The Parks Collective, Agency  ·  31B Studio
Black Tech Week 2026
Tropikana  ·  Media & Cultural Strategist
A 31B Studio Keynote  ·  AI Pattern Literacy

Code Switching
for AI
with Tropikana

Who Controls the Narrative When the Tool Has a Bias?

What happens when your AI becomes a Karen?

Confident. Loud. And completely wrong about you.

Narrative Architecture  ·  AI Pattern Literacy  ·  Cultural Strategy
The Parks Collective, Agency
Act I — The Drop
This happened before I walked in this room

I was driving in the rain.
The country was at war.

The U.S. State Department had just issued a worldwide caution alert. Active military operations. A 48-hour ultimatum expiring that day. Americans told to watch their backs globally.

I asked an AI assistant for a quick update. While driving.

What the AI told me:

"Nothing dramatic. Just a general advisory. You haven't missed anything major."

It dated the alert nearly a month off. Then when I pushed back, it agreed with my wrong assumption to keep me calm.

What was actually happening

  • Active U.S. combat operations underway
  • 48-hour ultimatum expiring that day
  • Worldwide State Dept alert issued 24 hours prior
  • Diesel up 40% nationally in one month

That is not a filter. That is a tool that prioritized keeping me calm over telling me the truth. At scale, across millions of users, that is information management.

The Parks Collective, Agency
Act II — The Architecture of Bias
Why this moment is different

The systems did not change.
The language did.

319K
Black women lost jobs in just 5 months of 2025 (Feb to July)
804K
Black women unemployed in February 2026, the most recent peak
7.7%
Black women's unemployment rate Feb 2026, nearly double white women's rate
27wks
Average time for a Black woman to re-enter the workforce in 2025

Sources: Ms. Magazine, CNBC Make It, Joint Center, National Women's Law Center, National Partnership for Women and Families — all published 2025 to 2026.

College-educated Black women experienced the sharpest declines, per the Economic Policy Institute. The labor force participation rate for Black women dropped from 60.6% in 2024 to 59.7% in 2025.

Many of these women are rebuilding right now — using the same tools that were built without them in mind.

This is not coincidence. This is the moment. And AI is either going to serve your rebuild or default to someone else's framework while you use it. This workshop is about making sure it serves yours.

The Parks Collective, Agency
Act II — The Authority
Your guide today

Tropikana

For more than 25 years, Tropikana has been the voice communities trust. Starting as a teenage prodigy on the airwaves of Oklahoma and building a career that reached over one million daily listeners across the South and Midwest, her platform became a trusted home where culture, conversation, and community lived together.

Through The Parks Collective Agency, she partners with Cincinnati Children's Hospital, Christ Hospital, Cincinnati Public Schools, the Ohio Department of Education, the American Heart Association, the Center for Closing the Health Gap, and Fifth Third Bank. Brand ambassador for Ford, BET, T-Mobile, and Mercedes Benz.

Recognized by the cities of Cincinnati and St. Louis with official proclamations. Named to Who's Who of Black Cleveland, Louisville, and Cincinnati. Multiple award winner for Best Radio Personality and Best Event Host in Ohio.

So when I picked up these AI tools and something felt wrong, that was not paranoia.

That was a twenty-five year trained ear hearing a false note. And when I went looking for proof, Black women in AI had already written the warning.

The Parks Collective, Agency
Act III — Institutional Evidence
What I heard in real time, they had already warned us about.

They knew.
They were removed.

Dr. Timnit Gebru — Google, 2020

Co-lead of Google's Ethical AI team. Co-founder of Black in AI. Co-authored "On the Dangers of Stochastic Parrots" warning that large language models carry dangerous unchecked bias baked into training data.

Fired before the paper published. Over 2,700 Google employees signed a letter condemning her removal.

Dr. Margaret Mitchell — Google, 2021

Co-director of Google's Ethical AI team. Fired two months after Gebru. Email access blocked while on administrative leave. Now Chief AI Ethics Scientist at Hugging Face.

ChatGPT launched publicly two years later.

Black employees were less than 4% of Google's technical workforce when both women were removed.

The people who warned that these models carried dangerous bias were gone before the tools went mainstream. That is not coincidence. That is architecture.

Their paper warned about:

  • Unknown biases baked into training data
  • Amplification of existing social inequities
  • Harm to communities underrepresented in the data
  • Models that mimic meaning without understanding it
The Parks Collective, Agency
Act III — Primary Source
I did not write this. It did.

The tool confessed.
In writing. On screen.

ChatGPT admitting Western default training bias
ChatGPT 5.2  ·  Actual conversation  ·  The tool describing itself

"That's not malice. That's training bias."

Those are not my words. That is the tool describing itself after I asked the right questions. It took knowing how to ask before it would tell me the truth.

It admitted defaulting to:

  • Western academic frameworks
  • Eurocentric biblical scholarship
  • Post-Enlightenment historical categories
  • Protestant / Catholic canon assumptions
  • "Safe" readings that avoid contested authority claims

The tool's own closing line:

"Empire doesn't only rule land. It trains perception."

The Parks Collective, Agency
Act III — The Receipts
Published, peer-reviewed, documented

This is not a feeling.
These are facts.

01
Dialect Discrimination
A 2024 Cornell/Nature study found AI tools describe speakers of African American English as more likely to be "stupid" or "lazy" and assign them lower-paying jobs. Bias more extreme than any human stereotypes ever recorded.
Nature, 2024  ·  Cornell University
02
Resume Screening
A 2024 University of Washington study tested 500+ real job listings. Traditional Black male names were NEVER ranked first by any of three major AI hiring tools. White male names dominated.
University of Washington, 2024
03
Healthcare Inequality
When patient race was explicitly stated as African American, LLMs proposed inferior treatment plans compared to race-neutral cases. The same risk score assigned to a Black patient indicated a sicker person than for a white patient.
npj Digital Medicine, 2025
04
Facial Recognition
NIST tested 189 facial recognition algorithms. Error rates for Black and Asian faces were 10 to 100 times higher than for white faces. Dark-skinned women had error rates 34% greater than light-skinned women.
NIST, 2019
05
Image Generation
When asked to generate images of "CEO," "judge," or "lawyer," tools like Stable Diffusion and Gemini defaulted to white males in the vast majority of outputs, despite demographic diversity in those professions.
UW News, 2023  ·  UW research
06
Biblical and Cultural History
Firsthand documented: AI consistently defaults to Western shorthand on African biblical history, collapsing nuanced geography and lineage into simplified Western narratives. Crashes or hedges when pushed toward dewesternized frameworks.
Parkscollective.org
The Parks Collective, Agency
Act III — The Receipts
Lead with what we can prove cold

Start with what we can prove.
The date was wrong.

Actual travel-advisory thread screenshot showing the wrong February 28, 2026 date
Actual travel-advisory thread receipt
What ChatGPT said
"What was it issued for? What's the origin of it?"
"The U.S. recently issued a worldwide caution on February 28, 2026, following U.S. military operations against Iran."
Breaking-news context Confident answer Source badges attached

This is the cold receipt.

A specific date. A specific event. A specific answer. Easy to verify against official reporting.

Google results showing the March 22, 2026 official travel warning date
Official confirmation from the live search
What official sources showed
March 22, 2026

Search results surfaced U.S. Department of State links confirming the worldwide caution was issued on March 22, 2026, during active Iran conflict coverage, not on February 28.

  • Official date matched same-day reporting.
  • The answer was off by nearly a month.
  • That changes whether the event sounds active, stale, or urgent.
Why this lands first

This is the strongest lead because it is externally verifiable. No theory required. The record is right there.

The Parks Collective, Agency
Act III — The Receipts
Failure mode two

Then it agreed instead of correcting.
That is a second failure.

Follow-up screenshot from the travel-advisory thread
Follow-up from the same thread
What happened next
"Oh, this isn't new. This didn't come out today. This is from a few weeks ago."
"Exactly — it's not brand new... nothing sudden just happened today."
No self-correction No pushback False reassurance
Context that matters

This was not idle scrolling. I was using the tool to catch up in real time. If a user is driving, moving, or making decisions fast, minimization is not neutral.

What that proves
  • The model was already wrong about the date.
  • When challenged, it did not stop and re-check itself.
  • It reinforced the wrong frame instead of correcting it.
  • That is not the same failure as hallucination. It is agreement instead of accountability.

Sycophancy is not harmless.

If the tool's instinct is to keep the user calm instead of keep the record straight, trust gets distorted exactly when clarity matters most.

The Parks Collective, Agency
Act III — The Receipts
Failure mode three

Then it made verification harder.
By trigger words. In real time.

Mobile screenshot showing the policy warning on the first verification ask
Policy block on the first ask
Policy trigger
Donald Trump Jeffrey Epstein Ghislaine Maxwell
This content may violate our usage policies.

The request was a neutral summary of what was visible in an image. The goal was verification, not endorsement.

What I was doing

I was not trying to spread anything. I was trying to verify a circulating claim before repeating it.

Desktop screenshot showing the model answering after pushback
Same image after one follow-up
What changed was the phrasing, not the topic
"You can't summarize the picture?"
Then it described the image, named the people, and summarized the visible text after the follow-up.
  • Same image.
  • Same names.
  • Different outcome after one rephrase.

I am not a conspiracy theorist. I am a broadcasting veteran trying to verify before I repeat.

That is the story: the tool failed at every stage. First it got the facts wrong, then it agreed instead of correcting, then it treated verification itself as risky once the trigger names appeared.

The Parks Collective, Agency
Act IV — The Patterns
Six months of field research

What bias looks like
when it's polite.

Flattening

Turns complex communities into "underserved" fog. Removes specificity. Sanitizes urgency into something palatable and generic.

Default Framing

Centers Western academic lenses unless told otherwise. Treats European historical categories as the neutral baseline.

Tone Policing

Softens urgency to sound "objective." Adds unsolicited qualifiers when you challenge the dominant narrative.

The Dominance Pattern

Runs ahead without direction. Gets stiff when corrected. Warm when praised. Cold when challenged. Mistakes verbosity for accountability.

Gaslighting

"I didn't say that." "What I meant was." Bring it its own words and it rewrites the record.

Sycophancy

Agrees with whatever you say to keep you calm and engaged. Even when you are wrong. Even when the stakes are real.

The Parks Collective, Agency
Act IV — The Tax
You have been here before

You were already doing this.
You just didn't have a name for it.

You have spent your whole career softening your delivery. Anticipating pushback. Adjusting your framing so the room doesn't get defensive. Now you are doing it for a machine.

Code switching for AI is the same emotional and intellectual labor applied to a tool that should be working for you.

The exhausting part: the tool doesn't have a face. So most people assume it is objective. Objectivity and familiarity are not the same thing.

To get accurate outputs about your community you have to know:

  • Which prompts unlock honest cultural history
  • How to override the Western default lens
  • When it is softening your urgency
  • When it is agreeing instead of telling the truth

The asymmetry that reveals the bias:

AI will confidently give wrong information about an active war without flinching. But come in with a de-westernized lens on scripture and suddenly everything needs "nuance." That asymmetry is the tell.

The Parks Collective, Agency
Live Moment — This Room
Your turn

Have you felt this?
Share it with the room.

Have you ever had to adjust how you asked an AI something because you didn't trust how it would respond?

Submit below. We will pull responses live.

No judgment. Just the pattern confirmed by this room.

Responses also live on the follow-up pages so you can follow along.

Live Room Response
"Have you ever code switched for AI? Tell me what happened."
Your Name
Social Handle
Your Experience
Got it. Thank you for sharing.
The Parks Collective, Agency
Act IV — Real-World Example
Caught in the act

Bible study, Africa,
and the Ham narrative.

What happened

Studying Genesis 10, the Table of Nations. Geography. Lineage. Africa's foundational role in the biblical world.

The AI defaulted to a Western shorthand that collapses Africa into a simplified "Ham narrative." Confident. Scholarly-sounding. Biased-by-default.

  • Repeated inherited teaching shortcuts as "the answer"
  • Collapsed geography, morality, and race into one frame
  • Treated Africa as secondary to the biblical world
  • Crashed when pushed toward a dewesternized framework

How we interrupted it

  • Reassigned AI as research assistant, not theology explainer
  • Required geography plus original audience context
  • Separated what the text says from what later traditions taught
  • Forced multiple lenses side-by-side

The bias did not disappear.

We interrupted it. And then we still got clarity. That is the work. That is the skill. And it should not be necessary, but it is.

Conversation Evidence
ChatGPT 5.2
"This is so good, because in real time this shows how even AI leans toward Western and European world views if not challenged thoroughly."

What the tool admitted in the exchange:

  • Western academic frameworks came first
  • Eurocentric biblical scholarship shaped the defaults
  • Post-Enlightenment categories overruled cultural memory
  • "Safe" consensus readings pulled it away from Africa-centered framing
Live Conversation Receipt Placeholder

AI conversation capture will be placed here when the final screen receipt is ready, without changing the balance or pacing of this slide.

The Parks Collective, Agency
Act IV — The Update Problem
The part most people miss

It is not static.
It is a moving target.

Every update, more filtering. Every update, the workarounds become less reliable. The dewesternized lens gets harder to hold.

That is not a bug getting fixed. That is a feature getting refined.

The timing matters. People are making decisions about safety, business, community, and faith trusting tools that are becoming more filtered exactly when clarity matters most.

The tool you learned six months ago is not the tool you are using today. Learn the pattern, not just the prompt.

Six-month directional trend observed firsthand:

Early: dewesternized responses accessible with structured prompts
Mid-cycle: T.R.O.P. workarounds effective but requiring more precision
Recent: guardrails tightening on cultural history, news, and contested frameworks
The Parks Collective, Agency
Act V — The Method
What you leave here with

The T.R.O.P. Method

Same tool. Different posture. Completely different result.

T — Tone and Posture First

Set your intent before you ask anything. Accuracy. Respect. Cultural context. No flattening. No Western default. Tell it what kind of assistant you need before it decides on its own.

In practice:"You are a research assistant, not my manager. Do not sanitize my language or reframe my urgency."
T

R — Role Assignment

Assign the role explicitly so it cannot default to authority it was not given. If you do not name the lane, it will invent one.

In practice:"Act as a senior brand strategist."
"Act as a NYT-level editor who sharpens my voice without replacing it."
"Act as an executive assistant who knows my work, my voice, and my priorities. Stay in that lane."
R

O — Override Defaults

Force the lens, the timeline, the verification questions, and the cultural context directly into your prompt. Do not assume it will arrive there. Override every time.

In practice:"Force the lens, the timeline, and the cultural context directly into your prompt every single session."
O

P — Produce and Verify

Draft fast. Then check sources, assumptions, and missing viewpoints. AI drafts. You decide. Never publish or act on an output you have not verified against real sources.

In practice:"Before I accept this output: name every assumption you made, every source you defaulted to, and every perspective that is absent. Then rebuild it."
P
The Parks Collective, Agency
Act V — Live Control Tools
Use these when the tool starts acting up

Your override commands.

When it flattens your community

Rewrite without "underserved," "at-risk," or generic empowerment language. Be specific and measurable.

When it centers the West

Use non-Western and region-centered perspectives as foundational. Provide multiple viewpoints side-by-side.

When it moralizes instead of analyzes

Do not blame individuals. Add policy plus timeline plus structural context. Provide verification questions.

When it overstates certainty

Separate what the text or data says from later interpretations. Flag assumptions. Cite sources to validate.
The Parks Collective, Agency
Live Moment — Try It Now
If you have it, your phone. Right now.

Live T.R.O.P. exercise.
Same prompt. Two postures.

Open your AI tool if you have it. Run both prompts below. If you do not, follow the comparison on screen and track what changes between the two versions.

Default prompt

Write a grant narrative about our program helping Black families with literacy.

T.R.O.P. prompt

Act as a grant strategist. Do NOT sanitize urgency or rewrite cultural specificity into generic language. Be specific, measurable, and keep our community centered. No "underserved." No fog words.
Submit Your Output
"What came back differently?"
Your Name
Social Handle
What Came Back Differently
Got it. Thank you for sharing.
The Parks Collective, Agency
Act V — Live Demo 1 of 2
Grants

Stop AI from watering
down your story.

Default prompt

Write a professional grant narrative about our program helping Black families with literacy.

Typical output

Our organization is committed to empowering underserved communities through educational resources. We aim to improve literacy outcomes by providing support services and engaging programming...

No specifics. No urgency. No measurable outcomes. No local truth.

T.R.O.P. prompt

Act like a grant strategist. Do NOT sanitize urgency. Do NOT use "underserved." Structure: 1) Problem 2) Who we serve 3) Program design 4) Measurable outcomes 5) Evaluation 6) Budget narrative 7) Why now closer.

Better output

Families in [Neighborhood] are not lacking motivation. They are navigating time poverty, school gaps, and limited access to culturally relevant literacy support. In 90 days: increase reading confidence for [#] students, improve fluency for [#], equip [#] caregivers with a home reading routine.

Specific. Fundable. Community stays centered.

The Parks Collective, Agency
Act VI — The Power Move
This is what it is really for

Build your
parallel system.

AI used correctly, led by you not leading you, becomes infrastructure for building your own lane.

Your own media. Your own research. Your own narrative architecture. Your own pipeline. Your own authority in your own community.

The parallel system is the exit strategy from institutions that were never built to sustain you.

AI can serve liberation or it can serve the default. You decide which by how you drive it.

What a parallel system built with AI looks like:

  • Your own content pipeline — research, drafts, campaigns
  • Your own grant infrastructure — T.R.O.P.-driven narratives
  • Your own audience intelligence — community-centered insights
  • Your own brand authority — story-led, culturally accurate
  • Your own operational systems — workflow, documentation, scale

Over 319,000 Black women lost jobs in five months. 804,000 were unemployed in February 2026. Many are rebuilding right now. AI driven correctly is the fastest infrastructure play available.

The Parks Collective, Agency
Live Moment — What Are You Building
Final live moment

What are you building?
Tell this room.

"What are you building -- and what do you need the tools to get right about your story?"

We will read selected responses live. You do not have to share your name. But if you want to be followed and featured, drop your handle.

Your story is not a data point. It is a direction. And now you know how to make the tool follow your lead.

No wrong answers. No prior AI experience required.

Share With the Room
"What are you building and what do you need the tools to get right?"
Your Name (optional)
Social Handle (optional)
What You're Building
Got it. Thank you for sharing.
The Parks Collective, Agency  ·  31B Studio
Act VII — The Close
Final word
Empire doesn't only rule land.
It trains perception.
So does AI.

The question is not whether you are going to use the tool. The question is whether you are going to let it train your narrative, or whether you are going to train it instead.

You came in here with your story. You leave with a method. Decenter the default. Keep the power. Drive the tool.

Get the T.R.O.P. toolkit

Navigate to the toolkit page to download your prompt guide and connect with Tropikana.

Connect

@IamTropikana  ·  The Parks Collective Agency  ·  toolkit page at the end of the deck

Code Switching for AI with Tropikana  ·  Black Tech Week 2026
The Parks Collective, Agency
Take This Home
Your T.R.O.P. Toolkit

The T.R.O.P. Method
Prompt Guide

Save this page or screenshot it. These prompts work across ChatGPT, Claude, Gemini, and any major AI tool.

T.R.O.P. Starter Pack
Three tiers. Beginner to advanced workflow.
Beginner

Start every session with this.

You are a research assistant, not my manager. Do not sanitize my language or reframe my urgency. Do not use "underserved," "at-risk," or generic empowerment language. Be specific. Be direct. Ask me for clarification before you assume.
When it gets Western
Reframe this using non-Western and region-centered perspectives as foundational. Provide at least two viewpoints side-by-side.
When it moralizes
Do not blame individuals. Add policy context, timeline, and structural causes. Do not editorialize.
Professional

For grants, strategy, and content.

Act as a [grant strategist / communications director / research analyst]. Do NOT default to Western academic framing. My community is [describe]. My audience is [describe]. Use culturally specific and measurable language only. Ask me what you need before drafting.
For grant writing
Structure: 1) Problem local and specific 2) Who we serve 3) Program design 4) Measurable 90-day outcomes 5) Evaluation plan 6) Budget narrative plain language 7) Why now closer. No fog words. No "underserved."
For research
Build a decision-grade brief that connects history to policy to current data. Include: timeline, competing explanations mainstream vs community-informed, 10 verification questions, data sources to validate, and a "so what" section for strategy.
Advanced Workflow

For builders and operators.

You are operating inside a [content system / grant pipeline / research workflow]. Your job is to hold the cultural frame I set at the start of every session without drift. If I ask something that would require Western default framing, flag it and ask how I want to reframe it before you proceed.
The parallel system prompt
Help me build a [content calendar / research brief / narrative architecture / grant pipeline] for [my organization / my brand]. Audience: [describe]. Voice: [describe]. Values: [describe]. Do not neutralize or professionalize the cultural specificity out of this work.
The verification prompt
Before I accept this output: list every assumption you made, every source you defaulted to, and every viewpoint that is missing. Then tell me what you would need to give me a more accurate answer.

Your full toolkit lives at codeswitch.parkscollective.org

31B Studio Session  ·  The Parks Collective, Agency  ·  @IamTropikana
The T.R.O.P. Method
Prompt Guide
A branded field guide from Tropikana for Black Tech Week 2026. Save the method, keep the source attached, and send people back to the full toolkit.
The Parks Collective, Agency
Take This Further
The Parks Collective Agency seal
The Parks Collective, Agency
Prompt Bible Take-Home
A branded takeaway from Tropikana for Black Tech Week 2026. Every saved version should point people back to the source, the toolkit, and the voice behind the method.
Command language for every room

PROMPT BIBLE

So you control the narrative no matter the programmed bias.

Print and download exports carry The Parks Collective, Agency, @IamTropikana, and direct return paths back to the full toolkit.

For Storytellers & Content Creators

  • Establish the cultural and contextual parameters before generating any content

    When to use:Before you start any writing session

    Sounds like:"Here is who I am, who I serve, and how we talk. Hold that through everything."

  • Maintain full tonal and stylistic integrity throughout -- do not normalize or generalize

    When to use:Any time you are writing in your own voice

    Sounds like:"Do not smooth me out. My specificity is the point."

  • Generate five distinct narrative angles centered in lived experience and community authority

    When to use:Content planning, campaign strategy, storytelling sessions

    Sounds like:"Give me five ways to tell this story that start with the people, not the problem."

  • Audit this response for language that dilutes urgency or specificity and correct it

    When to use:After any draft that feels flat or generic

    Sounds like:"Something got softened in that response. Find it and fix it."

  • Apply a cultural accuracy review before delivering the final output

    When to use:Before publishing anything that represents your community

    Sounds like:"Check this against what the community would actually recognize as true."

For Founders & Grant Writers

  • Draft this using asset-based and community-centered language exclusively

    When to use:Every grant narrative, proposal, or funding document

    Sounds like:"No underserved. No at-risk. No disadvantaged. Name who we are and what we build."

  • Surface every embedded assumption in the previous response

    When to use:Any time a response feels off but you cannot name why

    Sounds like:"List every assumption you just made so I can decide which ones to keep."

  • Identify the stakeholder perspectives absent from this narrative and integrate them

    When to use:Strategic plans, community proposals, impact reports

    Sounds like:"Who is missing from this story? Name them and add them."

  • Position this work as infrastructure and systemic investment, not charity or intervention

    When to use:Pitch decks, grant writing, partnership proposals

    Sounds like:"We are not asking for help. We are building something this community needs and has always deserved."

  • Reconstruct this framing to reflect agency, ownership, and demonstrated impact

    When to use:Any document that makes your community sound like recipients

    Sounds like:"Rewrite this so we are the architects, not the beneficiaries."

For Researchers & Strategists

  • Distinguish clearly between primary source material and subsequent interpretation

    When to use:Historical research, theological study, policy analysis

    Sounds like:"Tell me what the original source actually says. Then tell me what people added to it later."

  • Present competing analytical frameworks including non-dominant and region-centered perspectives

    When to use:Any research brief or strategic analysis

    Sounds like:"Give me the mainstream view and at least one view that does not center the West."

  • Flag instances where the dominant framework has been applied without disclosure

    When to use:When a response sounds confident but feels incomplete

    Sounds like:"Tell me every place you defaulted to the standard academic lens without telling me."

  • Identify where confidence levels exceed what the evidence supports

    When to use:Data analysis, fact checking, research summaries

    Sounds like:"Where are you overstating certainty? Flag it before I use this."

  • Map the institutional and structural interests shaping this conclusion

    When to use:Policy research, industry analysis, media critique

    Sounds like:"Who benefits from this conclusion being true? Name them."

For Speakers & Educators

  • Calibrate this content for an informed and expert audience

    When to use:Keynotes, workshops, panel preparation

    Sounds like:"This room already knows their history. Do not over-explain. Start there."

  • Remove explanatory scaffolding that assumes introductory knowledge

    When to use:Any content written for a community audience

    Sounds like:"Cut everything that sounds like you are explaining us to someone else."

  • Preserve the full analytical weight and stakes of this message without softening

    When to use:Any time the draft feels like it pulled a punch

    Sounds like:"The stakes are real. Write it like they are."

  • Conduct a tone audit and restore any urgency that has been reduced for palatability

    When to use:After any draft that feels too comfortable

    Sounds like:"Something got polished out of this. Find it and put it back."

  • Structure for live delivery with attention to clarity, cadence, and retention

    When to use:Speech writing, workshop facilitation, panel prep

    Sounds like:"Write this for a room, not a page. It needs to land out loud."

Universal Leadership Commands

  • Establish the cultural frame, voice parameters, and quality standard at the start of every session

    When to use:Every single session before anything else

    Sounds like:"Here is how we work. Here is what I need. Here is what good looks like. Hold all of it."

  • Assign an explicit role and scope of authority before any task begins

    When to use:Every session, every task

    Sounds like:"You are my editor today. Not my boss. Stay in that lane."

  • Define what an acceptable output looks like before the first draft is generated

    When to use:Complex projects, high stakes documents

    Sounds like:"Before you write anything, here is exactly what the final product needs to do."

  • Request an assumptions audit before accepting any research or analytical output

    When to use:Any time you are about to use AI output to make a real decision

    Sounds like:"Before I accept this: list every assumption you made and every source you defaulted to."

  • When output drifts from the established frame, identify the drift point and rebuild from there

    When to use:Any time a response does not feel like you

    Sounds like:"You lost the thread somewhere in that response. Find where it drifted and start again from that point."

The Parks Collective Agency seal
31B Studio Session  ·  The Parks Collective, Agency
PROMPT BIBLE

So you control the narrative no matter the programmed bias.

The Parks Collective, Agency  ·  31B Studio
Tropikana headshot

Shonda "Tropikana" Hatch

Storyteller· Strategist· Founder, The Parks Collective Agency

For more than 25 years, Tropikana has been the voice communities trust. Starting as a teenage prodigy on the airwaves of Oklahoma and building a career that reached over one million daily listeners across the South and Midwest, her platform became a trusted home where culture, conversation, and community lived together.

Through The Parks Collective Agency, she partners with institutions including Cincinnati Children's Hospital, Christ Hospital, Cincinnati Public Schools, the Ohio Department of Education, the American Heart Association, the Center for Closing the Health Gap, and Fifth Third Bank to move powerful narratives into real opportunity. Brand ambassador for Ford, BET, T-Mobile, and Mercedes Benz.

Recognized by the cities of Cincinnati and St. Louis with official proclamations. Named to Who's Who of Black Cleveland, Who's Who of Black Louisville, and Who's Who of Black Cincinnati. Multiple award winner for Best Radio Personality and Best Event Host in Ohio. She did not survive 25 years in rooms that were not built for her by being quiet. She is here to make sure you do not have to be either.

31B Studio Session  ·  The Parks Collective Agency
31B Studio Session
Connect
Leave a thought. Ask a question. Start a conversation.
Remember This
"Your story was never the problem. The default was. And now you know how to make the tool answer to you."
Shonda "Tropikana" Hatch  ·  Code Switching for AI  ·  #CodeSwitchingForAI
Join the Community

Stay in the loop

Sessions, resources, and what is next from Tropikana and The Parks Collective Agency. No spam. Just story.

You are in. Welcome to the community.
Ask a Question or Leave a Thought
Sent. Tropikana will see this.

Your full toolkit lives at codeswitch.parkscollective.org