# Role Responsibilities, Processes, and Human–AI Collaboration

- Author: Andrew David W.Y.
- Source: https://yutengfei.com/en/journal/role-process-performance-model/
- Version: V2.0
- Published: 2026-11-11
- Revised: 2026-11-11
- Content status: Full text published
- Content type: Models (models)
- Research topics: Stakeholder Capital Governance (stakeholder-governance)
- Ecosystem perspectives: Operations & Management Ecosystem (operations) · Resource Ecosystem (resource)

## Use this resource

**Purpose**: Use Four-Way SOP to connect what each role delivers to business results. Work back from the outcome, organize execution by people and AI, check acceptance and readiness separately, and test changes before adopting them.

**Conditions for use**: Check that the subject, objective and constraints fit the task. Confirm any conditions that the article does not specify.

**Models · Elements, relationships and assumptions**

**Evidence and validation limits**: The first version dates to April 10, 2023. This article extends it to human–AI collaboration, two checks at handover and tested updates. Assess practical effects using operating records. No separate external-validation result is recorded for this article. Publication does not establish effectiveness; check the evidence, assumptions and limits in the text.

This is a readable, citable knowledge resource, not an installed Skill. When using it with AI, retain its sources and limits and review the output.

## Full article

The previous step delivered its result, and the receiving team accepted it. Why can't the next step start? Ask a little further and you may find that some data is missing, or that authorization hasn't come through. Record the acceptance decision now and start the next step once the conditions are in place. Once you know what's missing, you know who to ask.

When I look at a responsibility, I first ask who needs the result and what they will do with it. Then I ask what this role has to deliver, what it needs from the previous step, and what people, tools and permissions the company has to provide. Ask those questions about an actual piece of work and the connection between roles, processes and business results starts to become clear.

## 1\. Get clear on the result you need

On April 10, 2023, I drew the first version of the role-design diagram for Lianchuang Jicai. It connected work results, the transfer of value between roles, process coordination, requirements for upstream outputs and the assessment of objectives. This time I take the model further into human–AI collaboration. How should we combine people's abilities with those of AI? What can each do under its authorization? Who checks the result at handover? After the work is done, how do we know whether the approach is worth using again?

Start with the result the business needs. Work backwards to the processes and role responsibilities, then match the tasks to the capabilities they need. Do the work to an agreed standard and have someone check the deliverable. The next process can use it once its starting conditions are met. Finally, look at the business outcome and decide which tested practices belong in the next version of a standard, work manual or Skill.

Once an action is complete, check that it meets the delivery requirements of the role. Once the role delivers, check whether the downstream process actually uses the result. After the business process has run, there is a basis for assessing how much of the objective was achieved.

## 2\. Locate the four SOP scopes on the master diagram

[![Role Responsibilities, Workflow, and Performance Assessment](https://yutengfei.com/journal/role-process-performance-model/materials/en/original-2023.svg)](https://yutengfei.com/journal/role-process-performance-model/materials/en/original-2023.svg)

Figure A　Role Responsibilities, Workflow, and Performance Assessment · Redrawn from the source diagram

[Open at full size ↗](https://yutengfei.com/journal/role-process-performance-model/materials/en/original-2023.svg)

The four horizontal bands show role responsibilities, work manuals, performance SOP and performance assessment. They let us examine responsibility, execution guidance, workflow and evaluation. To understand the Four-Way SOP, look at where each scope appears on the diagram.

| SOP scope | Where to look on the master diagram | Question it answers |
| --- | --- | --- |
| Action SOP | The “noun + verb” in the work description | What do we do to which object, and what should it produce? |
| Role SOP | The set of actions that makes up a role responsibility | What must this responsibility deliver, and which tasks fulfill it? |
| Business Process SOP | The nodes for Processes 1, 2 and 3 | What does this process need as input, who works together, and what does it hand over? |
| Performance SOP | The end-to-end workflow linking the business processes to the objective | Which processes have to work together to deliver customer value and meet the business objective? |

When I look at a performance SOP, I start by asking which business processes produce the intended result. Revenue, costs and customer acceptance come into the performance assessment. Processes 1, 2 and 3 illustrate how one process connects to the next. The actual business determines the number of processes and who needs to work together.

The master diagram places the role SOP in the role responsibilities band. Figure 2 takes a responsibility and breaks it down into tasks a role can take on, deliver and have checked. A role may have several tasks, and a business process may need several roles to work together. List what each responsibility must deliver and the tasks that deliver it. Then decide which capabilities and Skills suit each task.

A noun and verb can identify an action. The noun names the object and the verb says what to do. Before handing it to a person or AI, add the inputs, operating requirements, output, quality standard, permissions and exception handling. The work manual and Skill carry those requirements into practice for the person and the AI respectively.

## 3\. Follow a piece of work through the overview

[![Overview — From responsibilities and processes to business capability](https://yutengfei.com/journal/role-process-performance-model/materials/en/01-structure.svg)](https://yutengfei.com/journal/role-process-performance-model/materials/en/01-structure.svg)

Figure 1　Overview — From responsibilities and processes to business capability · Redrawn from the source diagram

[Open at full size ↗](https://yutengfei.com/journal/role-process-performance-model/materials/en/01-structure.svg)

Figure 1 starts with the value objective and the business result. In design, work backwards from the result to the processes, the deliverable of each role and the specific actions. In execution, move forward from each action's output. Have it checked, then find out whether the next step can use it.

-   **Role responsibilities and organization.** Decide who will use the result, then who is accountable, who does the work, who receives it and who approves it. Assign responsibility for arranging the process and providing the capabilities too.
-   **Work manuals and capabilities.** Write the result requirements into an SOP people can use while doing the work. People and AI work to those requirements within their respective permissions.
-   **Coordination between processes.** Hand the previous step's deliverable to the next one. After acceptance, confirm that the inputs and the other starting conditions are ready before proceeding.
-   **Evaluation and retained learning.** Review the deliverables, handovers and business feedback. Find the problem, test a change and then update the standard, manual, Skill and training.

## 4\. To define a role, ask who needs the result

[![Work back from the result to define responsibilities](https://yutengfei.com/journal/role-process-performance-model/materials/en/02-role-design.svg)](https://yutengfei.com/journal/role-process-performance-model/materials/en/02-role-design.svg)

Figure 2　Work back from the result to define responsibilities · Redrawn from the source diagram

[Open at full size ↗](https://yutengfei.com/journal/role-process-performance-model/materials/en/02-role-design.svg)

When defining a role, put people's day-to-day work alongside the results the business actually needs. It becomes easier to see which work should continue and where no one owns a result.

1.  **Define the result.** Write down what has to be delivered, who uses it, what they use it for and what qualifies for acceptance.
2.  **Check the interfaces.** What can the previous step supply? What else is needed here? What does the next step actually need?
3.  **Assign tasks and responsibilities.** Name the person accountable for the result, the executor, the receiver and the approver. Identify who handles exceptions.
4.  **Provide the capabilities and permissions.** Decide which work people do, where AI can help and where other roles need to contribute. Confirm the data, tools, permissions and time together.
5.  **Put the assignments in the work manual.** Group the tasks a role takes on. Say where they come from, who handles the handover, what to deliver and how to check it.

If a role is taking responsibility for another deliverable, agree on the resources and authority that go with it.

## 5\. Managers prepare the conditions, and executors deliver to the standard

[![Organize capabilities around the result](https://yutengfei.com/journal/role-process-performance-model/materials/en/03-capabilities.svg)](https://yutengfei.com/journal/role-process-performance-model/materials/en/03-capabilities.svg)

Figure 3　Organize capabilities around the result · Redrawn from the source diagram

[Open at full size ↗](https://yutengfei.com/journal/role-process-performance-model/materials/en/03-capabilities.svg)

The master diagram puts critical success factors, process management and moving the work forward on the manager's side. In practice, organizing the work means getting four things done. Define the result, arrange the process, provide the capabilities, and put authorization and coordination in place. Agree on the delivery standard, deadline, cost, dependencies and risk limits before starting. Decide who can change the arrangement in an exception. This is how the master diagram's ideas about management through systems and leadership through change show up in the work.

The people doing the work have to get it right and consider efficiency. They bring professional skills, judgment, communication and exception handling. Where appropriate, AI retrieves information, generates material, analyzes it and calls tools. Internal and external collaborators deliver through the agreed interfaces and own their part. Once data, knowledge, tools, permissions, budget and time are available, check how well those capabilities work together on the actual task.

Managers explain the plan and its trade-offs. Executors report missing inputs, exceptions and actual effects while there is still time to act. This is where “managing up” fits in the diagram. The same person may organize one task and execute another. For this task, make clear who owns the objective, authorization, delivery and exceptions.

## 6\. People and AI work to one standard on the same task

[![Execution with SOPs, work manuals, and Skills](https://yutengfei.com/journal/role-process-performance-model/materials/en/04-human-ai.svg)](https://yutengfei.com/journal/role-process-performance-model/materials/en/04-human-ai.svg)

Figure 4　Execution with SOPs, work manuals, and Skills · Redrawn from the source diagram

[Open at full size ↗](https://yutengfei.com/journal/role-process-performance-model/materials/en/04-human-ai.svg)

I start by lining up the kind of work with this particular task. The SOP is the shared standard for a class of work. It covers inputs, steps, outputs, quality, permissions and exceptions. This task also needs a specific deliverable, owner, actual inputs and due date. Routine, recurring, triggered and reactive describe how tasks arise. More than one can apply to the same task.

A work manual should help a person understand what is required, how to act and judge the result, and how to communicate, confirm or take over when an issue arises. A Skill gives AI an applicable method to use with the tools and data it is authorized to access. A task can draw on several capabilities. Decide how to combine them from the needs of the particular deliverable.

Before starting, verify inputs, capabilities and permissions. If something is missing, pause and tell the owner what needs to be confirmed. Start when the conditions are ready. During execution, people handle professional operations, judgment, communication and exceptions. AI checks the inputs, uses appropriate capabilities and submits its result together with the evidence behind it. Specify which instructions require human confirmation and which exceptions must go back to a person.

Submit the result with its record. The record should identify the task, input version, executor, capabilities used, basis for the output, current status and exceptions. Send AI-generated results to the receiver for the agreed checks too. Once a result has been used in the business, see whether it helped achieve the objective.

## 7\. Make both decisions at handover

[![Acceptance and readiness at handover](https://yutengfei.com/journal/role-process-performance-model/materials/en/05-handover.svg)](https://yutengfei.com/journal/role-process-performance-model/materials/en/05-handover.svg)

Figure 5　Acceptance and readiness at handover · Redrawn from the source diagram

[Open at full size ↗](https://yutengfei.com/journal/role-process-performance-model/materials/en/05-handover.svg)

Before Processes A and B work together, agree on who supplies and receives the deliverable, what it includes, when it is due and who checks it. Agree on the acceptance standard and where to take exceptions or disputes.

**First, check this deliverable.** Process A supplies the result, its evidence and its version. The receiver checks it against the standard. If it passes, record acceptance. If something is missing, identify the gap, complete or rework the result, then submit it again.

**Second, check whether Process B can start.** With the deliverable accepted, check whether valid inputs, authorization, resources and other dependencies are ready. If they are, start. If something is missing, wait, coordinate with the people involved or take it to someone who can make a decision.

When a process stalls, I ask which of the two checks it is stuck at. If there is a problem with the result, describe the issue and get it fixed. If the result has already been accepted, secure what the next step needs. To see whether a role's deliverable created value, trace who received it, who used it and what they achieved with it. Individual economic contributions and entitlements require fuller records and the applicable rules.

## 8\. Carry the learning from one job into the next

[![Evaluate, diagnose, test, and update](https://yutengfei.com/journal/role-process-performance-model/materials/en/06-validation.svg)](https://yutengfei.com/journal/role-process-performance-model/materials/en/06-validation.svg)

Figure 6　Evaluate, diagnose, test, and update · Redrawn from the source diagram

[Open at full size ↗](https://yutengfei.com/journal/role-process-performance-model/materials/en/06-validation.svg)

At the bottom of the master diagram, work delivery assessments, management by objectives and KPIs all help assess results. Pages 13–15 of the _BS Diagram Collection_ also discuss updating SOPs and manuals in light of why work succeeded or failed, then putting the changes into practice through training. Figure 6 goes further into verifying the cause and deciding which changes qualify for a formal release.

First identify the level where the problem occurred. Was the action done correctly? Did the role fulfill its responsibility? Did the handover work? Did business results improve? Look separately at quality, time, cost and risk, and compare what you find with business feedback. Choose KPA, OKR or KPI as needed, taking into account where the measure applies and how much work it takes to use and maintain.

Start with the facts. Were inputs missing? Were the capabilities sufficient? Were the tools suitable? Did the processes fit together? AI can organize evidence and suggest possible causes. Test those explanations against actual records. Once the evidence supports a change, write down what to change, why, what you expect to improve, what it will take and when it applies.

Run a small trial. Compare the old and new results, including side effects and operating costs. If the evidence shows improvement and the risk is manageable, seek approval, assign a version and arrange training and rollback. If the effect is still unclear, keep the current version while you check further or revise the hypothesis. Assign an owner after release. When conditions change, review whether to revise or retire the method.

In a review I look at two things. Did the business result improve this time? Can we use the tested method again? Once it works in subsequent jobs, there is evidence that the company has gained a capability.

## 9\. Walk one piece of work through all seven figures

Take a job that is already in progress. Locate it in the business on the master diagram, then use the atlas to check its responsibilities, execution and handovers. Write down what you can confirm now. Assign each missing condition to a specific person and action.

| Step | What to clarify | Figure |
| --- | --- | --- |
| Set the objective | Who needs the result, what the company needs to achieve and which business processes connect them | Master diagram, Figure 1 |
| Define responsibilities | What each process delivers and who owns, executes, receives and approves the work | Figure 2 |
| Provide capabilities | How people, AI and collaborators work together and whether data, tools, time and authorization are sufficient | Figure 3 |
| Set execution standards | What goes in the SOP, work manual and Skill, which versions are recorded and who handles exceptions | Figure 4 |
| Define handover | What qualifies for acceptance and what the next step needs to start | Figure 5 |
| Revise after running the work | Verify causes from actual evidence, run a small trial and then decide whether to release, retain or stop | Figure 6 |

For any responsibility, ask which process gets its result and which objective it supports. You can also trace back from an objective to the processes, responsibilities and actions that need to work together. If they don't connect, go back to the job in question. Clarify the requirements, rearrange the work or verify whether it is still needed.

## 10\. Where this method applies

This method is for designing role responsibilities, business processes and execution involving people and AI. Business strategy and capital governance have their own decisions, which depend on the circumstances. How far to break down a task and how much AI should do depend on the deliverable, available capabilities, authorization and risk.

This version sets out the structure, responsibilities and ways to test them. To know whether it reduces costs, improves performance or yields a method we can keep using, put it to work and examine the delivery records and business results.

* * *

### Versions and sources

The positions of the four SOP scopes follow _Role Responsibilities–Process–Performance Coordination Topology_, V3 annotated edition. The execution mechanism draws on the six-page _Role Responsibilities–Processes–Human–AI Collaboration Model_ atlas, V2. In the master diagram, the role SOP corresponds to a role responsibility. Figure 2 breaks that responsibility down into tasks a role can own, deliver and have checked.

The historical prototype appears on pages 12–15 and 22 of the _BS Diagram Collection_. This study extends it to cover the allocation of human and AI capabilities, authorization and human takeover, the two decisions at handover, and tested releases with a way to roll back.

[Upgrade atlas 00–06 (Draw.io, Chinese)](https://yutengfei.com/journal/role-process-performance-model/materials/upgrade-00-06.drawio) · [Annotated first-version redrawing (Draw.io, Chinese)](https://yutengfei.com/journal/role-process-performance-model/materials/original-2022.drawio)

**Further reading**

[Four-Way SOP](https://yutengfei.com/en/journal/four-way-sop/) · [From responsibilities to work manuals](https://yutengfei.com/en/journal/co-builder-role-handbook/) · [Process handover and starting conditions](https://yutengfei.com/en/journal/process-handover-and-entry-conditions/) · [From SOP to Skill](https://yutengfei.com/en/journal/sop-to-ai-skill/)

**Original method · 2023-04-10** · [A Method for Identifying and Designing Role Responsibilities](https://yutengfei.com/en/journal/role-responsibility-design-method/)

[CC BY-NC-ND 4.0](https://creativecommons.org/licenses/by-nc-nd/4.0/)

Applies to original text unless otherwise stated; excludes third-party materials. Downloading or machine reading does not extend the existing license or separately authorize derivatives, commercial use or model training.

[AI usage guide](https://yutengfei.com/en/journal/ai/)