For faculty and staff5-minute readDepth 1–2: quick or light check

Is AI worth it for my work?

The framework judges AI on Cost, Use, and Outcomes. The same three questions work for one person and one task. Answer them honestly, then decide.

A

Cost: what does it take?

For an individual, the main cost is your time. The tool may be free to you, but the minutes are not.

What to count
  • Learning the tool and building prompts that work
  • Time spent checking every output
  • Time fixing errors or redoing work
  • Any fee, if you pay for a tool yourself
Common mistakeTiming only how fast the AI generates. A draft in 30 seconds that takes 20 minutes to fix is not a saving.
Ask yourselfIncluding checking and fixing, how many minutes does this task take with AI?Full cost guidance →Full checklist →
B

Use: how do I actually use it?

Use is not whether you have access. It is which tasks AI is part of, how often, and how much it does.

What to count
  • The specific tasks you use it for
  • How often, on each task
  • Its role: it drafts, it assists, or it acts for you
  • Whether you kept using it after trying it
Common mistakeSaying “I use AI” without naming the task. Value depends on the task, so a tool can be worth it for one and not another.
Ask yourselfWhich of my tasks is AI really part of, and what role does it play in each?Full use guidance →Full checklist →
C

Outcomes: what changed?

Outcomes are the result, not the activity. Faster is one kind. Better is another, and the one that usually matters more.

What to count
  • Time per task, before and with AI
  • Quality: errors, rework, or feedback from others
  • Effect on the people you serve, such as students or colleagues
  • Effect on you: easier, or more tiring to check
Common mistakeCounting output instead of outcome. Producing more drafts is not the same as better work.
Ask yourselfCompared with how I did it before, is the result faster, better, or neither?Full outcomes guidance →Full checklist →

Put it together: two weeks, one task

Pick one task. Jot down the same few numbers before and with AI. Example from the meeting summaries case.

DomainWhat was notedBeforeWith AI
A · CostMinutes per summary, including review3514
B · UseSummaries drafted with AI0 of 109 of 10
C · Outcomes: qualityMissed action items22
C · Outcomes: youEasier or harder?—Easier; owners need checking

Reading it: 21 minutes saved per summary, used on most of them, quality unchanged. Worth keeping, with a check on who owns each action item.

Then decide

The individual-level question in the framework: should I use AI for this task, and under what conditions?

Keep usingCost is lower or the result is better, and you use it regularly.
Use with conditionsIt pays off on some tasks, or only with closer checking. Limit it to those.
StopChecking and fixing cost more than it saves, or the result is no better.

Time saved is capacity for other work, not a cut. Why that matters · If a use spreads to your team or unit, the same questions get asked there with formal measures. The unit guide · Decision guide · How teams, units, and the institution decide

Examples of individual use

Seven people, seven tasks, the same three questions. The first is a full walk-through. Not every answer is “keep using.” Illustrative examples.

Application developer · Full walk-throughWriting and refactoring code with a coding assistant
AAbout 35 min of prompting and fixing per change
BUsed on 14 of 18 changes; not for security code
C3.2 → 2.5 days per change; defects unchanged
Continue with conditionsSee every step →
Instructor

Drafting quiz questions from lecture notes

A10 min to generate and fix 10 questions, vs 40 min by hand
BEvery weekly quiz; AI drafts, instructor edits
CSame quality after edits; 2 of 10 needed rewriting for accuracy
Keep usingClear time saving, quality held after checking.
Academic advisor

Summarizing a student’s file before a meeting

AFast, but each summary needed a full re-read to check
BTried on 6 meetings
CNo time saved; one summary missed a hold on the account
StopChecking cost as much as reading the file, and an error slipped through.
Researcher

Cleaning and commenting analysis code

AAbout 15 min per script, including review
BUsed on most scripts; AI suggests, researcher accepts or rejects
CScripts easier for students to follow; bugs caught in review
Keep usingBetter result for the people who use the code, at low cost.
Department coordinator

Drafting routine email replies

A2 min per reply instead of 5
BUsed for scheduling and FAQ replies only
CFaster replies; tone needed adjusting on sensitive messages
Use with conditionsWorth it for routine replies. Not for anything personal or sensitive.
Faculty member

Translating course announcements

AMinutes per announcement, plus a check by a fluent colleague
BWeekly announcements to a multilingual class
CMore students read them; two phrases corrected in the first month
Use with conditionsHelps students, but only with a fluent reviewer.
Staff analyst

Writing first drafts of an annual report

ASaved drafting time, but added hours fact-checking figures
BUsed once
CDraft read well, but three figures were wrong
StopErrors in numbers are costly to find. Faster to write from the data directly.

Safe use comes first

Good numbers don’t count if a basic check fails. These are the framework’s five lenses, at your level.

Approved toolCampus tools, campus account
Allowed dataNo confidential data unless cleared
You check itVerify before sharing or acting
Better than beforeIf not, stopping is fine
Report problemsErrors, bias, or data slips

If your use changes, such as new data, more people, or AI acting on its own, check with your unit lead before continuing. Record your use

Common questions

Do I have to report my numbers?No. The self-check is for your own decision. Teams and units measure formally; you can share what you learned if you choose.
Am I being monitored?Use is counted in aggregate, by tool and role. It is visibility, not surveillance.
If AI saves me time, does my job shrink?No. Saved time is capacity for work you and your team choose, not a budget cut.