Don’t Start With AI. Start With the Patterns of Work.
How Pattern AI Can Help Close the Growing Gap Between AI Adoption and Business Value
Executive Summary
Organizations do not appear to have an AI adoption problem.
They increasingly have an AI-to-value problem.
McKinsey's 2025 global survey found that 71% of respondents said their organizations regularly used generative AI in at least one business function. Yet more than 80% said their organizations were not seeing tangible enterprise-level EBIT impact from generative AI. (McKinsey & Company)
That contrast deserves attention:
71% regularly using GenAI
80%+ not seeing tangible enterprise-level EBIT impact
This does not mean AI is failing. McKinsey also found cost reductions within individual functions and reported that 17% of respondents attributed at least 5% of organizational EBIT to generative AI. The more important point is that adoption and enterprise-wide value are not the same thing. (McKinsey & Company)
The evidence increasingly points toward what separates them.
McKinsey tested 25 organizational attributes and found that workflow redesign had the strongest relationship with reported EBIT impact from generative AI. Yet only 21% of respondents whose organizations used GenAI said their organizations had fundamentally redesigned at least some workflows. (McKinsey & Company)
The evidence has strengthened. In 2026, McKinsey reported that leaders were 5.3 times more likely to report enterprise value capture when workflows were redesigned rather than left unchanged — 32% versus 6%. (McKinsey & Company)
BCG has identified a similar execution gap. Its July 2026 CEO research found that nearly nine in ten CEOs reported some cost or revenue benefit from AI in targeted areas. Yet only 14% clearly defined P&L impact for all AI initiatives, fewer than one-third funded process and skills redesign, and high performers were roughly seven times more likely to redesign workflows and reshape the business end-to-end with AI. (BCG)
Taken together, these findings suggest a hypothesis:
The next enterprise AI challenge may be less about obtaining AI capability and more about discovering where that capability belongs within work.
That is the opportunity I see for Pattern AI.
Its proposition is simple:
WORK → PATTERN → FRICTION → MATCH → REDESIGN → OUTCOME
Instead of beginning with:
“Where can we use AI?”
Pattern AI begins with:
“What patterns repeatedly occur in our work, where do those patterns create friction, and where could AI improve the outcome?”
That small change in starting point could make a large difference.
1. The AI Adoption-to-Value Gap
The first phase of generative AI was dominated by a question:
What can this technology do?
The next question was:
How do we get people to use it?
Organizations responded with enterprise licenses, copilots, experimentation programs, AI academies, prompt-engineering workshops, innovation challenges and hundreds of pilots.
Adoption accelerated.
But adoption is an activity.
Business value is an outcome.
The difference matters.
Imagine a company in which 10,000 employees begin using an AI assistant.
Employees draft emails faster.
Meetings are summarized automatically.
Presentations take less time to prepare.
Reports are easier to produce.
All of those gains may be useful.
But the company may still operate with essentially the same:
processes, handoffs, approvals, roles, bottlenecks and decision structures.
AI has made individual tasks faster without necessarily changing the economics of the larger system.
McKinsey describes this distinction clearly. Its 2026 research argues that enterprise value comes from redesigning how work gets done, not simply giving employees access to AI tools. (McKinsey & Company)
This suggests an important progression:
AI Access → AI Use → AI Integration → Workflow Redesign → Business Outcome
Many organizations have made significant progress on the left.
The larger opportunity increasingly lies toward the right.
2. Why “Where Can We Use AI?” May Be the Wrong Question
Imagine an organization buys an enterprise AI platform.
Management asks every department:
“Find five AI use cases.”
HR finds five.
Finance finds five.
Marketing finds ten.
Operations finds another fifteen.
IT creates twenty.
Within weeks the organization has an impressive spreadsheet containing dozens of AI opportunities.
But consider what has happened.
The organization started with the technology and went looking for problems that fit it.
The resulting portfolio can easily become:
AI → Idea → Pilot → Demonstration → ?
The final question mark is where value frequently becomes difficult.
Pattern AI reverses the direction:
Work → Pattern → Friction → AI Match → Redesigned Work → Outcome
AI comes later.
The work comes first.
3. The Pattern AI Hypothesis
Every organization contains thousands of tasks.
But underneath those tasks are recurring patterns.
People repeatedly:
search
retrieve
compare
classify
match
summarize
generate
predict
check
prioritize
route
monitor
recommend
coordinate
decide
Consider two seemingly unrelated professions.
A recruiter screens candidates.
A procurement manager evaluates suppliers.
Their domains are different.
But underneath the surface, both may perform:
Requirements → Options → Compare → Rank → Select
Similarly, a professor evaluating student work and an insurance employee reviewing a claim might both perform:
Input → Criteria → Compare → Exception → Judgment
This leads to the central Pattern AI proposition:
Recurring work patterns can serve as an index for discovering AI opportunities across otherwise very different functions.
Patterns do not tell us automatically that AI should be deployed.
They tell us where to look.
4. Example: Recruitment
Consider recruitment:
Requirement → Applications → Screening → Shortlisting → Interviews → Selection → Onboarding
At first glance, this is an HR process.
Look at it through a Pattern AI lens and different structures appear.
Application screening might involve:
Read → Extract → Compare → Classify → Rank
Scheduling involves:
Constraints → Match → Coordinate → Confirm
Candidate communication involves:
Context → Generate → Personalize → Review → Send
Now ask the second question:
Where is the friction?
Perhaps recruiters do not struggle with writing emails.
Perhaps the real bottleneck is manually comparing hundreds of applications against mandatory requirements.
The opportunity is therefore not simply:
“Use generative AI in recruitment.”
It becomes:
High-volume comparison pattern + significant human effort → investigate AI-assisted candidate analysis.
Now we can define outcomes:
Screening time ↓
Time-to-interview ↓
Consistency ↑
Recruiter capacity ↑
Quality of shortlist → measured
The technology is attached to a business hypothesis.
5. Example: Customer Service
A customer-service workflow might look like:
Query → Understand → Retrieve → Respond → Resolve/Escalate
Its underlying patterns include:
Classification — What kind of problem is this?
Retrieval — Where is the relevant information?
Generation — How should the response be communicated?
Routing — Where should an exception go?
Judgment — Can the issue be resolved automatically?
A technology-first approach might conclude:
“We need an AI chatbot.”
Pattern AI asks something different:
Where does the greatest friction occur?
Suppose agents are reasonably fast at communicating with customers but lose substantial time searching across multiple knowledge repositories.
The pattern becomes:
Question → Intent → Knowledge Search → Evidence → Response
Now retrieval or a RAG architecture may become appropriate.
Notice what happened.
We didn't start with RAG.
We discovered the work pattern that made RAG relevant.
That distinction is fundamental.
6. Example: Finance
Consider invoice processing:
Receive → Extract → Validate → Match → Identify Exception → Approve → Pay
Its patterns include:
Extraction
Comparison
Rule checking
Exception detection
Routing
For illustration, imagine that a large majority of invoices are routine while a smaller proportion contain exceptions requiring expert attention.
Why should humans spend equal attention on both?
A redesigned process might become:
Invoice → AI-assisted extraction → Matching → Confidence/Exception Check
Then:
Routine → Straight-through processing
Exception → Human review
The objective isn't necessarily to eliminate the employee.
It is to redirect scarce human attention.
That suggests another important Pattern AI principle:
AI value may often come not from replacing an entire workflow, but from changing where human attention is applied within it.
7. Example: Education
Consider faculty work.
A faculty member may:
prepare a course, analyze a syllabus, create assessments, prepare question papers, evaluate responses, give feedback and maintain academic documentation.
At the task level these appear different.
At the pattern level:
Course planning
Syllabus → Concepts → Outcomes → Sequence → Learning Plan
Assessment
Outcome → Cognitive Level → Question → Evaluation Criteria
Feedback
Student Work → Criteria → Compare → Gap → Feedback → Improvement
Now AI opportunities become visible.
AI might assist with extracting syllabus structure, mapping learning outcomes, generating draft assessments, comparing student work with criteria, or drafting formative feedback.
But the human remains responsible for academic judgment.
The objective therefore changes from:
“Teach faculty how to use ChatGPT.”
to:
“Help faculty discover where AI can improve the patterns through which academic work gets done.”
AI enters the workflow because its purpose has become visible.
8. Example: Manufacturing
Consider maintenance.
Traditional reactive maintenance follows:
Failure → Diagnosis → Repair
But sensor-rich manufacturing environments can potentially reveal another pattern:
Signals → Deviation → Risk → Intervention
Historical data may reveal combinations of signals associated with equipment problems.
Where appropriate data and predictive performance exist, the workflow can move toward:
Operational Data → Pattern Recognition → Risk Estimate → Human Decision → Preventive Action
The business outcome isn't:
“We implemented AI.”
It is something measurable such as:
Unplanned downtime ↓
That is the difference between an AI implementation metric and a business outcome.
9. Example: Sales
Consider a sales workflow:
Lead → Research → Qualify → Prioritize → Contact → Follow Up → Convert
The patterns include:
Retrieval → Classification → Ranking → Generation → Monitoring
A superficial AI opportunity might be:
“Use AI to write sales emails.”
That can save time.
But Pattern AI asks:
Where is the economically important friction?
Perhaps salespeople spend substantial time researching prospects who have little likelihood of converting.
Then the higher-value opportunity might be:
Lead Data → Pattern Analysis → Prioritization → Human Attention
The scarce resource isn't the AI model.
It is the salesperson's:
time + attention
This illustrates why finding the pattern before choosing the AI intervention matters.
10. Pattern AI Opportunity Discovery
These examples suggest a repeatable methodology.
I call it the:
PATTERN AI OPPORTUNITY LOOP
1. WORK
Begin with what people actually do.
Not the AI platform.
Not the organizational chart.
Not the latest model.
Map the real workflow.
Ask:
What enters?
What happens?
Who participates?
Where are the handoffs?
What comes out?
2. PATTERN
Now ask:
What keeps happening?
Strip away domain terminology.
“Screen applications” may contain:
Extract → Compare → Classify
“Review contracts” may contain:
Retrieve → Compare → Detect Exception → Judge
“Prepare lessons” may contain:
Retrieve → Sequence → Generate → Evaluate
Pattern recognition creates a common language across work.
3. FRICTION
A repeated pattern is not automatically an AI opportunity.
Look for friction.
Where does the organization lose:
Time?
Money?
Quality?
Consistency?
Speed?
Capacity?
Attention?
Opportunity?
This step is critical.
Without friction, AI experimentation can easily become technology tourism.
11. MATCH — Only Now Bring in AI
Once the pattern and friction are understood, ask:
What form of intelligence might improve this?
| Work Pattern | Potential AI Capability |
|---|---|
| Repeated searching | Semantic search / retrieval |
| Repeated summarizing | Generative AI |
| Repeated classification | ML / LLM-assisted classification |
| Repeated comparison | AI-assisted analysis |
| Repeated prediction | Predictive models |
| Repeated generation | Generative AI |
| Repeated matching | Recommendation / matching |
| Repeated monitoring | Detection systems / agents |
| Repeated coordination | Workflow automation / agents |
| Repeated knowledge retrieval | Enterprise search / RAG |
The important word is potential.
Pattern AI is not:
Pattern detected → Automate.
It is:
Pattern detected → Opportunity identified → Intervention tested.
12. REDESIGN — The Step That Changes the Economics
This may be the most important part of the framework.
Putting AI into an inefficient process can simply create:
a faster inefficient process.
The workflow itself may need to change.
The research is striking here.
McKinsey's 2025 survey found that, among 25 organizational attributes tested, workflow redesign had the strongest relationship with EBIT impact from GenAI. (McKinsey & Company)
Its 2026 research found leaders were 5.3 times more likely to report enterprise value capture when workflows were redesigned — 32% versus 6% when they remained unchanged. (McKinsey & Company)
BCG's July 2026 CEO research points in the same direction: high performers were approximately seven times more likely to redesign workflows and reshape the business end-to-end with AI. (BCG)
That allows us to distinguish two levels of AI intervention.
AI Augmentation
How can AI help someone perform the existing task?
versus:
AI Transformation
Why does this workflow need to operate this way at all?
Both matter.
But the second is where much larger organizational value may emerge.
13. OUTCOME — Close the Loop
Every Pattern AI opportunity should eventually answer:
What changed?
Not merely:
How many employees used AI?
How many prompts were generated?
How many pilots were launched?
How many agents were deployed?
Instead:
Cycle time ↓
Cost ↓
Errors ↓
Waiting ↓
Revenue ↑
Capacity ↑
Quality ↑
Decision speed ↑
Customer satisfaction ↑
Employee time returned ↑
BCG's 2026 research specifically warns that metrics such as numbers of AI users or automated tasks reveal relatively little about financial value. It recommends defining the expected P&L impact, baseline, KPIs, ownership and value logic before deployment and continually testing those assumptions against results. (BCG)
The full loop therefore becomes:
WORK → PATTERN → FRICTION → MATCH → REDESIGN → OUTCOME ↻
The final arrow is important.
Outcome becomes evidence.
Evidence informs the next decision.
Pattern AI becomes a learning loop, not a one-time AI implementation.
14. From AI Use Cases to an AI Opportunity Portfolio
Most organizations don't suffer from a shortage of AI ideas.
They need a way to determine which ideas deserve investment.
Pattern AI could transform a conventional AI use-case list into an:
AI Opportunity Portfolio
Each opportunity can be represented through eight questions:
| Dimension | Question |
|---|---|
| Workflow | Where does this happen? |
| Pattern | What repeatedly occurs? |
| Friction | What value is being lost? |
| AI Match | What capability could help? |
| Intervention | What exactly will change? |
| Human Role | Where must judgment remain? |
| Metric | What outcome will demonstrate value? |
| Evidence | What actually happened? |
Opportunities can then be evaluated against:
VALUE × FEASIBILITY × EVIDENCE
This produces an important management discipline.
A technologically spectacular autonomous agent may deserve a lower priority than a relatively simple AI-assisted comparison process that can save thousands of hours.
Therefore:
The best AI opportunity is not necessarily the most sophisticated AI application.
It is the intervention that produces meaningful value at acceptable cost, risk and complexity.
15. Why This Matters for AI Strategy
BCG's 2026 research provides an interesting warning about experimentation.
It found that 64% of surveyed CEOs said their companies pursued AI pilots, while only 26% embedded AI within a broader business transformation. (BCG)
That difference is revealing.
A pilot asks:
Can AI do this?
Transformation asks:
If AI can do this, how should our work change?
Pattern AI adds another question before both:
Is this the right part of the work to change?
The sequence becomes:
Discover → Prioritize → Experiment → Redesign → Measure → Scale
This is fundamentally different from:
Buy → Train → Prompt → Pilot → Hope
16. Pattern AI Changes AI Training
Enterprise AI training often starts with the technology.
What is generative AI?
How does ChatGPT work?
How do you write prompts?
How do you use Copilot?
Those capabilities remain useful.
But consider reversing the sequence.
Ask participants to bring one actual workflow.
Then:
Map the work
↓
Find the pattern
↓
Locate friction
↓
Identify possible AI intervention
↓
Learn the AI capability needed
↓
Redesign the work
↓
Measure the outcome
Now employees are not learning AI in the abstract.
They are learning it in the context of accomplishing something.
That may create a much stronger reason for adoption.
17. Pattern AI Changes Consulting Too
Imagine beginning an AI consulting engagement without asking:
“Which AI platform are you considering?”
Instead ask:
“Show me how your work actually flows.”
That creates the possibility of a different kind of engagement:
Pattern AI Opportunity Discovery
Teams bring actual workflows.
Together they:
Map Work
↓
Discover Patterns
↓
Find Friction
↓
Match Intelligence
↓
Redesign Work
↓
Define Outcomes
The deliverable isn't a 100-slide presentation explaining AI.
It is a portfolio of testable, prioritized AI opportunities tied to work and outcomes.
This moves consulting from:
AI Evangelism
toward:
AI Opportunity Engineering
18. Human Intelligence Remains Inside the Pattern
Pattern AI should not imply that every recognizable pattern should be automated.
Some patterns contain exactly the human capabilities organizations need to preserve.
Judgment
Empathy
Accountability
Context
Ethics
Negotiation
Leadership
Trust
BCG's workforce research describes a progression from individual tool-based AI adoption toward workflow transformation and eventually agent-led orchestration. As AI assumes more execution, human work increasingly shifts toward strategy, design, oversight, problem framing, systems thinking and judgment. (BCG)
The Pattern AI question therefore isn't simply:
Can AI perform this pattern?
It is:
Which part of this pattern belongs to AI, and which part should remain meaningfully human?
That can create arrangements such as:
Machine recognizes → Human judges
Machine generates → Human evaluates
Machine monitors → Human intervenes
Machine retrieves → Human decides
Machine handles routine → Human handles exception
AI transformation therefore isn't merely automation.
It is the:
redistribution of intelligence across the workflow.
19. The Evidence for an Outcome Era Is Growing
BCG's broader AI maturity research provides another useful perspective.
Its study of 1,250 companies worldwide found that only 5% generated measurable AI value at scale, defined in terms including revenue, cash-flow and process/workflow improvements. Another 35% were scaling and seeing some returns, while 60% were not achieving material value. (BCG)
Importantly, those figures describe BCG's broader company-level AI maturity study; they should not be interpreted as measuring only technology departments.
BCG's analysis of the tech function also found that leading companies focus AI investment on outcomes rather than outputs and reinvent core processes rather than simply adding automation. (BCG)
Put this beside McKinsey's evidence:
GenAI adoption is widespread.
Enterprise-wide financial impact remains much less widespread.
Workflow redesign is strongly associated with greater value capture.
And a larger pattern begins to emerge.
20. From the AI Adoption Era to the AI Outcome Era
We can think about enterprise generative AI development in three broad questions.
Era 1 — Capability
What can AI do?
Era 2 — Adoption
How do we get people to use AI?
Era 3 — Outcomes
Where exactly does AI change the economics of our work?
The eras overlap.
Organizations will continue to need better models, better infrastructure, adoption, governance and AI literacy.
But the center of gravity appears to be moving.
The management challenge is increasingly not simply:
Can AI do something useful?
It is:
Can we systematically discover, redesign and measure the places where AI changes work?
That is a different problem.
And it requires a different lens.
21. Pattern AI as That Lens
Pattern AI does not need to compete with ChatGPT, Claude, Gemini, Copilot, RAG, agents, machine-learning platforms or enterprise AI systems.
Those technologies provide capabilities.
Pattern AI addresses the question that precedes capability selection:
Where should intelligence enter the workflow?
Its proposition becomes:
PATTERN AI
Discover where AI creates value.
Work → Pattern → AI Opportunity → Outcome
This can develop into a practical family of methods:
Pattern AI Opportunity Discovery
Pattern AI Opportunity Canvas
Pattern AI Opportunity Portfolio
Pattern AI Workflow Redesign
Pattern AI Outcome Measurement
and ultimately:
Pattern AI Opportunity Engineering
22. The Pattern AI Opportunity Canvas
A manager should be able to describe an AI opportunity on one page.
A simple canvas might ask:
WORK
What are we trying to accomplish?
PATTERN
What repeatedly happens?
FRICTION
Where are we losing time, money, quality, capacity or opportunity?
MATCH
What AI capability might help?
REDESIGN
How would the workflow change?
HUMAN
Where is human judgment required?
OUTCOME
What measurable result should improve?
EVIDENCE
What happened when we tested it?
That turns Pattern AI from an interesting concept into an operating discipline.
23. The Bigger Opportunity: Teaching Organizations to See
There may be an even more significant implication.
AI capability is increasingly accessible.
But understanding how work really happens remains deeply contextual.
Frontline employees know things consultants, executives and AI models often do not.
They know:
where customers wait,
where approvals stall,
which spreadsheet gets copied manually,
which report nobody reads,
where information repeatedly disappears,
which exceptions consume entire afternoons,
where employees search for the same information every week,
and
where human judgment genuinely matters.
These are not merely inefficiencies.
They are patterns waiting to be discovered.
Once employees learn to see them, AI opportunity discovery can become distributed throughout the organization.
That is potentially far more powerful than having a small central AI team search for use cases.
24. A New Management Capability: Pattern Thinking
Pattern AI therefore leads naturally to something broader:
Pattern Thinking
Instead of immediately asking:
“Which AI tool should I use?”
people learn to ask:
What repeats here?
What varies?
What enters the system?
What transformation happens?
Where does information move?
Where does it stop?
Where does judgment occur?
Where are the exceptions?
Where is feedback missing?
What outcome are we actually trying to improve?
Only then:
What role could AI play?
This turns AI literacy into something more valuable than tool literacy.
It becomes the ability to recognize opportunities for intelligence inside work.
25. The Strategic Opportunity for Pattern AI
The research does not prove Pattern AI.
Pattern AI is a proposed methodology.
That distinction matters.
McKinsey and BCG provide evidence that:
AI adoption alone is insufficient. (McKinsey & Company)
Workflow redesign is associated with substantially greater value capture. (McKinsey & Company)
Organizations frequently struggle to connect AI initiatives with financial outcomes. (BCG)
High-value AI transformation requires changes in work, roles and operating models—not merely technology deployment. (BCG)
Pattern AI builds a hypothesis on top of that evidence:
Perhaps organizations need a systematic way to discover the patterns inside work before deciding where AI should be introduced.
That hypothesis can be tested.
And that is exactly what makes it interesting.
26. The Question Leaders Should Ask
The question for executives may therefore be changing.
Not:
How much AI are we using?
Not:
How many employees have Copilot?
Not:
How many agents have we built?
Not even:
How many AI projects do we have?
Instead:
Where exactly is AI changing the economics of our work?
Answering that question requires understanding the work before selecting the intelligence.
The organizations that do this well may not be those with the greatest number of AI tools.
They may be those that learn to see their work differently.
They will see workflows.
Inside workflows, they will see patterns.
Inside patterns, they will see friction.
Inside friction, they will discover opportunities.
Against those opportunities, they will deploy the appropriate intelligence — human, artificial, or both.
And then they will measure what changed.
One Idea to Take With You
Don't start by asking where you can use AI.
Start by looking for the patterns in your work.
Find the friction.
Match the intelligence.
Redesign the workflow.
Measure the outcome.
WORK → PATTERN → FRICTION → AI → OUTCOME
Because the next competitive advantage may not come from having more AI.
It may come from:
seeing more clearly where AI belongs.
Sources
The empirical claims in this article draw principally on current research from McKinsey & Company — The State of AI: How Organizations Are Rewiring to Capture Value, McKinsey & Company — From Adoption to Impact: Three Horizons of AI Transformation, BCG — CEOs Are Starting to See Value from AI. Now Comes Execution, BCG — How AI Is Paying Off in the Tech Function, and BCG — AI Is Moving Faster Than Your Workforce Strategy. Are You Ready?.
Pattern AI, Pattern AI Opportunity Discovery, the Pattern AI Opportunity Loop, Pattern AI Opportunity Canvas, Pattern Thinking, and Pattern AI Opportunity Engineering are proposed conceptual frameworks and interpretations; they should not be read as methodologies developed or endorsed by the cited research organizations.
No comments:
Post a Comment