Research and practice · Amber Liu
Amber Liu focuses on how enterprises turn AI-assisted work into repeatable operating capability. The work starts with one business workflow, defines its results and responsibilities, and tests whether someone else can deliver to the same standard.
Stakeholder capital theory asks what resources, capabilities, and relationships can accumulate when people contribute and create value together. It also asks how contributions, rights, and governance fit together.
In enterprise work, the practical questions are who needs the result, who delivers it, what standard applies, and how the next team can use it.
Yu Tengfei’s V2.0 research outlines a design to test. Its evidence fields, permission rules, and cross-organization interfaces still need to be refined through use.
The starting point is a specific operating problem along a real workflow. The work asks where results, handoffs, and decisions fail to connect.
Six topics help structure the discussion:
Six areas of focus:
01 · Define value
Name the stakeholder, the result they need, and how the result will help them. Decide what a successful delivery would let them do.
02 · Design the handoff
Check the previous result against its standard. Separately decide whether the next process has permission, resources, and the other conditions it needs to start.
03 · Set responsibilities
Use the same process requirements to write role responsibilities, the work manual, and performance criteria. Agree how people and AI will work and who takes over when needed.
04 · Verify and reuse
Compare the result with the baseline, including waiting, rework, human takeovers, and cost. Have a second person use the manual or Skill to check that the method works again.
Update the SOP, responsibilities, manuals, Skills, and training together if the test supports the change. Document the scope, owner, and way to roll it back.
Pick a workflow with an observable result and one person who can take responsibility for the test. Agree how to establish a baseline and when to stop.
For the first discussion, describe what you need the workflow to deliver, which team uses the result, and where the work currently gets stuck.
Read the framework overview, or share a workflow on the contact page.
These questions separate what a pilot can show at each stage.
Did the work actually happen? Check that the task ran under the conditions and permissions you agreed to, and record any exceptions or manual takeovers.
Did the result pass acceptance? Compare the delivery against the same standard, whether a person, AI, or both did the work.
Was it used downstream? A passed result may still wait for permission, resources, or dependencies. Check that the next team was able to use it.
Did the business result improve? Look for changes against the baseline, as well as extra costs or side effects. Check what the responsible team could control and allow time for the effects to appear.
Bring one workflow to discuss. We can define a practical first test and what you would need to decide whether to continue.
Book a Consultation