ROC / STRATEGIC INQUIRY PRACTICE

When AI takes part of the work,
what happens to the work that remains human?

ROC is a strategic inquiry practice investigating where judgment, exceptions, coordination and responsibility move as AI becomes operational.

See where your question begins ↓

A SELF-LOCATION MAP

Where does your question really begin?

Not every problem is ready to be built or optimized.

HIGHER UNCERTAINTY
LOWER UNCERTAINTY
MYSTERY

What is actually going on?

We can see what is happening, but cannot reliably explain why.

INVESTIGATE
HEURISTIC

Is this really why?

We have plausible explanations, but they still need testing.

TEST
ALGORITHM

Do we understand enough to act?

The relationships are understood well enough to act.

BUILD / OPTIMIZE
ROC self-location map, adapted from Roger Martin’s Knowledge Funnel.

Don’t choose the most advanced state.

Find where your question honestly begins.

When AI handles part of the work, what happens to the rest? AI can take over some tasks while human oversight, exception handling and changes in attention may remain. Judgment and responsibility remain open questions. AI ENTERS WORK When AI handles part of the work, what happens to the rest? Some human work may be reduced. What remains may change. WORK BEFORE AI AI ENTERS THE WORK WHAT MAY CHANGE TASKS Some tasks can transfer to AI. OVERSIGHT Human evaluation or supervision can remain. EXCEPTIONS People can remain involved when cases escalate or require intervention / repair. ATTENTION Where human attention goes may change. What changes depends on the work and the deployment. STILL OPEN — UNDER INQUIRY Where does judgment move? Where does responsibility land? Evidence base: reviewed production-agent research and field experiments. ROC Phase 3, 2026.
When AI handles part of the work, what happens to the rest? AI may take over some tasks while oversight, exception handling and changes in attention can remain. Judgment and responsibility remain open questions. AI ENTERS WORK When AI handles part of the work, what happens to the rest? Some human work may be reduced. What remains may change. WORK BEFORE AI AI ENTERS THE WORK WHAT MAY CHANGE TASKS Some tasks can transfer to AI. OVERSIGHT Human evaluation or supervision can remain. EXCEPTIONS People can remain involved when cases escalate or require intervention / repair. ATTENTION Where human attention goes may change. What changes depends on the work and the deployment. STILL OPEN — UNDER INQUIRY Where does judgment move? Where does responsibility land? Evidence base: reviewed production-agent research and field experiments. ROC Phase 3, 2026.
Workflow is not the whole of work A formal workflow showing role, task, handoff, decision and output sits within a wider field of judgment, exceptions, local knowledge, coordination, trust, norms and responsibility. WORK AS A SYSTEM Workflow is not the whole of work. A mapped workflow captures the path, not every condition that makes it function. A workflow shows the path. ROLE TASK HANDOFF DECISION OUTPUT TRUST EXCEPTIONS JUDGMENT LOCAL KNOWLEDGE COORDINATION NORMS RESPONSIBILITY Work includes what makes the path function in practice. Examples of what may shape work in practice — not a complete model. So what exactly are we asking AI to enter? Adjacent evidence: ReD Associates + Jigsaw, Collective Intelligence Online (EPIC Proceedings, 2025). One concrete setting illustrating how work can exceed its mapped workflow — not a universal model.
Workflow is not the whole of work A vertical workflow sits inside a wider field of contextual relationships and conditions that can shape how work functions. WORK AS A SYSTEM Workflow is not the whole of work. A mapped workflow captures the path, not every condition that makes it function. A workflow shows the path. ROLE TASK HANDOFF DECISION OUTPUT TRUST LOCAL KNOWLEDGE EXCEPTIONS COORDINATION JUDGMENT NORMS RESPONSIBILITY Work includes what makes the path function in practice. Examples of what may shape work in practice — not a complete model. So what exactly are we asking AI to enter? Adjacent evidence: ReD Associates + Jigsaw, Collective Intelligence Online (EPIC Proceedings, 2025). Illustrative context — not a universal model.
The question is part of the work. The same AI deployment is viewed through three different questions. Each question directs attention toward different parts of the work. HOW ROC INQUIRES The question is part of the work. How a problem is framed shapes what becomes visible — and what stays outside the inquiry. THE SAME SITUATION An AI system entering an existing workflow. Same situation, different questions. QUESTION How do we automate this workflow? TASKS · PROCESS · CAPABILITY QUESTION What work is this workflow actually holding together? JUDGMENT · EXCEPTIONS COORDINATION · LOCAL PRACTICE ROC INQUIRY What changes when AI takes part in it? ? WHAT CHANGES · WHAT REMAINS WHAT IS STILL UNCLEAR DIFFERENT QUESTIONS MAKE DIFFERENT PARTS OF THE WORK VISIBLE. A SMALLER EXAMPLE “Why aren’t people using the AI?” SEES: USAGE “What are people still checking, correcting or protecting that adoption metrics do not show?” SEES: CHECKING · CORRECTION · PROTECTION · LOCAL PRACTICE One possible reframe — not a universal one.
The question is part of the work. The same AI situation is shown through three different questions, each directing attention to different parts of the work. HOW ROC INQUIRES The question is part of the work. How a problem is framed shapes what becomes visible — and what stays outside the inquiry. THE SAME SITUATION Same AI system. Different questions. QUESTION How do we automate this workflow? TASKS · PROCESS · CAPABILITY QUESTION What work is this workflow actually holding together? JUDGMENT · EXCEPTIONS COORDINATION · LOCAL PRACTICE ROC INQUIRY What changes when AI takes part in it? ? WHAT CHANGES · WHAT REMAINS WHAT IS STILL UNCLEAR DIFFERENT QUESTIONS MAKE DIFFERENT PARTS OF THE WORK VISIBLE. A SMALLER EXAMPLE “Why aren’t people using the AI?” SEES: USAGE “What are people still checking, correcting or protecting that adoption metrics do not show?”
This way of working did not begin with AI — an anonymized Korea field research case A past field research project for a global spirits brand in Korea. Twenty-two interviews and direct observation changed the explanatory framing of solo drinking and informed strategic work used across APAC. FROM PAST WORK This way of working did not begin with AI. Enter the field. Understand the world around the problem. Build an explanation a decision can use. A GLOBAL SPIRITS BRAND KOREA 22 INTERVIEWS + DIRECT OBSERVATION SIX-WEEK FIELD STUDY CONTINUOUS CAPABILITY FIELD INQUIRY CONTEXTUAL INVESTIGATION EXPLANATION BUILDING DECISION RELEVANCE INITIAL QUESTION PARAPHRASED How should the brand understand and reach women drinking alone? FIELD INQUIRY WHAT PEOPLE SAID WHAT PEOPLE DID Stated preference did not fully explain lived behavior. WHAT THE FIELD MADE VISIBLE PRIVATE RECOVERY EMOTIONAL FUNCTION SELF-PRESENTATION EXPLANATION SHIFT Not only — what do they want to drink? Also — what role does this occasion play in their lives? STRATEGIC IMPLICATION The resulting framework informed an RTD revival strategy adopted across five APAC offices. THE DOMAIN CHANGED. THE INQUIRY DID NOT. Consumer behavior then. AI-enabled work now.
This way of working did not begin with AI — an anonymized Korea field research case A past qualitative field project for a global spirits brand in Korea. Twenty-two interviews and direct observation changed how the consumer occasion could be explained and informed strategic work across APAC. FROM PAST WORK This way of working did not begin with AI. Enter the field. Understand the world around the problem. Build an explanation a decision can use. A GLOBAL SPIRITS BRAND KOREA 22 INTERVIEWS + DIRECT OBSERVATION SIX-WEEK FIELD STUDY INITIAL QUESTION PARAPHRASED How should the brand understand and reach women drinking alone? FIELD INQUIRY WHAT PEOPLE SAID WHAT PEOPLE DID Stated preference did not fully explain lived behavior. WHAT THE FIELD MADE VISIBLE PRIVATE RECOVERY EMOTIONAL FUNCTION SELF-PRESENTATION EXPLANATION SHIFT Not only — what do they want to drink? Also — what role does this occasion play in their lives? STRATEGIC IMPLICATION The resulting framework informed an RTD revival strategy adopted across five APAC offices. THE DOMAIN CHANGED. THE INQUIRY DID NOT. Consumer behavior then. AI-enabled work now.
CURRENT INQUIRY

What are you trying to understand
before you build?

ROC is currently interested in AI deployments where the technology is becoming operational, but the work around it is not yet fully understood.

Especially when the question sounds like:
“

We know what we want the AI to do.
But do we understand the work well enough?

“

The system works.
We’re less certain about what people now have to do around it.

“

We can measure adoption.
We’re not sure what those metrics are missing.

If you're deploying AI into real work — or investigating what changes when you do — I'd be interested to hear what you're seeing.

COMPARE NOTES
ryan@reasonofcreativity.com
Seoul · Working internationally