, inconsistent, or frustrating. Many of those ideas are legitimate.
The hard part is not generating the list.
The hard part is finding the first workflow worth funding.
The right first initiative is valuable enough to matter, focused enough to finish, and owned by someone accountable for the result.
It must do more than create a good demonstration.
It must earn the right to the second workflow.
Do not choose the most impressive demo
The use case that looks best in a boardroom is not always the one that creates the best business result.
A broad conversational assistant can feel transformational because it can answer many kinds of questions.
A narrow operational workflow may look less exciting while changing the economics of a department.
The first workflow should not be chosen based on how futuristic it appears.
Choose it based on whether changing the work will matter.
At World Emblem, the first major opportunity was not a general-purpose company assistant. It was a production bottleneck processing roughly 1,000 custom design requests a day.
At a national food distributor, the first opportunity was not an enterprise chatbot. It was the purchasing decision around perishable inventory, where every wrong assumption had a deadline and a financial consequence.
The best first workflow is often hiding inside ordinary work with extraordinary economics.
Start with a workflow, not a department
A department is too broad.
"Transform customer service" can include dozens of workflows, systems, roles, channels, and outcomes.
A task may be too narrow.
"Summarize calls" may save time but leave the larger process unchanged.
The useful unit is a bounded workflow.
A workflow has:
- A trigger
- A clear outcome
- A group of people responsible for completing it
- Information and systems it depends on
- Decisions and exceptions
- A business measure that can improve
Examples include:
- Turn a qualified opportunity into a complete proposal
- Move a new customer from signed agreement to successful onboarding
- Review a vendor submission and make an approval decision
- Convert a product request into a production-ready design
- Investigate an operating exception and recommend the next action
- Prepare a management report from multiple systems
- Resolve a customer request from intake through closure
The boundary makes the opportunity understandable, buildable, and measurable.
Six criteria for the first workflow
1. Meaningful business value
The improvement should materially affect at least one important measure:
- Cost
- Capacity
- Cycle time
- Quality
- Revenue
- Risk
- Customer experience
- Management visibility
"Employees would like this" is not enough.
The team should be able to explain why changing the workflow matters to the business.
That does not require a perfect ROI model before discovery. It requires a credible economic hypothesis.
2. Repeated work
Frequency creates leverage.
A ten-minute improvement performed once a year is not a transformation.
A ten-minute improvement performed thousands of times can change capacity.
Repeated work also creates the production evidence needed to improve the system. The team can observe enough examples, measure quality, identify exceptions, and learn where AI creates or destroys value.
3. A committed owner
The first workflow needs a business owner, not merely an interested executive.
That person should own the outcome, bring the right users and subject-matter experts into the work, help make tradeoffs, reinforce adoption, and use the results to decide what comes next.
A strong technology team cannot substitute for absent business ownership.
The owner does not need to understand the models.
They need to care deeply about the result.
4. A practical path to production
The workflow must be accessible enough to change.
Can the team reach the required systems?
Can the necessary information be assembled?
Can real users participate?
Can the organization define the approvals and controls?
Can a useful first version launch without waiting for a company-wide data or platform program?
The right first workflow can contain complexity. It should not depend on solving every enterprise constraint before the first value appears.
5. A measurable result
The company should know what it will compare before and after launch.
Possible measures include:
- Time to complete the workflow
- Volume handled per employee
- Number of handoffs
- Rework rate
- Quality score
- Response time
- Cost per transaction
- Conversion
- Backlog
- Exception rate
- User adoption
- Customer satisfaction
A measure does not need to capture every benefit.
It needs to be credible enough to tell the organization whether the new way is working.
6. Leverage beyond the first workflow
The first initiative should stand on its own business case.
It should also teach the organization something valuable.
Will it establish a useful integration?
Will it clarify important business context?
Will it create a reusable approval pattern?
Will it develop evaluation methods?
Will it teach the team how to launch AI in production?
Will it reveal adjacent workflows worth changing?
Reuse potential should break ties between strong opportunities. It should not rescue a weak business case.
Avoid the two bad extremes
Too small to matter
A tiny task can be easy to launch and impossible to care about.
If the result saves a few employees several minutes but creates no meaningful change in capacity, cost, speed, quality, or customer value, the initiative may never build organizational momentum.
The first success should be visible enough that leaders and users believe the next investment is justified.
Too broad to finish
A company-wide assistant or enterprise-wide transformation can sound strategic while containing too many unresolved decisions.
Which users?
Which workflow?
Which information?
Which systems?
Which actions?
Which controls?
Which business result?
Without a useful boundary, the team spends months creating infrastructure and alignment before a real user receives value.
The first workflow should be narrow enough to launch, not narrow enough to be irrelevant.
Score the opportunity, then use judgment
A practical opportunity inventory can compare workflows across:
- Business value
- Frequency
- Pain
- Ownership
- Information readiness
- Integration complexity
- Risk
- Measurability
- User readiness
- Reuse potential
Scoring helps expose assumptions and compare unlike opportunities.
It does not make the decision automatically.
A workflow with the highest theoretical value may have no committed owner. Another may be slightly smaller but capable of reaching production quickly and creating strong reusable patterns.
The point of the scorecard is not mathematical certainty.
It is a better executive conversation.
Understand the current work before promising the future
Once the first candidate is selected, observe how the work actually happens.
Do not rely only on the official process diagram.
Sit with the people doing the work.
Ask them to show:
- Where the request arrives
- What they search for
- Which systems they open
- What they copy
- Which judgment they apply
- Which exception causes the most delay
- What they do when information is missing
- Who reviews the work
- Which output the next person needs
- How they know the process is complete
The highest-value opportunity is often not the obvious AI task.
It is the hidden coordination work between the visible steps.
At Ryder, the first transformation work began inside one business function. The team mapped how the process moved across people, information, systems, decisions, and handoffs before defining what AI should do.
At Motorola Solutions, the work began by mapping the complete support journey and deciding what to enhance, replace, or augment with AI before a large build investment.
Clarity is not the opposite of action.
It is how the company avoids building the wrong thing quickly.
Build the business case around the new workflow
The business case should describe the operating change, not just the technology.
Current state:
- How many people participate?
- How much time does the workflow require?
- How much volume moves through it?
- Where does work wait?
- Where does quality break?
- What does the current friction cost?
- What opportunity is delayed or lost?
Future state:
- Which steps disappear?
- Which work does the system prepare?
- Which decisions remain human?
- Which actions happen automatically?
- What becomes faster, more consistent, or more visible?
- Which new capacity or revenue becomes possible?
This is a more durable business case than a generic estimate of "AI productivity."
The first workflow has to earn the second
The first initiative should create three forms of evidence.
Business evidence
Did the measure improve?
User evidence
Did the new experience become part of the real work?
Organizational evidence
Did the company learn how to build, govern, launch, and improve AI in production?
When all three exist, the organization does not need another theoretical argument about AI transformation.
It has a working pattern.
The first workflow has earned the second.
Ten questions for leadership
Before funding the first workflow, ask:
- What outcome does this workflow produce?
- Why does changing it matter financially or operationally?
- How often does the work happen?
- Who owns the result?
- Which people and systems participate?
- Which steps should disappear before anything is automated?
- Which decisions require human judgment?
- What can launch without an enterprise-wide dependency?
- How will the result be measured?
- What will the organization be able to reuse or do better afterward?
A strong first workflow does not need to be the biggest AI idea in the company.
It needs to be the one most capable of becoming a real result.
