Introducing Pattern Value Engineering (PVE): From Pattern Recognition to Verified Value
A Proposed AI YING Framework for Systematically Creating Measurable Value with AI
PVE v0.1 — September 2026
Executive Summary
Artificial intelligence is extraordinarily capable of working with patterns.
AI systems can recognize patterns in data, language, images and behaviour; generate new outputs from learned patterns; retrieve and compare knowledge; identify relationships; make predictions; and increasingly support decisions and actions.
But there is an important distinction:
AI capability is not the same as value creation.
An AI system can produce an impressive answer without changing an important outcome.
An organization can deploy AI without improving productivity.
A professional can use AI extensively without becoming more effective.
And research increasingly shows that AI can create substantial value for some patterns of work while providing little benefit—or even reducing performance—for others.
This creates an important practical question:
How can we systematically convert AI's ability to work with patterns into measurable value?
This article proposes an emerging framework:
Pattern Value Engineering — PVE
Pattern Value Engineering (PVE) is the systematic practice of discovering, understanding, evaluating, leveraging and reimagining patterns with AI to create and verify measurable value.
Its primary unit of investigation is:
PATTERN
Its technological lever is:
AI
Its governing capability is:
HUMAN JUDGMENT
Its mechanism is:
ACTION
Its test is:
EVIDENCE
Its destination is:
VERIFIED VALUE
PVE therefore proposes a progression:
Pattern Recognition → Pattern Understanding → Pattern Leverage → Pattern Intervention → Pattern Accomplishment → Verified Value
PVE v0.1 is presented here as a proposed AI YING framework and candidate emerging discipline.
It is not being presented as an already established academic or professional field.
The purpose of this first version is to define the proposition clearly enough for it to be criticized, tested, applied, measured and refined.
1. The AI Value Problem
AI adoption is accelerating.
But adopting AI and creating value with AI are fundamentally different things.
Much AI activity still begins with questions such as:
Which AI tool should we use?
Where can we deploy generative AI?
Which tasks can we automate?
Which prompts should employees learn?
These are legitimate questions.
But they begin with the technology.
Pattern Value Engineering proposes beginning somewhere else:
Where is value being created, lost, trapped or left unrealized—and what patterns influence that outcome?
Only then should we ask:
Can AI help us recognize, replicate, improve, predict, combine or reimagine those patterns?
This reverses the usual direction of AI adoption.
Instead of:
AI → Application → Hope for Value
PVE proposes:
Desired Value → Outcome → Pattern → Appropriate AI → Intervention → Evidence → Verified Value
The difference is significant.
The first approach begins with what AI can do.
The second begins with what matters.
2. Why PATTERN?
Pattern is not merely a convenient metaphor for discussing AI.
Pattern recognition, statistical regularity and learned representations are foundational ideas across machine learning and artificial intelligence.
Machine-learning systems learn from examples and data in order to perform tasks such as classification, prediction, generation and decision support.
The world in which these systems operate is also full of patterns:
patterns in language,
patterns in images,
patterns in customer behaviour,
patterns in machine performance,
patterns in software,
patterns in expertise,
patterns in failure,
patterns in learning,
patterns in demand,
patterns in decisions,
patterns in workflows,
patterns in organizations.
But PVE introduces an important distinction:
A pattern is not valuable merely because it exists.
A pattern becomes relevant to value creation when understanding, predicting, reproducing or changing it can influence an outcome that matters.
Therefore:
Pattern Recognition = Capability
Pattern Understanding = Insight
Pattern Intervention = Action
Improved Outcome = Potential Value
Measured Improvement = Evidence of Value
This is the territory PVE seeks to systematize.
3. Defining Pattern Value Engineering
The proposed foundational definition is:
Pattern Value Engineering is the systematic practice of discovering, understanding, evaluating, leveraging and reimagining patterns with AI to create and verify measurable value.
Each part of the definition matters.
SYSTEMATIC
PVE is not simply noticing interesting patterns.
It proposes a repeatable method.
DISCOVERING
Some valuable patterns are obvious.
Others are hidden across large amounts of data, behaviour, knowledge or experience.
AI may help surface them.
UNDERSTANDING
Recognizing a statistical relationship is not necessarily understanding it.
Context matters.
Correlation is not automatically causation.
EVALUATING
Not every pattern deserves intervention.
Some are irrelevant.
Some are unstable.
Some may be artifacts of poor data.
Some may be ethically inappropriate to exploit.
LEVERAGING
AI may help recognize, replicate, improve, predict or combine useful patterns.
REIMAGINING
Sometimes the greatest value will not come from improving an existing pattern.
AI may make an entirely different pattern of accomplishment possible.
MEASURABLE
PVE requires the intended value to be defined.
VERIFIED
Where meaningful measurement is possible, the outcome should be compared with an appropriate baseline or counterfactual.
This leads to the first foundational PVE principle:
AI use is not evidence of AI value.
4. Pattern + AI Does Not Equal Value
A central PVE proposition is:
Pattern + AI ≠ Value
A more useful conceptual expression is:
Useful Pattern + Appropriate AI + Human Judgment + Action + Verified Outcome = VALUE
This is not intended as a literal mathematical equation.
It is a reasoning discipline.
Every component matters.
A genuine pattern applied to an irrelevant outcome creates little value.
A relevant pattern processed by an unsuitable AI system can create errors.
A good AI recommendation without action produces no operational improvement.
An intervention without measurement may create perceived rather than demonstrated value.
And an intervention that improves one metric while creating unacceptable consequences elsewhere may destroy overall value.
PVE therefore requires us to complete the journey:
Pattern → AI Leverage → Intervention → Outcome → Evidence → Value
5. The Evidence: AI Value Depends on the Pattern of Work
One of the strongest reasons for a Pattern Value Engineering approach comes from empirical research showing that AI's effects vary substantially according to the nature of the task, worker and implementation.
Consider research involving consultants at Boston Consulting Group.
A field experiment involving 758 consultants examined realistic knowledge-work tasks using GPT-4.
For tasks that were within the AI system's capability frontier, consultants using AI completed more work, worked faster and produced higher-quality results.
But the researchers deliberately included a task outside that frontier.
On that task, participants using AI were more likely to arrive at an incorrect answer.
The researchers described the boundary as a:
“Jagged Technological Frontier.”
This finding is fundamental to PVE.
It tells us that the useful question is not simply:
Can we give this task to AI?
The better question is:
What pattern of work does this task represent, and how well does the AI's capability fit that pattern?
Therefore:
AI Availability ≠ AI Suitability ≠ Value
AI value is conditional.
Pattern-task fit matters.
Human judgment matters.
Verification matters.
6. REPLICATE: Turning Expert Patterns into Shared Capability
Consider customer support.
Researchers Erik Brynjolfsson, Danielle Li and Lindsey Raymond studied the staggered introduction of a generative-AI conversational assistant using data from 5,179 customer-support agents.
They found that AI assistance increased productivity, measured as issues resolved per hour, by approximately:
14% on average.
The gains were considerably larger for novice and lower-skilled workers:
approximately 34%.
The researchers also found suggestive evidence that AI assistance helped disseminate some of the practices associated with more capable workers.
These are study-specific results and should not be interpreted as guaranteed gains in other organizations.
But the mechanism is particularly interesting through the PVE lens.
The organization already possessed valuable capability.
It existed partly in the behaviours of experienced employees.
AI helped make useful aspects of those patterns more accessible to others.
The PVE interpretation becomes:
Expert Behaviour
→ Successful Pattern
→ AI Mediation
→ Pattern Replication
→ Improved Capability
→ Measured Outcome
The pattern existed before AI.
AI changed the economics of accessing and reproducing it.
This gives us the first major Pattern Value Move:
REPLICATE
What successful pattern can AI help us reproduce?
7. Six Fundamental Pattern Value Moves
PVE v0.1 proposes six fundamental mechanisms through which patterns can become sources of value.
1. RECOGNIZE
What important pattern is present?
AI can help identify:
recurrence,
similarity,
anomaly,
sequence,
relationship,
classification,
structure,
or deviation.
Examples might include:
recurring causes of customer dissatisfaction,
abnormal machine behaviour,
common software defects,
successful learning behaviours,
unusual financial activity,
or patterns associated with quality problems.
Value begins with:
VISIBILITY
We cannot deliberately act on a pattern we cannot see.
8. REPLICATE
What successful pattern can be reproduced?
If experts repeatedly demonstrate effective behaviours, AI may help identify, retrieve or make those patterns available to others.
Potential applications include:
customer service,
sales,
engineering troubleshooting,
education,
software development,
professional services,
maintenance,
and organizational knowledge.
The value mechanism is:
Expertise → Pattern → Accessibility → Replication → Capability
AI does not necessarily replace expertise.
It may increase the:
REACH OF EXPERTISE.
9. IMPROVE
Which existing pattern of work can perform better?
Consider software development.
GitHub conducted a controlled experiment involving 95 professional developers.
Participants completed the same JavaScript HTTP-server programming task with or without GitHub Copilot.
Developers using Copilot completed that particular task approximately:
55% faster.
The Copilot group averaged roughly 1 hour 11 minutes, compared with about 2 hours 41 minutes for the control group.
This result concerns a specific programming task and experimental setting. It should not be generalized into a claim that AI makes all software engineering 55% faster.
But the mechanism is useful.
Programming contains recurring structures:
syntax,
boilerplate,
common functions,
API conventions,
tests,
documentation structures,
and implementation patterns.
AI can generate many of these rapidly.
The deeper PVE interpretation is therefore not:
AI writes code faster.
It is:
AI can change the distribution of cognitive effort within the programming pattern.
Human attention can potentially shift toward:
architecture,
requirements,
security,
testing,
system understanding,
constraints,
and judgment.
This is the IMPROVE move:
Existing Pattern → AI Assistance → Improved Performance
10. PREDICT
What pattern indicates what may happen next?
Prediction creates value because it can change the timing of action.
Consider industrial maintenance.
A conventional maintenance pattern may be:
Failure → Detection → Diagnosis → Repair
But equipment may produce warning signals before failure:
temperature,
vibration,
sound,
pressure,
energy consumption,
cycle behaviour,
error frequency,
or combinations of these.
AI systems can potentially identify patterns associated with developing problems.
The operating pattern becomes:
Signals → Pattern → Prediction → Intervention → Avoided Failure
The value is not simply better analysis.
It is:
EARLIER ACTION.
This general value pattern appears in areas including:
maintenance,
fraud detection,
cybersecurity,
quality control,
demand forecasting,
supply chains,
and risk management.
11. COMBINE
What becomes visible when multiple patterns are connected?
Many important decisions do not depend on a single signal.
Consider retail demand.
Relevant patterns might include:
historical sales,
seasonality,
weather,
inventory,
pricing,
local events,
customer behaviour,
and logistics constraints.
Humans and organizations often examine these through separate systems and departments.
AI may help reason across several signals simultaneously.
The value pattern becomes:
Pattern A + Pattern B + Pattern C + Context
→ Integrated Interpretation
→ Better Decision
This suggests a larger organizational opportunity.
Organizations are divided into departments.
Reality is not.
Sales sees customer patterns.
Finance sees payment patterns.
Operations sees production patterns.
Service sees complaint patterns.
Supply chains see inventory patterns.
PVE asks:
What new value becomes possible when previously separated patterns can be considered together?
12. REIMAGINE
The sixth Pattern Value Move may ultimately be the most important.
What new pattern of accomplishment becomes possible because AI exists?
Suppose an organization produces a monthly management report.
AI could help employees produce that report faster.
That is:
IMPROVE.
But PVE asks another question:
Why does the report exist?
Perhaps the underlying requirement is simply:
Management needs to know when something important changes.
If continuous AI-assisted monitoring becomes feasible, the pattern might change from:
Collect → Analyze → Write → Publish → Read → Decide
to:
Monitor → Detect Exception → Explain → Human Judgment → Act
AI has now done something more important than accelerating the existing process.
It has enabled a different pattern of accomplishment.
That is:
REIMAGINE.
This produces an important PVE warning:
Do not use AI merely to automate a pattern that AI itself may make obsolete.
13. Precision Agriculture: Pattern Recognition Connected to Precise Action
John Deere's See & Spray provides a useful real-world illustration.
The system uses cameras, computer vision and machine learning to identify target weeds and support selective spraying.
For See & Spray Select, John Deere reports average herbicide savings of approximately:
77%
under its stated testing conditions.
This is a company-reported operational result, not an independent controlled estimate, and should be interpreted accordingly.
But the underlying pattern is illuminating.
Conventional operating pattern:
Field → Broad Spraying
Pattern-enabled operating model:
Observe → Recognize → Classify → Target → Act
This suggests another important PVE category:
PRECISION VALUE
When AI lowers the cost of recognizing relevant patterns at fine granularity, action can sometimes move from:
one action applied broadly
toward:
context-specific action for individual instances.
This principle potentially extends into:
manufacturing inspection,
personalized learning,
maintenance,
customer service,
resource allocation,
and many other domains.
14. Google Data Centres: Complex Patterns as Economic Assets
Google DeepMind reported an important industrial example involving data-centre cooling.
Data centres contain complex interactions among variables such as:
temperature,
equipment state,
pump speeds,
setpoints,
pressure,
environmental conditions,
and energy consumption.
DeepMind reported training neural networks on historical sensor data to model relationships between operating conditions and energy efficiency.
In its reported live application, the company stated that the system reduced energy used for cooling by:
up to 40%.
It also reported that this corresponded to approximately a 15% reduction in overall PUE overhead, after accounting for electrical and other non-cooling inefficiencies.
These are company-reported operational results.
The PVE pattern is:
Operational Data
→ Complex Pattern
→ Prediction
→ Optimized Action
→ Reduced Energy
→ Value
Here, the ability to model the pattern becomes an economic asset.
15. AI Can Change the Pattern of Collaboration
Pattern Value Engineering becomes even more interesting when AI changes not simply individual productivity but the way expertise is assembled.
A preregistered field experiment involving 776 professionals at Procter & Gamble studied product-development challenges.
Participants worked individually or in two-person teams, with or without AI.
One important result was that individuals working with AI could achieve performance comparable with teams without AI on the studied tasks.
Researchers also found evidence that AI helped participants bridge functional knowledge boundaries.
This does not establish that AI can generally replace teams.
The findings concern particular tasks and experimental conditions.
But they demonstrate something strategically interesting.
A traditional expertise pattern might be:
Specialist A + Specialist B → Combined Expertise
An emerging pattern might sometimes become:
Specialist A + AI-mediated complementary knowledge → Broader Capability
The PVE question becomes:
Can AI change the architecture through which expertise is assembled?
That moves Pattern Value Engineering beyond task productivity.
It enters organizational design.
16. Morgan Stanley: Knowledge-to-Action as a Value Pattern
Large organizations often possess enormous quantities of knowledge.
The challenge is not always knowledge creation.
Sometimes the problem is the distance between:
knowledge existing
and
knowledge being available at the moment it matters.
Morgan Stanley Wealth Management provides an illustrative example.
The company introduced an AI Assistant designed to help financial advisers retrieve relevant internal intellectual capital.
In June 2024, Morgan Stanley reported that:
98% of Financial Advisor teams
had adopted the AI Assistant.
It also introduced AI @ Morgan Stanley Debrief.
With client consent, the system can generate meeting notes, identify action items and prepare a draft follow-up communication for adviser review.
Again, the adoption figure is company-reported.
The PVE pattern is:
Organizational Knowledge
→ Retrieval
→ Contextualization
→ Human Judgment
→ Action
The value mechanism is:
KNOWLEDGE-TO-ACTION.
The knowledge already existed.
AI can reduce the friction between possessing knowledge and applying it when needed.
17. Relationship to Established Value Engineering
If Pattern Value Engineering is to develop into a serious framework, its relationship to existing disciplines must be explicit.
The most obvious comparison is:
Value Engineering.
Value Engineering is an established discipline.
SAVE International describes the Value Methodology as a systematic process used by multidisciplinary teams to improve the value of projects, products, processes, services and organizations through the analysis of:
FUNCTIONS.
SAVE International expresses the concept of value in terms of the relationship between function performance and the resources required to achieve it.
Its formal methodology includes structured stages such as:
information gathering,
function analysis,
creative exploration,
evaluation,
development,
presentation,
and implementation.
The central analytical object is therefore:
FUNCTION.
Pattern Value Engineering proposes another primary analytical object:
PATTERN.
The distinction can be expressed carefully as:
VALUE ENGINEERING
What function must be accomplished, and how can its value be improved?
PATTERN VALUE ENGINEERING
What pattern influences the desired outcome, and how can AI help us understand, leverage, modify or reimagine that pattern to create measurable value?
In shorthand:
Value Engineering: Engineer the FUNCTION for value.
Pattern Value Engineering: Engineer the PATTERN toward value.
PVE should not be positioned as a replacement for Value Engineering.
Indeed, an important future research question is whether:
FUNCTION ANALYSIS + PATTERN ANALYSIS
could become complementary approaches.
That relationship deserves serious investigation.
18. Is Pattern Engineering Already an Established Discipline?
This question requires careful treatment.
The words “pattern engineering” have been used in different technical contexts.
Computing has long-established traditions involving:
software design patterns,
pattern languages,
pattern recognition,
and reusable architectural patterns.
More recent AI and software projects also use variations of “pattern engineering” to describe reusable practices, agent patterns or structured AI-development approaches.
Therefore it would be inappropriate to claim that the words “Pattern Engineering” are entirely new.
However, the existence of the phrase is different from the existence of a standardized professional discipline.
At the time of writing, I have not found evidence of an established professional discipline called:
PATTERN VALUE ENGINEERING
with a widely recognized body of knowledge, professional methodology and institutional structure comparable with established Value Engineering, specifically in the sense defined in this article.
Therefore PVE is presented here as:
an AI YING proposed framework and candidate emerging discipline for systematically converting outcome-relevant patterns into verified value with AI.
This is deliberately a narrower claim.
Whether PVE deserves to mature into a distinct discipline should be determined by:
research,
application,
comparison with adjacent fields,
replication,
professional critique,
and evidence.
19. What Might Be Distinctive About PVE?
The proposed contribution is not simply putting the words:
Pattern + Value + Engineering
together.
Its proposed distinctive structure is the integration of:
1. PATTERN AS THE UNIT OF INVESTIGATION
What recurring or relational structure influences the outcome?
2. AI AS LEVERAGE
What becomes possible because AI can recognize, generate, retrieve, predict, combine or transform patterns?
3. HUMAN JUDGMENT AS GOVERNANCE
What should be done?
What should not be delegated?
What context or responsibility remains human?
4. INTERVENTION AS THE MECHANISM
What actually changes?
5. EVIDENCE AS THE TEST
Did the outcome improve?
6. VALUE AS THE DESTINATION
Did the change matter?
Together:
Pattern → AI Leverage → Human Judgment → Intervention → Evidence → Verified Value
That combination—not merely the terminology—is the PVE proposition.
20. PVE Is Not Prompt Engineering
Prompt Engineering asks:
How should we interact with an AI system to obtain better outputs?
That can be valuable.
PVE operates at another level.
It asks:
What pattern matters to the outcome, and how can that pattern be converted into measurable value?
A brilliant prompt can produce an impressive response.
But if nothing important changes:
PVE Value = Unproven.
Prompting can therefore be a technique used within PVE.
It is not the discipline itself.
21. PVE Is Not Simply Automation
Automation asks:
Can this activity be automated?
PVE asks first:
Why does this activity exist in this form?
Then:
What outcome matters?
What pattern drives it?
Should we automate the pattern?
Improve it?
Or replace it with a different pattern?
This distinction helps avoid a major AI-era mistake:
Automating Yesterday's Assumptions.
The objective should not necessarily be:
Old Process + AI = Faster Old Process
Sometimes the better result is:
Underlying Need + Pattern Understanding + AI = New Pattern of Accomplishment
22. The PVE Method
PVE v0.1 proposes an eight-stage operating method:
SEE → UNDERSTAND → VALUE → LEVER → REIMAGINE → ACT → PROVE → LEARN
SEE
What pattern exists?
Look for:
repetition,
variation,
relationships,
success,
failure,
sequence,
exceptions,
signals,
behaviour,
and structure.
UNDERSTAND
Why might the pattern matter?
Ask:
What evidence supports it?
How stable is it?
Could it be coincidence?
Is it correlation or causation?
What context affects it?
Where might the interpretation fail?
VALUE
What outcome matters?
Identify where value is:
created,
lost,
trapped,
or
left unrealized.
LEVER
How can AI appropriately work with the pattern?
Can AI:
RECOGNIZE
REPLICATE
IMPROVE
PREDICT
or
COMBINE
the pattern?
REIMAGINE
Ask:
If this activity were being designed today with AI available from the beginning, would we design it the same way?
If not:
What new pattern becomes possible?
ACT
Convert understanding into an intervention.
That could be:
a decision,
workflow,
recommendation,
product,
service,
artifact,
maintenance action,
learning intervention,
or operating change.
PROVE
Measure what happened.
Compare against an appropriate baseline wherever feasible.
LEARN
Every intervention produces new evidence.
Every outcome may reveal new patterns.
Therefore:
Pattern → Intervention → Outcome → Evidence → Learning → Better Pattern
PVE is therefore a learning loop, not merely an AI implementation method.
23. What Counts as Value?
PVE deliberately defines value broadly while requiring clarity about what is being measured.
TIME VALUE
Time saved or decision latency reduced.
COST VALUE
Expenditure or resource consumption reduced.
QUALITY VALUE
Accuracy, reliability or consistency improved.
REVENUE VALUE
Economic opportunity created or increased.
RISK VALUE
Failure, exposure, uncertainty or error reduced.
EXPERIENCE VALUE
Customer, employee, learner or citizen experience improved.
KNOWLEDGE VALUE
Relevant knowledge becomes more accessible and usable.
LEARNING VALUE
Understanding or capability develops more effectively.
PRECISION VALUE
Broad action becomes targeted intervention.
SCALE VALUE
Useful capability becomes available to more people.
CAPABILITY VALUE
Something previously difficult becomes feasible.
INNOVATION VALUE
A new product, service, workflow or operating pattern becomes possible.
The value category should ideally be defined:
BEFORE
the intervention.
Otherwise there is a danger of redefining success after seeing the results.
24. Baseline Before AI
This should become a mandatory PVE principle wherever meaningful measurement is possible.
Before asking:
How well did AI perform?
ask:
Compared with what?
Possible baselines include:
existing completion time,
current cost,
human-only performance,
quality,
error rates,
failure rates,
energy consumption,
conversion,
customer satisfaction,
learning outcomes,
or resource utilization.
The PVE sequence becomes:
Baseline → Intervention → Measurement → Comparison → Value Claim
Without a meaningful comparison, statements such as:
“AI improved productivity”
may remain impressions rather than evidence.
25. Evidence Is Mandatory
PVE should explicitly distinguish:
Pattern Observed
from
Pattern Hypothesized
from
Pattern Predicted
from
Pattern Demonstrated.
They are not equivalent.
An AI-generated explanation is not evidence.
A plausible relationship is not necessarily causal.
A correlation does not automatically justify intervention.
A successful pilot does not automatically prove scalability.
Therefore:
The strength of evidence required should increase with the consequence of the decision.
Low-risk experimentation may tolerate weaker evidence.
High-stakes decisions require much stronger validation, governance and domain expertise.
26. The Pattern-to-Value Canvas
A practical PVE exercise can begin with ten questions.
1. SITUATION
What are we trying to accomplish?
2. OUTCOME
What result actually matters?
3. PATTERN
What repeats, varies, predicts, succeeds or fails?
4. EVIDENCE
How do we know the pattern exists?
5. VALUE
What value could be created, protected or recovered?
6. AI LEVER
Should AI recognize, replicate, improve, predict, combine or help reimagine the pattern?
7. HUMAN ROLE
Where are context, ethics, responsibility, expertise, empathy and judgment essential?
8. INTERVENTION
What specifically will change?
9. METRIC
How will value be measured against an appropriate baseline?
10. LEARNING
What did the result teach us about the original pattern?
This converts PVE from an idea into a usable method.
27. From AI Use Cases to Pattern Value Libraries
Organizations frequently collect:
AI USE CASES.
That is useful.
But hundreds or thousands of use cases can become disconnected examples.
PVE asks whether we can identify the underlying:
VALUE PATTERNS.
Consider several examples.
VALUE PATTERN 1 — EARLY WARNING
Factory:
equipment failure.
Financial services:
suspicious activity.
Cybersecurity:
network anomaly.
Retail:
unexpected demand.
Underlying structure:
Normal Behaviour → Deviation → Detection → Interpretation → Intervention
VALUE PATTERN 2 — EXPERTISE TRANSFER
Customer service:
agent assistance.
Engineering:
troubleshooting.
Software:
coding assistance.
Education:
learner support.
Underlying structure:
Expertise → Pattern → AI Mediation → Wider Capability
VALUE PATTERN 3 — PRECISION ACTION
Agriculture:
targeted spraying.
Manufacturing:
targeted inspection.
Education:
individual learning intervention.
Customer service:
context-specific support.
Underlying structure:
Observe → Recognize → Classify → Contextualize → Target Action
VALUE PATTERN 4 — KNOWLEDGE-TO-ACTION
Finance:
adviser knowledge.
Engineering:
technical manuals.
Operations:
procedures.
Professional services:
organizational expertise.
Underlying structure:
Knowledge → Retrieval → Context → Judgment → Action
VALUE PATTERN 5 — MULTI-PATTERN DECISION
Demand,
weather,
inventory,
customer behaviour,
logistics,
pricing.
Underlying structure:
Multiple Patterns → Integrated Interpretation → Better Decision
VALUE PATTERN 6 — CONTINUOUS EXCEPTION MANAGEMENT
Traditional pattern:
Periodic Report → Review → Decision
Emerging pattern:
Continuous Monitoring → Exception → Explanation → Judgment → Action
These patterns may be more transferable than individual use cases.
That creates the possibility of:
PATTERN VALUE LIBRARIES
that can be tested and adapted across roles, professions and industries.
28. Human Judgment Is Not a Residual Role
PVE is not a methodology for maximizing AI involvement.
It is a methodology for maximizing:
VALUE.
Sometimes extensive AI involvement may create value.
Sometimes limited AI involvement may be preferable.
Sometimes the correct conclusion may be:
DO NOT USE AI HERE.
Humans remain responsible for questions such as:
What outcome matters?
Which pattern deserves attention?
Is the pattern meaningful?
What evidence is sufficient?
What context is absent?
What values are involved?
What risks are acceptable?
Is AI suitable?
Should action be taken?
Who is accountable?
Did value actually result?
AI may make these judgment questions more important, not less.
29. PVE and AI YING's TAJ
Pattern Value Engineering connects naturally with the AI YING discipline:
THINK → ASK → JUDGE
THINK
What are we trying to accomplish?
What outcome matters?
What pattern may influence it?
What would value mean?
ASK
Use AI appropriately to:
discover,
retrieve,
compare,
generate,
simulate,
predict,
and explore patterns.
JUDGE
Is the pattern valid?
Is AI suitable?
Is the output reliable?
What risks exist?
Should we act?
Did the intervention create value?
TAJ therefore provides a human operating discipline within PVE.
30. The PVE Value Ladder
PVE can also be expressed as a maturity ladder.
LEVEL 1 — SEE
We recognize a potentially important pattern.
Value:
Awareness.
LEVEL 2 — UNDERSTAND
We understand why the pattern may matter.
Value:
Insight.
LEVEL 3 — LEVER
AI can usefully work with the pattern.
Value:
Capability.
LEVEL 4 — ACT
Something changes because of the pattern.
Value:
Intervention.
LEVEL 5 — PROVE
The intervention produces measurable improvement.
Value:
Verified Value.
LEVEL 6 — REIMAGINE
AI enables a fundamentally different pattern of accomplishment.
Value:
Transformational Value.
Many AI initiatives demonstrate Level 3:
AI can do something.
PVE asks us to continue until we can answer:
Did something valuable happen?
31. The PVE Failure Test
Before declaring a PVE initiative successful, ask:
Did we identify a genuine pattern?
Is it relevant to an important outcome?
What evidence supports the pattern?
Is AI actually suitable for working with it?
What baseline are we comparing against?
Did the intervention materially change anything?
Did the desired outcome improve?
Can the improvement reasonably be associated with the intervention?
Did the intervention create negative consequences elsewhere?
Is the result repeatable?
Is it scalable?
Does the value exceed the total cost, risk and complexity introduced?
If these questions cannot be answered satisfactorily, the initiative should not yet be described as:
VERIFIED PATTERN VALUE.
32. Ten Foundational Principles of PVE v0.1
PRINCIPLE 1
Begin with value, not AI.
PRINCIPLE 2
Pattern is the primary unit of investigation.
PRINCIPLE 3
Patterns require evidence.
PRINCIPLE 4
AI suitability is pattern-dependent.
PRINCIPLE 5
Human judgment is part of the value system.
PRINCIPLE 6
Pattern recognition without action is incomplete.
PRINCIPLE 7
Action without measurement does not establish value.
PRINCIPLE 8
Improvement should not prevent reimagination.
PRINCIPLE 9
Every outcome creates new evidence and potentially new patterns.
PRINCIPLE 10
The ultimate measure of PVE is accomplishment—not AI usage.
33. The Simplest PVE Test
The framework can initially be compressed into three questions.
P — PATTERN
What pattern matters?
V — VALUE
What measurable value could result?
E — EVIDENCE
How will we know?
Therefore:
PVE = PATTERN → VALUE → EVIDENCE
If the pattern cannot be articulated,
if the value cannot be defined,
or if the evidence cannot be specified,
the initiative may not yet be ready for Pattern Value Engineering.
34. Why Pattern Value Engineering Could Matter
AI capabilities are becoming increasingly accessible.
That changes the competitive question.
Imagine two organizations with access to similar AI models.
One uses AI primarily to produce more:
emails,
documents,
presentations,
summaries.
The other discovers patterns that allow it to:
prevent recurring failures,
transfer expert capability,
reduce decision latency,
personalize interventions,
combine previously separated knowledge,
or redesign an entire operating model.
They may possess similar AI.
They do not possess the same:
PATTERN SKILL.
The emerging differentiator may increasingly become:
Who can discover the patterns through which AI creates value?
PVE attempts to turn that capability into a systematic discipline.
35. Research Agenda for Pattern Value Engineering
PVE v0.1 should not be treated as finished.
If it is to become a credible framework—or eventually justify recognition as a distinct discipline—it must expose itself to comparison and empirical testing.
Important questions include:
PATTERN TAXONOMY
What types of patterns are most relevant to value creation?
PATTERN EVIDENCE
How do we distinguish genuine patterns from coincidence, spurious correlation and AI-generated narratives?
AI–PATTERN FIT
Which AI capabilities reliably work with which pattern classes?
VALUE MEASUREMENT
How should different forms of Pattern Value be measured?
CAUSALITY
How can we establish whether changing the pattern actually caused the improvement?
HUMAN–AI ALLOCATION
Which components of different value patterns should remain human-led?
PATTERN TRANSFERABILITY
When can a Pattern Value demonstrated in one domain be adapted successfully to another?
VALUE DESTRUCTION
How should unintended consequences be incorporated into the calculation?
PATTERN VALUE LIBRARIES
How should reusable patterns be validated?
PVE MATURITY
Can organizations be assessed according to their ability to move from Pattern Recognition to Verified Value?
COMPARATIVE PERFORMANCE
Does PVE lead to better AI investment and implementation decisions than tool-first, automation-first or use-case-first approaches?
These should become:
research questions,
not assumptions.
36. Invitation to Critique, Test and Refine
Pattern Value Engineering v0.1 is deliberately being presented as:
A PROPOSED AI YING FRAMEWORK AND CANDIDATE EMERGING DISCIPLINE.
It is not being presented as an established academic or professional field.
It does not claim that related concepts such as pattern recognition, pattern languages, software patterns, Value Engineering, systems engineering, process improvement, decision science or other adjacent disciplines did not already exist.
Instead, PVE proposes a particular integration:
PATTERN
AI LEVERAGE
HUMAN JUDGMENT
INTERVENTION
EVIDENCE
→
VERIFIED VALUE
The framework should therefore be:
challenged,
compared,
tested,
applied,
measured,
refined,
and, where necessary,
changed.
A framework centered on evidence should itself remain:
EVIDENCE-SEEKING.
37. Foundational Statement of PVE v0.1
The proposed foundational definition is:
Pattern Value Engineering is the systematic practice of discovering, understanding, evaluating, leveraging and reimagining patterns with AI to create and verify measurable value.
Its primary object is:
PATTERN
Its technological lever is:
AI
Its governing capability is:
HUMAN JUDGMENT
Its mechanism is:
ACTION
Its test is:
EVIDENCE
Its destination is:
VERIFIED VALUE
Its larger purpose is:
ACCOMPLISHMENT.
The complete chain becomes:
Pattern → AI Leverage → Human Judgment → Intervention → Outcome → Evidence → Verified Value
Conclusion: From Pattern Recognition to Verified Value
AI has made powerful pattern capabilities increasingly accessible.
But access to intelligence is not accomplishment.
And AI activity is not automatically value.
Patterns are potential.
AI expands what can be done with those patterns.
Human judgment determines which possibilities deserve action.
Intervention changes what happens.
Evidence tells us whether the change mattered.
Therefore:
PATTERN RECOGNITION
is only the beginning.
The larger journey is:
Pattern Recognition
↓
Pattern Understanding
↓
Pattern Leverage
↓
Pattern Reimagination
↓
Pattern Intervention
↓
Pattern Accomplishment
↓
Verified Value
This is the territory that Pattern Value Engineering proposes to explore.
Pattern AI provides the lens:
See the world through patterns.
Pattern Value Engineering adds the discipline:
Discover the patterns that matter. Leverage them appropriately with AI. Change what happens. Measure what matters. Prove the value.
Or in its shortest form:
Patterns are potential. Outcomes are value.
PVE is the proposed discipline for systematically converting one into the other.
Sources and Evidence Notes
1. SAVE International — Value Methodology
SAVE International is the professional society associated with the Value Methodology and describes the methodology as a systematic process using function analysis to improve value.
Official source:
https://www.value-eng.org/
The comparison with PVE in this article is conceptual. PVE is not presented as part of, endorsed by, or a replacement for SAVE International's Value Methodology.
2. Brynjolfsson, Li & Raymond — Generative AI at Work
Erik Brynjolfsson, Danielle Li and Lindsey Raymond studied the deployment of generative AI among 5,179 customer-support agents.
The study reported approximately 14% average productivity improvement, with considerably larger gains among novice and lower-skilled workers.
NBER Working Paper 31161:
https://www.nber.org/papers/w31161
The paper was subsequently published in the Quarterly Journal of Economics.
Results are specific to the studied environment and should not be interpreted as guaranteed productivity improvements elsewhere.
3. Dell'Acqua et al. — Navigating the Jagged Technological Frontier
The research examined 758 Boston Consulting Group consultants performing realistic knowledge-work tasks with and without GPT-4.
The study demonstrated significant performance improvements on tasks inside the model's capability frontier while also showing that AI reliance could reduce correctness on a task outside that frontier.
Harvard Business School / Digital Data Design Institute:
https://aiinstitute.hbs.edu/navigating-the-jagged-technological-frontier/
The research was subsequently published in Organization Science.
4. Dell'Acqua et al. — The Cybernetic Teammate
A preregistered field experiment involving 776 Procter & Gamble professionals examined individuals and teams working with and without generative AI on product-innovation challenges.
NBER Working Paper 33641:
https://www.nber.org/papers/w33641
The research was published online in Organization Science in 2026.
Results should be interpreted within the specific tasks and experimental conditions studied.
5. GitHub Copilot Productivity Research
GitHub reported a controlled experiment involving 95 professional developers completing a specified JavaScript programming task.
Developers using Copilot completed the task approximately 55% faster on average.
The result applies to the studied task and should not be generalized to all software-development work.
6. Google DeepMind — Data Centre Cooling
Google DeepMind reported that its machine-learning system reduced energy used for cooling in its data centres by up to 40% in the described deployment.
Source:
https://deepmind.google/discover/blog/deepmind-ai-reduces-google-data-centre-cooling-bill-by-40/
This article identifies the figure as a company-reported operational result.
7. John Deere — See & Spray
John Deere describes See & Spray systems that use computer vision to identify weeds and enable selective spraying.
The company reports average herbicide savings of approximately 77% for See & Spray Select under its stated conditions.
Official source:
https://www.deere.com/
This article identifies the figure as a company-reported result rather than an independent experimental finding.
8. Morgan Stanley — AI @ Morgan Stanley
Morgan Stanley reported in June 2024 that 98% of its Financial Advisor teams had adopted its AI Assistant and described AI @ Morgan Stanley Debrief for generating meeting summaries, action items and draft communications for adviser review.
Official source:
https://www.morganstanley.com/press-releases/ai-at-morgan-stanley-debrief-launch
This article identifies the adoption figure as company-reported.
Evidence Disclaimer
The quantitative findings cited in this article come from different kinds of evidence.
Some are controlled or field experiments published through academic research channels.
Others are operational results reported by the companies deploying the technology.
They should not be treated as equivalent forms of evidence.
Experimental productivity improvements are specific to the populations, tasks, AI systems and implementation conditions studied.
Company-reported operational results are explicitly identified as such.
None of the numerical findings should be interpreted as guaranteed improvements from adopting AI elsewhere.
Terminology and Novelty Note
The terms pattern recognition, software patterns, design patterns, pattern languages and variations of pattern engineering already exist in computing, engineering and related literature.
This article therefore does not claim invention of the general concept of patterns in engineering or AI.
Nor does it claim that no one has previously used the words “Pattern Value Engineering.”
The narrower proposal made here is the PVE v0.1 framework defined as:
Pattern as the primary unit of investigation + appropriate AI leverage + human judgment + intervention + evidence + verified value.
At the time of publication, PVE is presented as an AI YING proposed framework and candidate emerging discipline, not as an established professional or academic discipline.
Its distinctiveness and usefulness should ultimately be determined through comparison with adjacent disciplines, practical application, empirical evidence and independent critique.
Pattern Value Engineering — PVE v0.1
September 2026
AI YING — Accomplishing with AI
This is the version I would publish as PVE v0.1. The most important change is the novelty statement: instead of making the difficult-to-prove claim that nobody has ever used “Pattern Value Engineering,” it precisely defines what this PVE framework means and invites evidence, comparison and criticism. That makes the proposal considerably more defensible.
No comments:
Post a Comment