The big idea: AI can widen your thinking and speed up your work, but your team owns every claim: name the tool, share the prompts, verify everything and keep client data out of public tools.
In 60 seconds
- AI can widen your thinking and speed up your work, but your team owns every claim and every line.
- Four rules apply: integrity, attribution, critical thinking and privacy. Name the tool, share your prompts and verify every source yourself.
- Never paste confidential client, company or personal data into a public AI tool. When in doubt, leave it out.
Do this: keep an AI log as you go, and end every deliverable where you used AI with the signed AI statement from the syllabus.
How to use AI in consulting research
The BUAD 497 syllabus welcomes AI for ideas, drafting, research support and practice. Use it to think wider and work faster. Then do the part only your team can do: check it, judge it and own it. Four rules from the syllabus keep your work trustworthy.
Integrity
AI never stands in for your work, invents sources or helps you skip the learning.
Ask: Could I explain and defend every line?
Attribution
Name the tool, say how it helped and share the prompts you used.
Ask: Could someone repeat what we did?
Critical thinking
Review it, verify it, then decide what to keep. You are 100% responsible for accuracy.
Ask: What did we check, and against what?
Privacy
No sensitive, personal or proprietary information goes into AI tools.
Ask: Would our client be fine seeing this pasted here?
The client test
Picture your client asking, “How did you get this?” If you could walk them through every step, AI steps included, without wincing, you are using AI the right way.
Right way, wrong way
Appropriate uses
- Brainstorm angles. Ask for possible research angles, then test them with your own interviews and action research.
- Summarize, then fact-check. Get a quick overview of public trends, then find and read the original sources.
- Generate options. Ask for ten names or headlines, then your team evaluates and picks.
- Practice. Role-play an interview with AI as the participant, then reflect on what you would ask differently.
Inappropriate uses
- Unreviewed work. Turning in AI text without review, edits or acknowledgment.
- Made-up facts. Using AI to fill in data, citations, quotes or case facts.
- Outsourced thinking. Letting AI do your reasoning, analysis or problem-solving.
- Leaked data. Entering confidential client, company or personal data into public AI tools.
Where AI fits in the design process
Inspiration
Good for: drafting an interview guide, finding questions you missed, practicing an interview, learning a new industry’s vocabulary fast.
Watch out: AI has never met your users. It cannot replace talking to them.
Ideation
Good for: getting volume, reframing a How Might We question, pushing past your first ten ideas, playing devil’s advocate.
Watch out: AI ideas trend toward the average. The insight that makes an idea fit comes from your research.
Implementation
Good for: stress-testing a Business Model Canvas, listing risks for a pilot, drafting success measures to debate.
Watch out: Never let AI invent costs, revenue or market numbers. Get real ones or say they are missing.
Privacy: what never goes into a public AI toolA stop, ask, go list for client work
Many public AI tools can store what you type, and some may use it to improve their models, depending on the tool and your settings. Treat anything you paste as something that could leave your control.
Stop
- Client internal documents, financials, strategy or contracts
- Anything marked confidential or shared under a nondisclosure agreement
- Interview recordings or transcripts with names
- Personal details: names, emails, phone numbers, health, student records
Ask first
- Interview notes with names and identifying details removed
- Anything the client told you in a meeting
- Photos from site visits
Watch out: removing a name is not always enough. A job title plus a small organization can identify someone.
Go
- Public information, like the client’s website or published reports
- Your own ideas and drafts
- General questions, like “What usually keeps volunteers coming back?” or “How do hospitals usually measure discharge delays?”
Ask your faculty, and your client, which tools (if any) are approved for client material. When in doubt, leave it out.
Verify before you trustA five-step check for anything AI tells you
- Find the original. Locate the actual article, report or data set, ideally through the library databases.
- Read it. Not only the title or abstract. Make sure it says what AI claimed it says.
- Match the details. Authors, year, numbers and quotes must match exactly.
- Check the date and the source. Old data and weak sources hurt your credibility with the client.
- Log it. Note what you checked in your AI log so your statement is easy to write.
If you cannot find the original, the claim does not go in your work. Say what is missing instead.
Prompts that keep your thinking in chargeWeak prompts next to better ones
| Weak prompt | Better prompt | Why it is better |
|---|---|---|
| “Write our insights.” | “Here are five de-identified quotes. What patterns might I be missing? Point to the quote behind each one.” | You stay the analyst. AI only widens your view, and every pattern ties to your evidence. |
| “Give me statistics on food insecurity in our county.” | “What kinds of public sources usually report local food insecurity data? I will look them up myself.” | You get a search map, not numbers you would have to trust blind. |
| “Is our idea good?” | “Argue against this idea as a skeptical volunteer coordinator. Give me your three strongest objections.” | It surfaces risks to test with real people, instead of flattering you. |
| “Analyze our client’s scheduling spreadsheet.” | “Here is a made-up table with the same columns as our data. Write the spreadsheet steps to compare wait times by shift. We will run them on the real data ourselves.” | Client data stays private, you learn the method and you can check every number. |
| “Write interview questions.” | “Here is our design challenge. Draft open questions about past behavior, then flag any that are leading.” | It follows good interview practice, and you still edit the guide. |
Keep an AI log as you goSo your statement takes two minutes, not two hours
Keep a shared table in your team folder. Add a row any time AI meaningfully shapes your work.
| Date | Who | Tool | Prompt | What we used | How we checked |
|---|---|---|---|---|---|
| Sept 9 | Teammate A | ChatGPT | “Draft 12 open questions about why volunteers stop coming. Flag leading ones.” | 4 questions, rewritten | Tested in 2 practice interviews |
Example only. The log is also where you paste full prompts, which the syllabus requires in your attribution.
Try it: AI Use Statement Builder
Tool · BUAD 497
AI Use Statement Builder
Mark how your team used AI. Get the statement the syllabus requires and any fixes before you submit.
- 1Mark uses
- 2Fix flags
- 3Copy
Your AI statement
Not ready to submit. Fix the Stop items with your faculty first. Never remove a use from the statement to hide it.
All checks
Put it to work: your AI statement in every deliverable
Any BUAD 497 deliverable where your team used AI ends with the AI statement from the syllabus. The HCD Technical Report has its own AI Utilization Statement section at the end, plus a short reflection on how your team used AI as a research tool. Build the habit in week one and the statement writes itself.
Build it in five steps
Agree on team AI rules
At kickoff, decide which tools you will use, what never goes in (open the privacy panel above) and who double-checks AI-assisted work.
Your deliverable: a three-line “How we use AI” note in an early Weekly Consulting Brief.
Log as you go
Add a row to your AI log any time AI meaningfully shapes your work. Paste the full prompt.
Your deliverable: an up-to-date AI log in your team folder, mentioned in your Weekly Consulting Brief when AI shaped that week’s work.
Verify everything
Every source, number and quote gets found and read by a person on your team before it goes in.
Your deliverable: a reference list in your HCD Technical Report where a teammate has opened and read every source.
Write the statement
Use the AI Use Statement Builder above. Each person who used AI signs.
Your deliverable: the AI statement at the end of your Weekly Consulting Brief, Summer Prototype Update, Final Prototype Presentation, HCD Technical Report or EUReCA poster and abstract, wherever the outline or template places it.
Reflect
Look back at your log. Where did AI help, where did it mislead you and what would you do differently?
Your deliverable: the AI Utilization Statement and reflection at the end of your HCD Technical Report.
Example statementWhat a complete one looks like, for a fictional client
Example only. Harbor Street Food Pantry is a made-up client.
“I Jordan Lee (example name) used ChatGPT to generate initial ideas in the following ways: brainstorm interview questions about volunteer retention and summarize public background on nonprofit volunteering, then refined these ideas through independent research and analysis by testing the questions in two practice interviews, rewriting them in our own words and finding and reading every source the summary pointed to. Prompts used: 1. ‘Draft 12 open questions about why food pantry volunteers stop coming back. Flag any that are leading.’ 2. ‘Summarize public trends in volunteer retention at nonprofits. List the kinds of sources I should check.’”
Notice what it does: it names the tool, lists every use, quotes the prompts and says exactly how the team checked the output.
Before you submitA final check for any deliverable
- [ ] Every AI tool we used is named.
- [ ] Every meaningful use is listed, including the ones that did not work out.
- [ ] Our full prompts are included.
- [ ] Every source, number and quote was found and read by a teammate.
- [ ] No confidential client, company or personal data went into a public AI tool.
- [ ] Our insights and recommendations come from our own research and reasoning.
- [ ] Each person who used AI has signed the statement.
- [ ] For the HCD Technical Report: the reflection on how we used AI as a research tool is written.
Go deeper
Back to all BUAD 497 topics · Next: Feedback, Imposter Syndrome and Blind Spots
In-class team activity · 25 min
Green Flag, Red Flag AI
By the end: An honest audit of how AI touched your client project, a three-line team AI rule and a draft AI statement for this week's Weekly Consulting Brief. Your instructor will give you the password in class.
Part of the Smith GLS human-centered leadership sequence: IB 407, BUAD 317 and BUAD 497, toward the Human Centered Consulting Certificate.
Frequently asked questions
What are the rules for using AI in BUAD 497 consulting research?
The BUAD 497 syllabus sets four rules: integrity, attribution, critical thinking and privacy. AI never stands in for your work or invents sources. You name the tool and share your prompts. You verify everything, because you are 100% responsible for accuracy. No sensitive, personal or proprietary information goes into AI tools.
How do you verify information from AI tools?
Use a five-step check: find the original source, read it, match the details, check the date and source, and log it. Authors, year, numbers and quotes must match exactly. AI tools can produce citations, quotes and statistics that look real and do not exist. If you cannot find the original, the claim does not go in your work.
What client information should never go into a public AI tool?
Client internal documents, financials, strategy, contracts, anything under a nondisclosure agreement, interview recordings or transcripts with names, and personal details never go into a public AI tool. Many public tools can store what you type. Ask first about de-identified notes or site photos, and remember that a job title plus a small organization can still identify someone.
What is the difference between a weak AI prompt and a better one?
A weak prompt hands AI your thinking, like "Write our insights." A better prompt keeps you as the analyst: share de-identified quotes and ask what patterns you might be missing, with the quote behind each one. Better prompts ask for search maps instead of statistics, skeptical objections instead of praise, and steps you run on real data yourself.
What should an AI use statement include?
A complete AI use statement names the tool, lists every way it was used, quotes the prompts and says exactly how the team checked the output. Keeping an AI log as you go, with date, person, tool, prompt, what was used and how it was checked, makes the statement quick to write. The HCD Technical Report also includes an AI Utilization Statement.