5 of 40

AI Is Not Replacing Planners

AI can reduce friction

Repetitive work

Blank pages

Tedious research

Administrative tasks

Planning still requires

Judgment

Ethics & public trust

Political awareness

Negotiation

Local knowledge

6 of 40

AI Is Not Replacing Planners

Planning Magazine - July 30, 2026s

AI Changed Our Company. It Didn’t Replace Our Expertise.

New tools may help organize and apply planning information, but local context, public trust, and professional judgment remain essential.

7 of 40

The Planning Intern Analogy

“Think of AI as a very fast planning intern—not your planning director.”

Use it to accelerate work. Keep responsibility with the planner.

8 of 40

Human Intelligence

~86 billion neurons. ~150 trillion synaptic connections. Estimated storage on the order of a petabyte — and that estimate jumped roughly tenfold once researchers realized synapses vary in size far more than assumed.

The kicker: all of it runs on about 20 watts. That's a dim light bulb. Continuously, since before you could talk. A data center running a large model does not.

You may have heard that "we only use 10% of our brains". That myth is busted. Imaging shows the whole brain is active over a full day. There is no dormant 90%.

What is true: neuroplasticity continues at any age. London cab drivers memorizing the city's street network showed measurable growth in the hippocampus.

Land it: "That's the machine you're bringing to this tool. AI isn't a replacement for it — it's an amplifier."

Transition: "So that's the wetware. Let's talk about what's actually happening inside the other thing."

9 of 40

AI IN PLAIN ENGLISH

What Is Generative AI?

Large language models (LLMs)

Pattern recognition and next-token prediction

Produces plausible language from patterns in training and supplied context

It can sound expert without having lived experience or local institutional memory

Time: ~50 sec

The one thing: It predicts plausible language. It does not retrieve verified facts.

What to say

Large language models. The mechanism: given everything so far, predict the next most likely piece of text. Then do it again. Millions of times.
The planning analogy: imagine someone who has read every comprehensive plan, every zoning code, and every staff report ever written — but has never walked a project site, never sat through a public hearing, never met your council.
They've absorbed the patterns of how planning documents are written. That's a real and useful skill. It is not the same as knowing anything about your community.
The sentence to land, slowly: "It's predicting what a good answer looks like. It is not retrieving what's true."
Foreshadow: "Hold onto that. It explains everything we're going to talk about in the risk section."

Delivery

Do not explain transformers, tokens, attention, or training runs. Nobody needs it and you'll burn 90 seconds you don't have.
"Next-token prediction" is on the slide — you can gesture at it without defining it.
This is the intellectual core of your segment. If you're running long, cut Slide 10 or 11, not this one.

Transition: "There's a shorthand for this that I find useful."

10 of 40

A Useful Mental Model

“A very well-read new hire—not a colleague with tenure.”

It may write like an experienced planning director, but it has never walked your site, attended your hearing, or learned your council’s history.

Time: ~30 sec

The one thing: Well-read new hire, not a tenured colleague.

What to say

Explicit callback to Jason: "Jason gave you the intern analogy. Here's the mechanical reason it holds."
It may write like a 20-year planning director, because it has read thousands of documents written by people like that. The prose is senior. The knowledge is day-one.
Make it concrete and local — this is what makes the room laugh and remember:

It doesn't know why last year's rezoning went sideways.
It doesn't know which two council members don't speak to each other.
It doesn't know that the applicant's father sits on the planning commission.

"That's not a bug you can prompt your way out of. It's what the tool is."

Delivery

Pick one of the three examples if you're short. The planning-commissioner's-son line is the strongest.
Compression candidate: this whole slide can be the callback sentence plus one example.

Transition: "Quick landscape check before we get into what it's actually good for."

11 of 40

THE LANDSCAPE

Common AI Tools

ChatGPT

Microsoft Copilot

Google Gemini

Claude

Perplexity

Time: ~25 sec

The one thing: Which tool matters far less than which tool your organization has approved.

What to say

Name them and move: ChatGPT, Microsoft Copilot, Google Gemini, Claude, Perplexity.
The honest framing: for the planning work in this presentation, these are more alike than different. You will not get a dramatically different staff report draft by switching vendors.
Two distinctions that do matter:

Free vs. paid tier is often a bigger capability gap than one vendor vs. another. Someone who tried the free version two years ago and was unimpressed is not evaluating the current tool.
Consumer account vs. enterprise/government account is the difference that matters for data handling — and that's a Section 5 conversation.

"Many of you will find Copilot is what your IT department has already licensed. That's usually the right place to start, because it's the one that's been through your procurement and security review."

Delivery

Stay vendor-neutral. You're the attorney on the panel; endorsing a product is off-brand and you may have client conflicts.
Do not do a feature tour. This is a 25-second slide.
Compression candidate: read the list, say "more alike than different," say "use what your organization approved," move on.

Transition: "So what is this category of tool actually good at?"

12 of 40

INFORMATION WORK

What AI Does Well

Summarize long material

Organize scattered information

Brainstorm alternatives

Draft first-pass content

Compare documents and standards

Rewrite for different audiences

Explain complex concepts

Time: ~45 sec

The one thing: AI is strong at reshaping information that already exists.

What to say

Do not read the seven bullets. The audience can read. Pick two and make them vivid:

Summarize. A 200-page comprehensive plan, or a 40-page legislative bill, into a two-paragraph brief you can actually absorb before a hearing at 6:00 when the packet landed at 4:30.
Compare. Line up setback and height standards from three neighboring jurisdictions side by side. That's a task that used to eat an afternoon.

Then name the pattern rather than the list: "Notice what all seven have in common. Every one of them is taking information that already exists and reshaping it — for a new purpose, a new format, or a new audience."
Optional third if time: Rewrite. Same concept, three audiences — engineers, elected officials, residents. Jason and I will show that later.

Delivery

The temptation to read all seven bullets will be strong. Two concrete examples plus the pattern beats seven abstractions.
"Draft first-pass content" is worth one clause: "so you're editing instead of staring at a blank page."

Transition: "Now the other side of the ledger — and this is the more important slide."

13 of 40

JUDGMENT WORK

What AI Does Poorly

Know local politics

Interpret public sentiment

Understand context you did not provide

Cite sources reliably without verification

Make accountable policy decisions

Time: ~45 sec

The one thing: No lived experience, no accountability, no verified memory.

What to say

Again, don't read all five. The two that matter most to this audience:

"Understand context you did not provide." It doesn't know why that variance was denied in 2019, or that "affordable housing" means something specific in your jurisdiction and something different one city over. You have to hand it that context. Every single time. It doesn't remember.
"Cite sources reliably." This is the one that ends careers, and I'll spend real time on it in the risk section. Preview only: it can generate a statute citation that looks perfect and does not exist. And here's the part that gets people — the confidence is identical whether it's right or fabricated. There's no tell.

Draw the distinction explicitly: "This is different from being wrong. A wrong answer is a mistake. This is confident invention."
Close with the pattern: "No lived experience. No accountability. No verified memory. It's working entirely from patterns in text — not from knowing your community."

Delivery

Slow down on "the confidence is identical." That's the sentence that makes the risk section work later.
Don't editorialize about AI being dangerous here. Stay descriptive. You have a whole section for the warnings and it lands harder if you haven't pre-spent it.

Transition: "Which gives us the distinction that organizes everything else today."

14 of 40

The Core Distinction

“AI is good at handling information. Planners are responsible for judgment.”

Use the machine for speed, structure, and iteration—not accountability.

10-20-102. Definitions.

(40) "Land use authority" means:

(a) a person, board, commission, agency, or body, including the local legislative body, designated by the local legislative body to act upon a land use application; or

(b) if the local legislative body has not designated a person, board, commission, agency, or body, the local legislative body.

Time: ~25 sec · Section closer

The one thing: AI handles information. Planners are responsible for judgment.

What to say

Say the line as written, then stop: "AI is good at handling information. Planners are responsible for judgment."
One beat of amplification: "Use the machine for speed, structure, and iteration. Not for accountability. Accountability doesn't delegate."
Frame the rest of the session: "Everything Jason and I show you for the next half hour is an application of that one line. Every use case is on the information side. Every risk is what happens when the line gets crossed."

Delivery

This is a punctuation slide. Resist elaborating. Say it, let it sit for a second, hand off.

Live example? ("here's a real statutory definition — watch what happens when I ask AI to explain it to a resident").

Transition to Jason: "Jason's going to walk you through where this actually shows up in your week."

15 of 40

REMOVE THE FRICTION

Where AI Fits Into Planning

Research & policy analysis

Staff reports

Code writing & review

Public communication

Economic development

Project management

Public meetings

16 of 40

USE CASE 1

Research & Policy Analysis

Summarize legislative bills

Compare zoning ordinances

Identify best practices

Summarize case law—then verify it

Explain new statutes for a specific audience

17 of 40

RESEARCH

Live Example: Utah ADUs

Prompt: “Summarize Utah’s ADU legislation for elected officials.”

Then ask for: key changes, local discretion, implementation issues, and source links

Planner task: verify statutory language and current effective dates

18 of 40

USE CASE 2

Staff Reports

Draft findings of fact

Analyze consistency with adopted plans

Summarize comprehensive-plan policy

Build meeting agendas

Create executive summaries

Turn rough notes into a structured first draft

19 of 40

USE CASE 3

Code Writing & Code Review

Rewrite zoning language

Simplify definitions

Identify potential conflicts

Compare development standards

Draft ordinance amendments

Ask AI to flag ambiguity—not to decide legislative intent

20 of 40

USE CASE 4

Public Communication

Public notices

FAQs

Application Checklists

Website content

Social media

Council presentations

Translate one concept for engineers, elected officials, and residents

21 of 40

USE CASE 5

Economic Development

Incentive summaries

Demographic analysis

SWOT analysis

Development comparisons

Market summaries

Convert technical material into decision-ready briefs

22 of 40

USE CASE 6

Project Management

Meeting summaries

Action items

Schedules

Task tracking

Presentation outlines

Convert notes into owners, deadlines, and follow-up items

23 of 40

USE CASE 7

Public Meetings

Generate difficult public questions

Prepare commissioner briefing packets

Draft neighborhood FAQs

Summarize hearing themes

Role-play competing stakeholder perspectives

24 of 40

LIVE DEMONSTRATIONS

Centaur Chess: Human + Machine

After Deep Blue, chess did not disappear—it evolved.

Human + engine teams could outperform either alone.

The advantage came from the process for combining machine speed with human judgment.

Planning with AI is a workflow design problem, not a replacement contest.

25 of 40

DEMONSTRATION

Demo 1: Conditional Use Findings

Start: “Draft findings supporting approval of a conditional use permit for a childcare center.”

Evaluate the generic first output

Add ordinance criteria, facts, role, audience, and constraints

Compare the improved output

Lesson: prompt quality—and supplied context—matter

26 of 40

DEMONSTRATION

Demo 2: Zoning Ordinance Review

Upload the ordinance

Summarize setbacks

Compare standards across zones

Identify inconsistencies or ambiguous provisions

Ask AI to cite the exact section for every conclusion

Verify against the source document

27 of 40

DEMONSTRATION

Demo 3: Meeting Notes → Work Product

Input meeting notes

Generate a concise summary

Extract action items and owners

Draft a follow-up email

Prepare a short council update

Review before distribution

28 of 40

USE AI RESPONSIBLY

The Risks Every Planner Should Know

Hallucinations

Bias

Privacy & confidentiality

Transparency

Professional responsibility

29 of 40

Hallucinations

“AI can confidently make things up.”

Verify statutes, cases, quotations, numbers, citations, and factual claims before relying on them.

30 of 40

RISK CONTROL

Privacy & Confidentiality

Do not upload confidential applications

Do not upload personnel information

Do not upload attorney-client communications

Do not upload protected records

Know your organization’s AI policy and approved tools

31 of 40

RISK CONTROL

Bias, IP & Transparency

AI can reproduce bias present in data and language patterns

Understand ownership and licensing implications

Consider public-records requirements

Decide when AI assistance should be disclosed

Document meaningful human review when appropriate

32 of 40

NON-DELEGABLE

Professional Judgment Stays Human

Recommendations

Findings

Decisions

Ethics

Balancing competing community values

Accountability to elected officials and the public

33 of 40

A Four-Part Prompt

ROLE — “You are a senior land use planner…”
CONTEXT — “This project is…”
TASK — “Prepare…”
AUDIENCE — “For elected officials…”

Example:

You are a municipal planner. Summarize this zoning amendment for residents using plain language at an eighth-grade reading level.

34 of 40

START LOW-RISK

Five Ways to Start Tomorrow

Summarize a long public document

Draft meeting notes and action items

Improve or simplify writing

Brainstorm alternatives

Create presentation materials

35 of 40

A Simple Operating Rule

“Start with low-risk tasks. Supply context. Verify the output. Own the final work.”

AI should shorten the distance from information to judgment—not replace judgment.

36 of 40

THE FUTURE OF PLANNING WORK

Closing Takeaways

AI will not replace planners.

Planners who use AI effectively may outperform those who do not.

The advantage comes from combining technical expertise, critical thinking, ethical judgment, and AI-assisted productivity.

37 of 40

AUDIENCE Q&A

Questions & Discussion

Where could AI remove friction in your current workflow?

What work should never be delegated to AI?

What guardrails does your organization need?

38 of 40

CONTINUE LEARNING

Resources & Further Reading

American Planning Association — ethics, PAS, technology resources

National League of Cities / ICMA / GFOA — local-government AI governance

Stanford AI Index — adoption, capabilities, limitations

NIST AI Risk Management Framework — governance and trustworthy AI

OECD AI Principles — accountability, oversight, fairness, transparency

Utah AI law — Artificial Intelligence Policy Act S.B. 149 (2024), S.B. 226 (2025)

Books: Co-Intelligence; Superagency; Prediction Machines; AI for Good: How Real People Are Using Artificial Intelligence to Fix Things That Matter; The Reverse Centaur's Guide to Life After AI

39 of 40

HIGH-VALUE TAKEAWAY

Build a Planning Prompt Library

Staff reports

Comprehensive plans

Code amendments

Development agreements

Public notices

Planning commission findings

Council presentations

Meeting summaries

RFQs / RFPs

Grant applications

40 of 40

PRESENTER REFERENCE

Presenter Run of Show

Run of Show Category	Time	Slides	Presenter	Focus
1. Welcome & Setting the Stage	5 min	1-7	Jason	Introduction, audience poll, AI is not replacing planners, “planning intern” analogy
2. AI in Plain English	5 min	8-14	Micah	Human brain opener, what AI/LLMs are, common tools, strengths vs. weaknesses
3. Where AI Fits Into Planning	15 min	15-23	Jason / Micah	Research, staff reports, code review, communication, economic development, PM, public meetings
4. Live Demonstrations	10 min	24-27	Micah	Centaur Chess concept + three live AI demonstrations
5. Risks Every Planner Should Know	5 min	28-32	Jason / Micah	Hallucinations, privacy, bias, IP, transparency, professional judgment
6. Practical Prompting Tips	3 min	33	Micah	Four-part prompting framework + example
7. Getting Started Tomorrow	2 min	34-36	Jason / Micah	Five practical starting points + operating rule
8. Closing & Q&A	Flexible	37	Jason / Micah	Key takeaways and audience discussion
9. Resources / Take-Home Material	Reference	38-39	Jason / Micah	Further reading and planning prompt library
Presenter Reference	Not presented	40	Jason / Micah	Run-of-show/reference material