Week-13-of-52: Your second prompt is almost always better.
AI Tuesdays · Week 13 · Iteration: Your Second Prompt Is Almost Always Better
Multifamily NEXT · AI Tuesdays
Week 13 · Iteration
Self-guided · ~25 min · Phase 2
Warmup
The Workshop
Power Preview
Resources & Tools
Take Action
Week 13 of 52 · Self-Guided · ~20 min
0 starred0 notes
Warmup · Week 12 Recap
Welcome back. Let us warm up first.
Last week (Week 12) was about specificity: the single biggest difference between a weak prompt and a strong one. Vague in, vague out. This week builds directly on top of it. Here is the part nobody tells you: even a specific, well-built prompt is rarely perfect on the first try. And the fix is not to write a better first prompt. The fix is to write a second one. That is what Week 13 is about.
New here?
Coming to AI Tuesdays for the first time? Start with Week 12 first. This workshop assumes you already know that vague prompts get vague answers. Week 12 teaches specificity in about 25 minutes. Come back here after.
Question 1 of 3
Week 12 recap: the core idea
Fill in the blank. "A vague prompt gives you a ____ answer."
Generic.
The more you leave AI to guess, the more it falls back on the average of everything it was trained on. Specific in, specific out. That was Week 12. Hold onto it, because this week is what to do when even your specific prompt does not land the first time.
Question 2 of 3
Week 12 recap: what makes a prompt specific
Which of these actually makes a prompt more specific?
Naming the audience, the format, and the constraints.
Length is not specificity. A short prompt that names exactly who the answer is for and what shape you want beats a long, vague one every time.
Question 3 of 3
Bridge into Week 13
You wrote a specific prompt and the answer is still not quite right. What does this week teach you to do?
Refine, based on what the first answer showed you.
That is the whole idea of Week 13. Your first prompt is a draft. Your second one, written after you have seen what AI gave you, is almost always better. Let us show you how to do that on purpose.
Multifamily NEXT · AI Tuesdays · Week 13
Your second prompt is almost always better.
Iteration: turning a rough first answer into the one you actually needed.
Here is the thing almost nobody tells you when you start using AI. The people who get great results are not the ones who write a perfect prompt on the first try. Nobody does that. They are the ones who look at the first answer, see what is missing, and ask again. The first prompt is a draft. The second prompt, written after you have seen what AI gave you, is where the good work happens. This week is about doing that second prompt on purpose, in a couple of deliberate moves, instead of ten random ones.
When you close this workshop, a mediocre first answer will stop feeling like a dead end and start feeling like information about what to ask for next.
Last Tuesday · Week 12
Specificity: The Skill That Separates Beginners from Builders
Vague prompts get vague answers. This week: what to do when even a specific prompt does not land on the first try.
Localize · Which of these is you?
Which role are you reading this in? Your answer sets up the examples you will see next.
Beginner-friendly · Read at your own pace · Star slides, add notes, print, or email yourself anytime
The Workshop · Agenda
Six parts. About 25 minutes. No timer.
Pause whenever you need. Star anything you want to come back to. Add your own notes as you go. This is designed to teach you the refinement loop, not race you through it.
Part 1 · ~3 min
Why the first prompt is a draft
Nobody writes a perfect first prompt, and the reason is mechanical, not a skill you are missing.
Part 2 · ~4 min
How to read the first answer
The gap between what you got and what you wanted is a map. It tells you exactly what to fix next.
Part 3 · ~4 min
The three moves
Add what you left out. Correct what it got wrong. Show it an example of right. That is refinement.
Part 4 · ~3 min
Refine, do not restart
Why staying in the same chat beats opening a new one, and the one time you should start over.
Part 5 · ~6 min
Power Preview: the loop, run once
One task. A first prompt. A refined prompt. Read what AI produced from each. The jump is the point.
Part 6 · ~5 min
Try This Today · Reference · Session Complete
One refinement drill for this afternoon. Updated cheat sheet. Board paragraph. Where to practice safely.
No timer. Pause whenever you need. The estimate above counts reading plus every tap and reflection.
The Workshop · Why You Should Care
The cost of accepting the first answer.
Two things happen when people do not know how to iterate. Both are quiet. Both are expensive. You have probably done one of them this week.
Scenario one
The give-up
A leasing professional asks AI to write a renewal email. The answer comes back generic and stiff. She thinks "AI is not good at this," closes the tab, and writes the whole thing herself from scratch.
She was one message away from a great draft. She just needed to tell AI the three things it did not know: the resident's name, the tone she wanted, and the renewal offer. She did not give up because AI failed. She gave up because she thought the first answer was the only answer.
Scenario two
The ten-round spiral
A marketing coordinator asks for a social caption. It is off. He rewrites the whole prompt. Still off. Rewrites again. Twenty minutes and nine prompts later he has something usable, but he could not tell you which change fixed it.
He was iterating, but blindly. He never stopped to read why each answer missed, so every new prompt was a fresh guess. Iteration without reading the answer is not refining. It is gambling faster.
Both are refinement problems. One person quit too early. One iterated without aiming. This week fixes both: read the answer, name the gap, change one thing on purpose.
Part 1 · Teaching A
Your first prompt is a draft. That is not a flaw.
When you write a prompt, you are guessing what AI needs to know. You cannot help it. And you do not find out what you left out until you see what comes back. So the first answer is not the finish line. It is the thing that shows you the finish line.
Why nobody writes a perfect first prompt.
Reason 1
The curse of knowing
You already know the resident's name, your property's tone, the offer, the deadline. It is so obvious to you that you forget AI cannot see any of it. Your first prompt leaves it out because your brain already filled it in.
Reason 2
You cannot see the target yet
Until AI shows you an answer, you do not actually know what "good" looks like for this exact task. The first answer makes the target visible. Now you can aim at it.
Reason 3
The first answer is cheap
AI produces a full draft in seconds. It costs you almost nothing to get a rough version, look at it, and ask again. That is a feature. Use it instead of fighting it.
The reframe
Stop trying to write the perfect first prompt. It slows you down, and it does not work anyway. Write a good-enough first prompt fast, read what comes back, and put your real effort into the second one. The first prompt's job is not to be right. Its job is to show you what is missing.
This is the opposite of how you were taught to write
You do not have to get it right before you hit send. AI is not a printer where a typo wastes the page. It is a conversation. The first message is allowed to be rough, because the whole point is that you get to respond to what it says.
Gut check · 15 seconds
You ask AI to "write a move-out letter" and the result is generic. What is the most useful thing to do next?
Tell it what was missing, in the same chat.
The generic answer is not a dead end. It just showed you exactly which specifics to add. That is the second prompt doing the real work.
Part 2 · Teaching B
The first answer is a map. Read it before you rewrite.
Most people react to a weak answer with a feeling ("that is not right") and immediately retype. Skip the retype. First, name what is wrong. The gap between what you got and what you wanted tells you exactly what to change.
Three questions to ask the first answer.
Question 1
What is missing?
Did it leave out a fact only you know? The resident's name, the real policy, the actual number, the deadline. Missing facts mean you need to add context.
Question 2
What did it get wrong?
Did it assume something false? Wrong tone, wrong audience, wrong length, or it invented a detail. Wrong assumptions mean you need to correct it with a constraint.
Question 3
What is the shape?
Is it a wall of text when you wanted three bullets? An essay when you wanted a short email? Wrong shape means you need to specify the format.
This is diagnosis, not judgment
"It is bad" is a verdict. "It is missing the pet policy and it is too formal" is a diagnosis. A verdict makes you start over. A diagnosis tells you the two things to fix. Every time you catch yourself thinking "that is not right," push one step further: not right how?
The trap
The fastest-looking move is to retype the whole prompt the second you dislike the answer. It feels productive. It is usually the slow path, because you are throwing away a perfectly good answer and the information it just handed you about what to fix.
Reflect · Just for you
Think of the last time AI gave you an answer you did not like. Name one specific thing that was wrong with it: a missing fact, a wrong assumption, or the wrong shape.
✓ Saved
Part 3 · Teaching C
Three moves. That is the whole refinement toolkit.
Once you have read the first answer, every fix is one of three moves. Learn these three and you have the skill for the rest of your working life with AI.
Move 1 · ADD
Give it the context it did not have
Hand over the facts you left out. "The resident is Maria. She has lived here two years. The offer is a 3% increase." You are not rewriting the request. You are supplying what was missing.
Move 2 · CORRECT
Tell it what to stop doing
Add a constraint. "Less formal. Do not mention the pool. Do not invent any numbers. Keep it under 120 words." Each correction removes one wrong thing the first answer did.
Move 3 · SHOW
Give it an example of right
Paste a sample. "Here is a renewal email we sent last year that worked. Match this tone." One good example teaches AI more than a paragraph of description.
The key move: change one thing at a time
You almost never need all three at once. Read the first answer, pick the one move that fixes the biggest problem, and send it. Then look again. Refinement is a series of single, deliberate changes, not one giant rewrite. Changing one thing at a time also tells you which change actually helped.
These are follow-ups, not new prompts
Notice all three are follow-up messages in the same chat. "Make it less formal and add the pet policy" is a complete, useful refinement. You do not repeat the original request. AI still has it.
Gut check · 15 seconds
AI wrote a delinquency notice that is too harsh and left out the payment-plan option. What is the best single next message?
The specific follow-up.
It names the correction (soften) and the addition (payment plan) in one clear message, in the same chat, so AI keeps everything else it already had right.
Part 4 · Teaching D
Stay in the chat. AI already has the context.
The single most common iteration mistake: getting a not-quite-right answer and opening a brand-new chat to try again. The moment you do that, you throw away everything AI already knew.
Why the same chat wins.
Same chat
AI still has everything
Every message in a chat is re-read by AI each turn. When you refine in place, AI still has your original request, the facts you added, and its own last answer. You build on all of it instead of rebuilding it.
New chat
A blank container
A new chat starts empty. AI has none of what you established. You re-explain everything just to get back to where you already were, and you usually lose a detail in the re-typing.
When you SHOULD start over
Refining in place is the default. Start a new chat only when one of two things is true: (1) the conversation has gone sideways and AI keeps repeating a mistake even after you correct it, so the wrong turns are now polluting the history, or (2) you are switching to a genuinely different task. If you are still on the same task and AI is getting closer each turn, stay put.
Read the signal
"Getting closer each turn" means your refining is working, keep going. "Going in circles" means either change your approach or start fresh, this time loading everything you have learned into the first prompt.
Reflect · Just for you
Name one kind of task where you have caught yourself opening a new chat instead of just refining the one you had open.
✓ Saved
Part 5 · Teaching E
Knowing when to stop refining.
Iteration has an opposite failure to giving up early: never stopping. Chasing a perfect answer through fifteen rounds when round three was already good enough to take and finish yourself.
The "good enough to finish by hand" line.
Truth 1
AI gets you about 80 percent
For most real tasks, AI is a drafting tool, not a finishing tool. When the draft is close enough that you can finish it faster by hand than by prompting again, stop prompting. Take it and edit.
Truth 2
Diminishing returns are real
The first refinement usually makes a big jump. The fifth makes a tiny one. If two rounds in a row barely changed anything, more rounds will not save it. Change your approach or finish it yourself.
Truth 3
You are the editor, always
The last stretch is judgment: your voice, your resident, your call. That was never AI's job. Handing AI the wheel for the finish is how you get answers that are technically fine and completely generic.
The loop has a brake
Ask. Read. Refine. Read. And the moment the draft is good enough to finish by hand, stop and finish it. Not the first answer, that is quitting early. Not the fifteenth, that is spinning. The one you can take from here.
Gut check · 15 seconds
You have refined a listing description three times. It is now solid, but the last edit only changed two words. What is the move?
Take it and finish by hand.
Two-word changes mean you have hit diminishing returns. The last polish is faster and better coming from you, the person who actually knows the property.
Part 5 · The Bright Line
One rule that never changes, no matter how many rounds you run.
The bright line
Refining a prompt improves the words. It never changes what data is allowed in. If a fact should not have gone into the first prompt, it should not go into the second one either.
Iteration is about quality. Data classification is about permission. They are two different questions, and getting better answers never loosens the rules about what you are allowed to paste. Round one or round five, the line is the same: public and non-sensitive is fine. Anything with a real resident's private information stays out of a general AI tool, no matter how much better the tenth draft would be with it in.
The move when the data is off-limits
You do not stop using AI. You refine with brackets. Ask for the draft with [resident name], [balance], or [date] left as placeholders, then fill the sensitive parts in yourself, by hand. That is iteration staying inside the rules. You practice this exact call on Slide 19.
This is the one hard rule of the week. Everything else is technique.
The Workshop · Default Rule
The default rule for this week.
Never accept the first answer as final. Read it, name the gap, and refine once, on purpose. Your second prompt is almost always better, so make the second prompt deliberately.
If you do nothing else with this workshop, do this. The next time AI hands you a mediocre answer, do not close the tab, and do not retype blindly. Say out loud what was missing or wrong, turn that into one clear follow-up, and send it in the same chat. That single habit is the difference between people who think AI is disappointing and people who get real work out of it.
One rule stays fixed
Refining makes the words better. It never changes what data is allowed into the prompt. See Slide 9 for the bright line, and Slide 19 to practice the call.
The Workshop · Q & A
Questions people ask about refining prompts.
Tap a question to open the answer.
Trying to get it perfect first is the slow path, because you cannot see what is missing until you see an answer. Two deliberate refinements beat ten minutes of staring at a blank prompt trying to anticipate everything. Write a fast rough draft, then aim.
Specificity is what you put into the prompt. Iteration is what you do after you read the answer. They work together: your first prompt is as specific as you can make it, and refinement adds the specifics you did not know you needed until the answer showed you. Week 20 goes deeper into structured multi-turn prompting.
Always explain what to change. "Try again" makes AI guess differently, which might be worse. "Make it shorter and remove the second paragraph" tells it exactly what to fix. Name the change, do not just signal that you are unhappy.
That is one of the two times to start fresh. When corrections stop working, the chat history is working against you. Open a new chat and put everything you have learned into the first prompt this time. You are not starting from zero, you are starting from what you figured out.
The same rule applies on every round: getting a better answer never changes what data is allowed in. Refine freely with non-sensitive information. A real resident's private details stay out of a general AI tool no matter which draft you are on. See Slide 9, and practice the call on Slide 19.
Power Preview · The Refinement Loop
Watch the loop run once.
The next two slides show the same task twice: a first prompt, then the refined prompt that came after reading the first answer. Same person, same tool, same task. The difference is not talent. It is one deliberate round of refinement.
The loop · Four steps
1. Ask. Write a good-enough first prompt and send it. 2. Read. Name what is missing, wrong, or badly shaped. 3. Refine. Send one follow-up in the same chat: add, correct, or show. 4. Stop. When it is good enough to finish by hand, take it and finish.
The task we will run.
A resident, Maria Alvarez, has a lease ending in about 60 days. You need a renewal email. Watch the first prompt, then the refined one.
Note before the runs
The answers below are composite examples of what these prompts typically produce. Real AI output varies every run. The point is the jump in quality from one round of refinement, not any exact wording.
Power Preview · Run 1
Run 1: the first prompt.
The first prompt (a reasonable place to start)
Write a renewal email to a resident whose lease is ending soon.
What AI produces (typical)
Dear [Resident Name],
We are reaching out because your lease is coming up for renewal. We value you as a resident and would love for you to continue living with us. Please contact the leasing office to review your renewal options and let us know your plans.
We look forward to hearing from you.
Sincerely, [Property Management Team]
Read it. What is missing?
Missing facts (needs ADD). No resident name. No property name. No renewal offer. No deadline. AI could not include what it was never told.
Wrong assumptions (needs CORRECT). Corporate-generic tone. No sense of your property's voice. It did not invent anything false here, but it defaulted to bland.
Shape is close but soft (format). It is an email, which is right, but there is no clear single next step the resident can act on.
The read
This is not a failure. It is a first answer doing its job: showing you the four things you forgot to say. Now you refine. Slide 14 shows the follow-up.
Power Preview · Run 2
Run 2: the refined prompt.
Same chat. One follow-up message. It does not repeat the original request. It adds the facts, corrects the tone, and names the shape.
The refined prompt (the follow-up you send)
That is close but too generic. Rewrite it for a resident named Maria Alvarez who has lived here two years. Warm and personal, not corporate. Mention her lease ends on [date], that we would genuinely like her to stay, and that if she renews within 14 days we will hold her current renewal offer. End with one clear next step: reply to this email or call the office. Keep it under 150 words. Do not invent any dates, rent amounts, or offers. Leave anything you do not know in [brackets] for me to fill in.
What AI produces (typical)
Hi Maria,
It is hard to believe you have been with us for two years already. As your lease approaches its end on [date], I wanted to reach out personally: we would genuinely love to have you stay.
If you renew within the next 14 days, we will hold your current renewal offer for you. No pressure at all, I just wanted to make sure you had the chance.
The easiest next step is to reply to this email or give the office a call, and we will take it from there.
Warmly, [Your name], [Property name]
What changed, and why it worked.
ADD did the heavy lifting. Maria's name, the two-year tenure, the 14-day offer, the deadline. The email went from "any property" to "this resident."
CORRECT fixed the tone. "Warm and personal, not corporate" removed the bland default.
The guardrail did quiet work. "Do not invent any dates or amounts. Use brackets." That is the hallucination lesson from Week 12, applied. The refined answer is safe to send because it could not make up a number.
Format made it usable. "One clear next step, under 150 words" gave you something a resident will actually read and act on.
The point
None of that came from being smarter than the first prompt. It came from reading the first answer and naming the four things it was missing. That is the entire skill. The second prompt was better because the first answer told you how to write it.
Power Preview · Try This Today
One refinement drill you can run this afternoon.
Fifteen minutes. Do it once and the loop becomes a habit.
The drill · One task, one refinement
Step 1. Pick one real writing task you would normally do by hand: a renewal note, a maintenance update, a listing blurb, a review reply.
Step 2. Write a quick, good-enough first prompt. Do not overthink it. Send it.
Step 3. Do not retype. Read the answer and write down the one biggest problem: a missing fact, a wrong assumption, or the wrong shape.
Step 4. Send one follow-up in the same chat that fixes just that: add, correct, or show. Read the new answer.
Step 5. Repeat Step 4 at most once more. The moment it is good enough to finish by hand, stop and finish it.
Step 6. Notice how many rounds it took. For most tasks it is two. That is the loop.
A warning about what sounds good.
When your refined prompt makes AI produce something fluent and confident, remember Week 12: fluent is not the same as true. A polished renewal email that invented a rent number is worse than a rough one that left it blank. That is exactly why the drill says "do not invent numbers, use brackets." Read what you get. Verify anything factual. Especially when the writing got good.
Where to practice safely
The refinement drill is safer to run in a governed workspace.
If your organization is not sure where your team should practice refining prompts on real resident material, this is the environment built for it. Data classification, human review, and the RETTC principles are built in, so "which tier of data is in this prompt" is handled before anyone hits send. Learn more from Multifamily NEXT at tami@mprodigitaledge.com. mPro Digital Edge: the AI workspace built for using AI with governance, not against it.
Practice this on us
You just spent 25 minutes learning to refine and to push on confident answers. Push on this one too. Ask us for a demo. Ask what "governance built in" actually looks like. Ask for third-party reviews. Any AI workspace that survives those questions is worth considering. Any that does not is worth walking away from. This is what Multifamily NEXT wants you doing with every AI product, including this one.
Share this workshop with a coworker
If Week 13 helped, someone on your team probably needs it too. Send them the link.
Reference · Bookmark this slide
Vocabulary cheat sheet: 30 core terms.
Every AI term this curriculum has introduced so far, in one place. Plain-language definition plus a multifamily read for each. New this week: Iteration. Print it. Bookmark it. Come back to it whenever you need to remind yourself what a term actually means.
Foundational (from Weeks 8-10)
Generative AI
AI that creates new content (text, images, audio, code) based on training data patterns.
Large Language Model (LLM)
The kind of AI behind ChatGPT, Claude, Gemini, and Copilot. Trained on vast text and predicts one word (token) at a time.
Token
A small chunk of text an LLM reads and writes. Roughly 0.75 words per token; 1,000 tokens is about 750 words.
Training Data
The vast collection of text an LLM learned patterns from. Not a searchable database, a compressed pattern archive.
Prediction
How AI decides what to say next: the most likely word based on the pattern of what came before, one word at a time.
Hallucination
AI producing a confident, fluent answer that is factually made up. Especially common when it does not know something you did not tell it.
Sycophancy
AI agreeing with what you just said instead of pushing back. Feels helpful, is often wrong.
Prompt anatomy (from Week 11)
Prompt
What you type to AI. Your input. The one part of the container you fully control.
System Prompt
A hidden instruction the AI tool sends the model before your prompt. Shapes tone, role, and constraints. You do not write it and cannot see it.
Conversation History
Every earlier message in the current chat, yours and AI's. AI re-reads all of it every turn. This is how AI "remembers."
Context Window
The ceiling on how much text AI can read in one conversation. When it fills up, the oldest content silently drops off.
Prompt Anatomy
The five parts of a well-formed prompt: instruction, context, input data, output format, examples.
Output Format
The shape you want the answer in: length, structure, tone. One line in your prompt saves you rewriting the whole reply.
Prompt quality (from Week 12)
Specificity
The single biggest lever on prompt quality. Naming the audience, the format, and the constraints so AI is not guessing.
Ambiguity
What makes AI fall back on average defaults. The parts of a prompt AI has to guess. Iteration cannot rescue what specificity should have prevented.
This week · The refinement loop (Week 13)
Iteration
Refining a prompt after reading the answer, in the same chat. Your first prompt is a draft. Your second, written on purpose, is almost always better.
Refinement
The act of sending a targeted follow-up: add what was missing, correct what was wrong, or show an example of right. One deliberate change beats a full rewrite.
The Three Moves
ADD (give it the context it did not have), CORRECT (tell it what to stop doing), SHOW (give it an example of right). Every refinement is one of these three.
Diminishing Returns
When each new round of refinement barely changes the answer. Your signal to stop prompting and finish the piece by hand.
Loop
The Ask → Read → Refine → Stop cycle. The four-step rhythm that turns first-answer disappointment into a usable draft.
Governance and data (running through the curriculum)
Data Classification
Sorting a piece of information by which AI tool it is allowed to enter. Tier 1 (safe for any AI), Tier 2 (enterprise only), Tier 3 (never in AI). Full treatment: Week 1.
Tier 1 · Safe
Public or fully de-identified information. Any approved AI tool is fine.
Tier 2 · Enterprise Only
Sensitive but useful information. Only in an approved enterprise AI workspace with data protections in place.
Tier 3 · Never
Information that never enters an AI tool. Real names tied to sensitive details, financial identifiers, protected-class data, and screening or eviction information.
Use Case
A specific way you plan to use AI: who is doing it, what task, what data goes in, what comes out. The Council approves use cases, not tools.
Guardrail
A rule inside your prompt or your workspace that limits what AI can do. "Do not invent numbers" is a guardrail. Full treatment: Week 17.
Human-in-the-Loop
A person reviews AI output before it goes to a resident or a decision. The RETTC principle behind almost every safe use case.
AI Council
The group in your organization that approves use cases, reviews new tools, and owns the response when something goes wrong. Full treatment: Week 7.
NIST AI RMF
National Institute of Standards and Technology AI Risk Management Framework. The four functions: GOVERN, MAP, MEASURE, MANAGE.
RETTC AI Governance Framework
Real Estate Technology and Transformation Center. Eight principles specific to rental housing. Detailed on Slide 18.
Same terms, same definitions, printable one-sheet. Matches this slide and Slide 21.
Reference · Bookmark this slide
The refinement loop, on one page.
Read down the left column when you are refining a prompt. Read down the right when a refinement is not working.
Do this on every round
ASK. Write a good-enough first prompt. Send it fast.
READ. Name what is missing, wrong, or the wrong shape.
REFINE. Pick one move: ADD, CORRECT, or SHOW.
STAY. Send the follow-up in the same chat. Do not restart.
STOP. When it is good enough to finish by hand, take it.
Watch for these traps
Retyping blindly. Sending a fresh guess without reading the answer.
Changing everything at once. You cannot tell which move worked.
Verdicts instead of diagnoses. "It is bad" teaches AI nothing.
Opening a new chat. Throwing away everything AI already had.
Chasing perfect. Round eight when round three was good enough.
Practice this instinct on us, too.
Week 12 taught you to test AI's confident claims. That instinct applies to this workshop just as much. We have been confident throughout. Push on it.
What Week 13 might be wrong about
1. "The Three Moves" is a teaching frame, not a fixed rule. Add, Correct, Show is one way to name what refinement does. Different guides split it differently. What matters is the behavior, not the count.
2. "Same chat almost always wins" is true today, but tools are changing. Some AI products are adding memory features that carry limited context across chats. Verify how your own tool handles this before you build a workflow around it.
3. "About 80 percent" is a rule of thumb, not a benchmark. How close AI gets on the first try varies by task, tool, and how specific your first prompt was. Use it as a mindset, not a number.
4. Iteration is not a substitute for specificity. If your first prompt was too vague, you may need to start over instead of refining. Week 12 first, then this.
5. We are confident everywhere. Same as Week 12. That is a feature of workshop writing, not a claim of certainty. If any specific claim struck you as too confident, that instinct is right. Verify it.
The habit we want you to keep
Verify any confident-sounding claim, including this workshop. If you spot something in Week 13 that does not match what your tool actually does, that is Week 12 working. Good.
Permanent · Standards Library
NIST AI Risk Management FrameworkNIST AI RMF (NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY AI RISK MANAGEMENT FRAMEWORK)A voluntary US government framework for managing risks associated with AI systems. Published January 2023 by the National Institute of Standards and Technology, a bureau of the US Department of Commerce. Not a law. Not regulation. A structured way to think about, measure, and manage AI risk that has become the de facto US standard for AI governance. The four functions: GOVERN (culture, policies, and accountability structures for AI risk). MAP (understanding context and categorizing what your AI system is). MEASURE (assessing and tracking AI risks in a specific implementation). MANAGE (prioritizing and responding to AI risks based on impact). Multifamily example: When your AI Council writes a policy that says "no autonomous decisions affect residents," that is GOVERN. When you categorize a leasing chatbot as "LLM with system prompt, Tier 2 data allowed," that is MAP. When you audit AI-drafted renewal letters quarterly for Fair Housing risk, that is MEASURE. When you decide which use cases get approved with guardrails vs. refused, that is MANAGE. Four functions. One rhythm. Link: https://www.nist.gov/itl/ai-risk-management-framework and RETTC AI Governance Framework.
Two frameworks. The NIST AI Risk Management Framework (AI RMF) from the National Institute of Standards and Technology, and the AI Governance Framework from RETTC (Real Estate Technology and Transformation Center, at rettc.org). US-anchored. Together they cover the risk side (NIST) and the operating side (RETTC) of governing AI in rental housing. Every Week 13 concept about refining prompts lands somewhere in one or both.
The RETTC AI Governance Framework · all 8 principles
Official titles as published at rettc.org/frameworks/ai-governance. These are the 8 principles the Framework asks every rental housing provider and technology partner to adopt.
#
Principle
1
Establish an Organizational Philosophy on AI Use and Oversight. Adopt a clear philosophy that defines how AI will be used, governed, and monitored across operations.
2
Promote Fairness & Consumer Protection. Ensure AI systems comply with the Fair Housing Act, Fair Credit Reporting Act, Americans with Disabilities Act, and state and local law.
3
Uphold Transparency & Explainability. Promote a general understanding of when AI influences key decisions, and provide clear explanations to renters and applicants of decisions affecting them as required by law.
4
Protect Privacy & Promote Data Integrity. Adhere to federal, state, and local consumer privacy laws, with an emphasis on data minimization, security, and informed consent.
5
Maintain Outcome-Based Accountability & Human Involvement. Engage humans in significant decisions, establish processes to review model outcomes, provide escalation channels for complaints and appeals, and educate staff on AI use and limitations.
6
Promote Positive Renter Experiences and Outcomes. Elevate technologies that enhance affordability, access, and housing stability, and use AI to improve resident experiences.
7
Strive for Innovation with Responsibility. Encourage development of AI tools that benefit residents, prospects, guests, and operators, without compromising trust.
8
Encourage Third-Party Partner Accountability and Due Diligence. Incorporate robust due diligence and risk management controls for third-party AI tools. Invite transparency from vendors regarding data sources, model limitations, and known risks.
This week bridged to the frameworks
Week 13 concept
NIST AI RMF anchor
RETTC AI Governance anchor
The refinement loop (Ask, Read, Refine, Stop)
Manage function. A named, teachable procedure for how staff interact with AI. Iteration written down is a governance artifact, not just a personal habit.
Principle 1 (Establish an Organizational Philosophy on AI Use and Oversight). How your team refines prompts is part of your written AI policy.
Reading the first answer before refining
Measure function. Reading AI output for what is missing or wrong is the smallest, most frequent quality check your team performs.
Principle 5 (Maintain Outcome-Based Accountability & Human Involvement). The human read is the accountability step. Every refinement round is a human-in-the-loop moment.
The three moves (Add, Correct, Show)
Map function. Naming the three kinds of prompt fix categorizes what kinds of AI output failures your team is actually seeing.
Principle 3 (Uphold Transparency & Explainability). A refinement log makes it possible to explain how a resident-facing message was shaped.
The Week 13 bright line (data allowed in does not change with the round)
Govern function. Data classification is a policy floor. Iteration is a quality technique. The floor never moves. This belongs in the AI Council's written standard.
Principle 4 (Protect Privacy & Promote Data Integrity). Refining more does not authorize more sensitive data. Data minimization applies to the tenth prompt just as much as the first.
Refine, do not restart (staying in the same chat)
Manage function. A short-loop workflow that avoids re-establishing context, so review effort concentrates where it matters: the final human read.
Principle 5 (Maintain Outcome-Based Accountability & Human Involvement). A single continuous thread makes it easier to review how a response evolved, and who was accountable at each step.
Knowing when to stop (finish by hand)
Govern function. The point where AI hands off to a person is a governance boundary. Staff should know it and their reviewer should confirm it.
Principle 6 (Promote Positive Renter Experiences and Outcomes). The final message a resident sees was written by a person. AI drafted, a human finished.
Refining is a technique. Data classification is a rule.
Eight scenarios. In each one, you are refining a prompt. Your job is to sort where that prompt is allowed to go: Green (any approved AI tool), Yellow (enterprise-only workspace), or Red (never in AI). Getting a better answer never changes which color a scenario is. That is the whole test.
1
You are refining a listing description for a two-bedroom that is publicly available on your website. Third round, still not right. You paste the address, square footage, and amenities into a public AI tool for another rewrite.
Correct. This is already public. No resident data. Any approved AI tool is fine. Refine as many rounds as you like.
Not quite. The right answer is Green. This is already public. No resident data. Any approved AI tool is fine. Refine as many rounds as you like.
2
Round two of refining a delinquency notice. You are about to paste the resident's real name and the exact balance into a public AI tool to make the tone "warmer."
Correct. A real resident name plus a balance is sensitive financial information tied to an identified person. It never goes into a general AI tool, no matter which round of refinement you are on. Ask for a template with [Resident name] and [balance], fill it in yourself.
Not quite. The right answer is Red. A real resident name plus a balance is sensitive financial information tied to an identified person. It never goes into a general AI tool, no matter which round of refinement you are on. Ask for a template with [Resident name] and [balance], fill it in yourself.
3
You are refining a Fair Housing training reminder for staff. No resident details, just a policy summary you want made more readable. Fifth try.
Correct. General policy content, no resident data. Refine freely.
Not quite. The right answer is Green. General policy content, no resident data. Refine freely.
4
Refining a maintenance email. You have already stripped out the resident's name and apartment number and replaced them with [name] and [apt]. You now want a shorter, warmer version.
Correct. You removed the identifying information yourself. What is left is generic. Iterate as needed and fill the brackets in by hand at the end.
Not quite. The right answer is Green. You removed the identifying information yourself. What is left is generic. Iterate as needed and fill the brackets in by hand at the end.
5
Round three of refining a review response. You are about to paste in the reviewer's real name from the site and details you have about their maintenance history.
Correct. A specific person plus their maintenance history is identifiable resident information. Round of refinement does not change that. Refine with placeholders, then personalize by hand before you post.
Not quite. The right answer is Red. A specific person plus their maintenance history is identifiable resident information. Round of refinement does not change that. Refine with placeholders, then personalize by hand before you post.
6
You are refining a resident satisfaction survey summary in your organization's enterprise AI workspace (Tier 2 approved). The summary contains individual apartment numbers and comment text.
Correct. Individual-level survey content is sensitive but usable in an approved enterprise workspace with data protections. Refine there, not in a public AI tool.
Not quite. The right answer is Yellow. Individual-level survey content is sensitive but usable in an approved enterprise workspace with data protections. Refine there, not in a public AI tool.
7
Refining a market analysis for your regional. You paste public rent comps from three competitor properties (from a public source) and want a cleaner summary.
Correct. Public market data. No resident anything. Any approved AI tool. Refine away.
Not quite. The right answer is Green. Public market data. No resident anything. Any approved AI tool. Refine away.
8
Round four of refining a screening-decision explanation. You want AI to write it more clearly. You are about to paste the applicant's name, credit score, and prior eviction record into a general AI tool.
Correct. This is the never-zone. Applicant name plus screening data plus eviction history is exactly the combination that never enters a general AI tool. Screening-decision content requires an approved, governed workflow. Even in the right workspace, a human writes the final wording.
Not quite. The right answer is Red. This is the never-zone. Applicant name plus screening data plus eviction history is exactly the combination that never enters a general AI tool. Screening-decision content requires an approved, governed workflow. Even in the right workspace, a human writes the final wording.
The pattern: Green scenarios are public or de-identified. Yellow scenarios include sensitive but usable content that belongs only in an approved enterprise workspace. Red scenarios name a real resident or applicant together with sensitive information about them: financial, screening, protected-class, or eviction. Refinement round does not change any of that. The bright line from Slide 9 sits under every round.
Take Action · Anonymous Quiz
Five questions. Anonymous. Just for the pattern.
Tap an answer for each question. The correct answer highlights green and a short explanation appears. Nothing is scored publicly, nothing is graded, no name is collected. When you finish, you can submit your selections anonymously so the curriculum can see where the class as a whole gets stuck. That is what tunes future weeks.
1. According to Week 13, what job does your first prompt actually do?
Answer: B. Your first prompt is a draft. Its job is to make the target visible. What comes back tells you exactly what the second prompt should say.
2. AI gave you a renewal email that is generic and left out the resident's name and the offer. What is the best next move?
Answer: C. Refine in place with a specific follow-up. Same chat means AI keeps everything it already had right. "Try again" teaches AI nothing. Starting over throws away context.
3. Week 13 names three refinement moves. Which set is correct?
Answer: A. Add what was missing. Correct what it got wrong. Show it an example of right. Every refinement is one of these three, and you almost never need more than one at a time.
4. You have refined a listing three times. The last two answers barely changed from each other. What should you do?
Answer: B. Diminishing returns. When two rounds in a row barely move the answer, more rounds will not save it. The final polish belongs to you: your voice, your resident, your call.
5. You are on round four of refining an eviction notice, trying to soften the tone. You are about to paste the resident's real name and their full ledger into a public AI tool because the earlier drafts were "not personal enough." What does Week 13 say?
Answer: C. This is the bright line from Slide 9. Better answers never buy you permission to paste sensitive data. Ask for a template with [Resident name] and [balance], then fill the sensitive parts in yourself.
No name. No email. Just which options you tapped, so we can see the pattern across everyone taking Week 13.
Take Action · Policy / Action Download
Take one thing back to your team.
Two artifacts. One quick-reference. One supervisor email. Both ready to send.
Artifact 1 · The Refinement Loop Quick-Reference (PDF)
A one-page card showing the four-step loop (Ask, Read, Refine, Stop), the three refinement moves (Add, Correct, Show), the three questions to ask the first answer, and the bright line about data. Print it. Tape it near your monitor.
If your supervisor has not seen Week 13 yet, send them this. The mailto link opens in your email client with a draft already filled in. Add your supervisor's address and send.
Draft (edit before sending)
Subject: Week 13 of AI Tuesdays · your second prompt is almost always better
Hi [supervisor],
I just finished Week 13 of the Multifamily NEXT AI Tuesdays workshop series. This one is about iteration: why your second prompt to AI is almost always better than your first one, and how to get to the good version on purpose instead of by accident.
It teaches a four-step loop (Ask, Read, Refine, Stop) and three refinement moves (Add, Correct, Show). I think our team would get value from walking through it together.
The workshop is free at multifamilynext.com. Password 1352.
Note: The email is drafted for you but the recipient field is left blank on purpose. Add your own supervisor's address. Multifamily NEXT never pre-fills recipients.
✔ A working mental model of why the first prompt is a draft, and why that is not a flaw.
✔ Three questions to ask any first answer: what is missing, what is wrong, what is the shape.
✔ The three refinement moves: ADD context, CORRECT with a constraint, SHOW an example.
✔ The Ask, Read, Refine, Stop loop, with a brake: stop when the draft is good enough to finish by hand.
✔ The bright line: refining never changes what data is allowed in. Round one or round ten, the tier is the same.
✔ Five new vocabulary terms (Iteration, Refinement, The Three Moves, Diminishing Returns, Loop), bringing your cheat sheet to 30 terms.
The board paragraph.
If your CEO asked you at Monday's meeting what our team learned in Week 13, here is what you say:
Our team learned that the first prompt to AI is a draft, not the finish line, and that the people who get real work out of AI are not writing perfect first prompts. They are reading the first answer and refining once, on purpose. We now have a short loop we use for every AI task: ask a good-enough first prompt, read what comes back, refine with one deliberate move (add missing context, correct a wrong assumption, or show an example), and stop when the draft is good enough to finish by hand. This replaces two failure modes we were seeing: people giving up on AI after one mediocre answer, and people spinning through eight or ten prompts trying to get something usable. One rule stays fixed under every round: getting a better answer never changes what data is allowed in. Sensitive resident details stay out of general AI tools no matter which draft we are on. Next Tuesday, Week 14 gives us the workhorse prompt structure that makes the first prompt land closer to the target: RTF, which stands for Role, Task, Format.
Where to practice this safely.
Governance handoff · mPro Digital Edge
The refinement loop, in a workspace built for it.
mPro Digital Edge is the enterprise AI workspace for multifamily: a governed environment where the NIST AI Risk Management Framework and the RETTC 8 Principles are built into every tool. Your team can run the ask-read-refine-stop loop on real material without every round becoming a data-classification argument. Data tiers are handled up front. Human review is built in. AI Council approvals are tracked. That means "where can our team practice this?" has a real answer, not a shrug.
One platform where every person in the organization gets measurably better at their job with every interaction.
Now that you know how to refine, next Tuesday teaches the workhorse prompt structure that makes the first prompt land closer to the target. Give AI a role, a specific task, and the exact output format you need. RTF is the structure Phase 3 opens with.
See you Tuesday. Trust me, this one levels up your leverage on AI.
Tami Siewruk
CEO, Multifamily NEXT and mPro Digital Edge
Built by Multifamily NEXT. Powered by mPro Digital Edge, the AI workspace for governed AI use in multifamily.
