PrerequisiteCompleted Module 1 (Executive Kickoff), with an agreed approved version, RACI, and stop authority
0 / 3 completed
Learning objectives
Re-walk how the current process actually runs today, instead of reusing a process map from IL 101/201 that may already be stale.
Record work-difficulty variance across case types and staff, rather than treating the baseline as a single uniform number.
Compile the concrete reasons users are reluctant to use AI assistance, sorted into skill, trust, process, and incentive types.
Re-walk the current state, see the variance, hear the resistance
60 min
Re-document the current process
IL 101 and IL 201 already mapped the current process once, but weeks or months typically pass between charter approval and pilot kickoff, and the process may have changed. IL 301's process immersion is not redrawing the diagram: it is re-verifying whether what people actually do today still matches what the document says.
Current-process re-verification checklist
Item
Question to re-verify
Step order
Does the actual execution order still match the old process map? Have new exception-handling steps been added?
Processing time
How long does each step actually take now? Has staffing turnover changed it?
Review and rework
What are the current review and rework rates? Higher or lower than the IL 101 baseline log recorded?
Systems and tools
Has the system changed, or have new tools or forms been introduced?
Warning
Common mistake: reusing IL 101's baseline log directly as the pilot's comparison baseline without confirming whether the current state has changed. If it has, resample: otherwise the "improvement" seen later may just be an artifact of a stale baseline.
Key takeaway: The point of process immersion is to re-verify whether the current state still matches old documentation, not to assume nothing has changed.
60 min
Record work-difficulty variance
Treating the whole process's handling time as a single average masks the variance that matters most: a novice and a veteran staff member may differ threefold in handling time, and a simple case and a complex case may differ tenfold. The pilot's comparison baseline must segment this variance first, or the average will make AI's effect look overstated or understated.
Common dimensions for segmenting work-difficulty variance
Dimension
SYN-CS-001 example
Case type
The six categories: normal, missing_context, conflicting_sources, outdated_source, no_answer, restricted_access: each carry different difficulty
Staff tenure
Handling-time difference between a new hire (<3 months) and a veteran (>1 year) on the same case
Peak vs. off-peak
Whether processing quality drops under peak-time pressure
System stability
How the manual process copes when the system is down or slow
Info
The point of recording variance is not to cherry-pick the hardest cases to dismiss AI's value, but to let Module 6's evaluation compare cases of matched difficulty, avoiding a false read of "AI only handled the easy cases" as "AI improved efficiency across the board."
Key takeaway: Without segmenting work difficulty first, later evaluation can easily mistake "the case mix got easier" for "the AI got better."
60 min
Find the real reasons for reluctance
A pilot commonly fails not because the technology can't do the job, but because users never really use it, or quietly route around it. Rather than discovering low adoption only after the pilot ends, listen for the concrete reasons behind resistance during immersion.
Four common types of reluctance to use AI assistance
Type
Concrete signal
Response direction
Skill
Unfamiliar with the new interface, unsure how to judge whether an AI suggestion is trustworthy
Training, and surfacing citable sources for cross-checking
Trust
Previously misled by a wrong suggestion, worried about being blamed for AI's error
Show confidence transparently, with a clear human sign-off step
Process
AI assistance turns one step into two (check the AI suggestion, then redo the work anyway)
Redesign the workflow so AI genuinely reduces, not adds, steps
Incentive
Worried that higher efficiency will shrink headcount or workload allocation
Align with the sponsor on staffing policy at kickoff and communicate it openly
1Exercise: run three synthetic interviews (using SYN-CS-001) Interview three synthetic customer-service staff at "Demo Service Company A": a new hire, a veteran, and someone who was previously complained about following a system issue.
2Interview 1: new hire Record her doubts about AI suggestion trustworthiness: is this skill or trust type?
3Interview 2: veteran Record whether he feels AI assistance turned the workflow into two steps: is this process type?
4Interview 3: previously complained-about staff Record his view on "who is responsible if the AI suggestion is wrong," and judge whether it is trust type or incentive type.
5Compile an immersion summary Classify each of the three interviews' concerns and write one concrete response direction for each.
Check yourself
A veteran customer-service staff member says, "Even with the AI suggestion, I still have to look it up myself to be sure: it feels like doing the work twice." Which type of reluctance does this mainly reflect?
Key takeaway: Reluctance to use AI is rarely simple "resistance to change": sorting it into skill, trust, process, and incentive types lets Module 4's solution design target the actual cause.
