Diagnosing adoption opportunity and work context

Review questions

Integrated summary: Diagnosing adoption opportunity and work context

We turned a demand to “use AI” into a defined job, outcome, guardrail, and non-AI alternative. The workflow map locates the suggestion, source of truth, and human authority; it also exposes waiting time that faster drafting cannot remove. A baseline fixes start and end points, denominators, and case mix. An overall mean can rise while each group improves when the mix changes. Stakeholder diagnosis separates skill gaps, incentives, access problems, and justified objections, testing competing explanations instead of labeling people resistant. For pilot selection, data permission, authority, and harm control are eligibility conditions before benefit ranking. Scores then need sensitivity checks and comparison with non-AI repair. The resulting decision names a bounded first use case, evidence that would reverse the choice, and a stop owner. Success in that use case does not transfer automatically to recruitment or customer decisions.

Field observation expands the picture to work missing from system records, including expert help and pre-ticket corrections. Separate effort from waiting and test a competing explanation before choosing a tool. Between branches, compare input, purpose, authority and case mix; test changed components while preserving sound evidence for unchanged parts. Commitment decisions weigh learning against waiting and reversal costs, treating past expenditure as accounting and learning information. Adoption also requires effective access to sources, help and escalation. Together these links support a bounded trial whose result can change the organizational decision.

Review questions

Test your understanding with these questions. Marking and saved results are available with a free account.

  1. Question 1More drafts alone prove that time to correct resolution has fallen.
  2. Question 2In a financial complaint workflow, a model can draft while an authorized person retains closure authority.
  3. Question 3A changed case mix can make an overall average misleading even when each class improves.
  4. Question 4Low use by a team without access permission proves psychological resistance.
  5. Question 5Data-use rights and decision authority should be checked before ranking eligible use cases by expected return.
  6. Question 6An executive sponsor may leave pilot stop authority unassigned until the first incident.
  7. Question 7A leader should disclose what remains undecided, when it will be decided and how staff can challenge it.
  8. Question 8Controls for a public-message assistant automatically fit a system advising on hiring decisions.
  9. Question 9When a role shifts to output review, time, training and performance criteria should reflect the new work.
  10. Question 10“Review cases when needed” is a sufficient operating agreement without triggers or an owner.
  11. Question 11If the ticket system changes alongside the assistant pilot, every later gain can be attributed solely to the assistant.
  12. Question 12Checking source version and access permission belongs in review instructions even when an output sounds confident.
  13. Question 13Passing a definitions quiz alone establishes ability to handle high-impact exceptions.
  14. Question 14When a pilot discloses information, containment precedes routine collection of more usage data.
  15. Question 15A critical-error limit can be offset by a sufficiently better average completion time.
  16. Question 16Success with an expert team and simple requests does not prove immediate fit in every branch.
  17. Question 17A usable export right for decision records reduces dependence on a single vendor.
  18. Question 18Local adaptation permits deleting a shared escalation rule when it shortens the message.
  19. Question 19Rewarding generated-reply volume can increase use even if rework rises.
  20. Question 20A favorable average can conceal harm concentrated in a subgroup or shift.
  21. Question 21A sound adoption scorecard shows usage alone because outcomes always arrive later.
  22. Question 22Improvement after launch is an observation, but it does not isolate the tool effect if other changes coincide.
  23. Question 23Fewer incident reports with steady errors calls for checking the reporting channel and incentives, not declaring risk gone.
  24. Question 24Handing over an instruction file without an owner and maintenance cadence sustains organizational capability.
  25. Question 25An integrated board decision joins net value, harm limits, uncertainty and the evidence that would change the decision.
  26. Question 26The dashboard starts at ticket creation, but staff prepare request data for five minutes beforehand. Which boundary fits service measurement?
  27. Question 27A researcher needs the sequence of handoffs across three departments for a request lasting two days. Which method best fits?
  28. Question 28All observation snapshots were collected during a quiet morning period, then search share was generalized to the day. What is the best repair?
  29. Question 29Drafting takes six minutes, waiting one hour and editing two minutes. What is the known active effort?
  30. Question 30An observer writes: the employee opened three screens, so the system caused distraction. What is a better record?
  31. Question 31Employee effort falls from 200 to 170 minutes while expert support rises from 40 to 70. What is the correct description?
  32. Question 32Observation needs request type, work stage and duration. The observer proposes full customer-screen screenshots just in case. Which option fits?
  33. Question 33Two branches call their service subscription renewal. One uses a fixed price and the other applies contract discounts. What should be compared first?
  34. Question 34A renewal document arrives as an image rather than text, with policy and approval authority unchanged. Which test addresses the direct difference?
  35. Question 35Use shifts from summarizing a request to recommending rejection, while the document remains the same. What is the main change?
  36. Question 36Branches share a policy, but permission to read a contract annex differs. Which test fits?
  37. Question 37Routine accuracy is 78 of 80 and exception accuracy 18 of 20. The new branch has 20 routine and 80 exception cases. What is expected accuracy if both rates transfer?
  38. Question 38A manager calls weighted expected accuracy a proven result for the new branch before testing it. What correction is needed?
  39. Question 39The transfer condition says complete readiness without defining it. Which condition is more operational?
  40. Question 40A discount expires today while customer-data handling is unknown. Which step is available?
  41. Question 41The team stops a tool after an incorrect reply reaches a customer. What remains to reverse the consequence?
  42. Question 42A contract says cancelable but does not clarify refunds for the year. What is the sound assumption?
  43. Question 43The service handles one hundred requests weekly and expects two released minutes per request. What is expected over two weeks?
  44. Question 44A team proposes another trial but will buy regardless of its result. What is the problem?
  45. Question 45Buying yields ten thousand on success and loses fourteen thousand on failure, each assumed equally likely. What is expected value?
  46. Question 46A manager assumes a fifty-percent success probability without evidence and presents a precise result. What should change?
  47. Question 47A phone user has low assistant use, but the escalation control is hidden on that device. What is the first diagnosis?
  48. Question 48An employee is authorized for a document but her screen reader cannot interpret its image. Which repair fits?
  49. Question 49A worker recognizes the button name but misinterprets the exception. Which test fits?
  50. Question 50A team requests detailed medical history before testing keyboard task navigation. Which information better fits the need?
  51. Question 51An adjustment works for one screen-reader user. What can be concluded?
  52. Question 52The assistant is online at night but the only approver leaves in the afternoon. What is the barrier?
  53. Question 53Management wants a shorter message explaining an eligibility exception. What condition should apply?