What if your marketing team could produce a polished social graphic in five minutes without a stock photo subscription? What if your branch manager could create a custom event flyer before lunch?

AI image generation is making both of those things happen at credit unions right now. The tools have gotten good enough that the output looks professional. The challenge is knowing which tool to use, how to prompt it well, and where the compliance guardrails are.

What Is AI Image Generation—and Which Tools Should Credit Unions Know?

AI image generators take a plain-English text prompt and produce an original image in seconds. No photographer. No stock library. No per-image licensing fee (with some caveats we'll cover below).

Three tools are worth knowing for credit union marketing work.

DALL-E, built into ChatGPT and available through OpenAI's API, is the most accessible starting point. If your team already uses ChatGPT, you have DALL-E built in. It handles photorealistic scenes and illustrated styles equally well and takes conversational revision prompts naturally.

Adobe Firefly lives inside Adobe Express and the broader Creative Cloud suite. If your marketing team works in Photoshop or InDesign, Firefly fits right into that workflow. Adobe has been deliberate about training Firefly only on licensed content, which matters for commercial use. As of 2024, Firefly is positioned as commercially safe by design—but verify current licensing terms directly with Adobe before using outputs in paid advertising.

Ideogram is newer and less well-known, but it has one skill the others struggle with: rendering legible text inside images. If you need a social graphic with a headline baked in—"Annual Meeting, March 14"—Ideogram handles that better than the competition.

The best AI image tool is the one your team will actually use. Start with whatever plugs into the workflow you already have, and expand from there.

How to Write Prompts That Actually Produce Usable Images

Vague prompts produce generic images. Specific prompts produce usable ones. A framework that works well for credit union marketing: subject + context + style + mood + technical specs.

Here's the difference in practice:

• Vague: "woman at bank"
• Specific: "Middle-aged woman smiling at a credit union teller window, modern branch interior, warm overhead lighting, photorealistic, shallow depth of field, horizontal 16:9 format"

The specific version wins every time. A few techniques to build into your prompts:

1. Name the setting explicitly. "Credit union branch lobby" or "outdoor community event" beats "financial services environment." The more grounded the context, the more relevant the image.
2. Specify diversity intentionally. AI models have default patterns. If your membership spans demographics—and most credit union memberships do—prompt for it. "Diverse group of adults ranging from 20s to 60s" is a reasonable starting phrase. Always review outputs before publishing.
3. Anchor the style. Words like "photorealistic," "flat illustration," "watercolor," or "editorial photography style" steer the aesthetic hard. Pick one and stick with it across a campaign.
4. Lock in dimensions early. "Horizontal 16:9" for social banners, "square 1:1" for Instagram posts, "vertical 9:16" for Stories. Cropping after the fact loses quality and composition.

Three Real Credit Union Use Cases

Social media posts

Your loan ops team wants to promote a HELOC special. Prompt DALL-E: "Happy homeowner couple standing in front of a freshly painted house exterior, suburban neighborhood, bright sunny day, photorealistic, square format." Generate three variations. Pick the strongest. Drop it into Canva with your rate and compliance disclosures. Done in under ten minutes.

Internal communications

Your HR team is rolling out a wellness initiative. Instead of a tired stock photo of someone jogging, prompt Ideogram: "Colorful illustrated graphic, text reads 'Your Wellbeing Matters', vibrant geometric background, modern flat design style." Ideogram's text rendering makes this practical where other tools fall short.

Event materials

Your branch is hosting a first-time homebuyer workshop. Prompt Firefly inside Adobe Express: "Illustrated flyer background, cozy community room setting, warm earth tones, no people, space for text overlay, horizontal banner format." Your designer adds the event details on top. Half the layout work is already done.

How to Build a Consistent Visual Style Across Campaigns

One of the most common frustrations with AI image generation is inconsistency. The image you generate on Monday looks nothing like the one from Friday.

The fix is a style brief you paste into every prompt. Write it once. Save it somewhere your whole team can find it. It might look like:

"Photorealistic, warm natural light, diverse community members, modern credit union branch or outdoor community settings, optimistic and approachable tone, no text in image, horizontal 16:9."

Every prompt starts with that block. Then you add the specific scene. Your outputs won't be identical, but they'll feel like they belong to the same family—which is what visual consistency actually means.

One more thing: AI tools don't know your primary blue is #003DA5 or that your brand never uses script fonts. Use generated images as a starting layer, then apply your brand standards in your design tool of choice.

What to Avoid: Compliance and Brand Guardrails

This is where you need to slow down.

Don't use AI-generated faces in lending promotions without a compliance review. Fair lending examiners may scrutinize promotional imagery that appears to target or exclude certain demographic groups. AI-generated faces look real. That's exactly what makes them worth checking carefully. Consult your compliance team before any AI-generated people appear in advertising subject to fair lending rules.

Zoom in before you publish. AI generators sometimes produce extra fingers, distorted text, or nonsensical background signage. A blurry sign in the background that reads something garbled is embarrassing at best and a compliance flag at worst.

Verify your usage rights. DALL-E, Adobe Firefly, and Ideogram each have different terms around commercial use. As of 2024, all three allow commercial use under their standard paid tiers—but terms change. Confirm directly with each vendor before using outputs in paid advertising or printed member materials.

What's the risk?

For most credit union marketing work—social posts, internal newsletters, event flyer backgrounds, board presentation graphics—the risk profile here is low. You're not exposing member NPI, the outputs are reviewed before publishing, and the tools are widely used in financial services marketing. Go ahead and start experimenting.

The higher-risk area is imagery in advertising subject to fair lending scrutiny, particularly lending promotions. AI-generated human faces can inadvertently skew toward certain demographics, and fair lending examiners do look at imagery. If your workflow touches any advertising covered by ECOA, the Fair Housing Act, or your state's equivalent, loop in your compliance team before you publish. That's a conversation worth having once, upfront, rather than after the fact.

Data privacy risk here is minimal—you're sending text prompts to these tools, not member data. The main risks are accuracy (artifacts in outputs), brand consistency (easy to solve with a style brief), and fair lending optics (manageable with a compliance review cadence). None of those should stop you from starting. They should just shape how you start.

This post was drafted with AI assistance and reviewed by a human at CU 2.0. AI makes mistakes; verify any specific claim before acting on it.

Frequently Asked Questions

This post was drafted with AI assistance and reviewed by a human at CU 2.0. AI makes mistakes; verify any specific claim before acting on it.

The post AI Image Generation for Credit Union Marketing: A Practical Guide to DALL-E, Firefly, and Ideogram appeared first on CU 2.0.

Still got "try Claude Desktop" sitting on your to-do list? Maybe a peer mentioned it at a conference. Maybe someone on your lending or marketing team brought it up. Either way, this is the post that gets it off the list.

Installing Claude Desktop on a Mac takes under five minutes and requires no technical background. Follow the steps below and you'll be signed in before your next meeting.

What Is Claude Desktop?

Claude Desktop is the Mac and Windows application for Claude, Anthropic's AI assistant. It lives in your dock, launches instantly, and—unlike a browser tab—stays put when you need it.

For credit union executives, that means you can open Claude Desktop to draft a board memo, summarize a vendor proposal, or pull together talking points before an examiner visit, without hunting through a dozen open tabs to find your place.

It can also connect to files and tools on your computer through what Anthropic calls Model Context Protocol (MCP) integrations—a feature worth exploring once you're comfortable with the basics.

Claude Desktop gives you a dedicated workspace for AI-assisted thinking. It's not a browser tab you'll accidentally close mid-draft.

The app is free to download. You'll need an Anthropic account, and some features require a Claude Pro subscription. Verify current pricing at anthropic.com.

What You Need Before You Start

Three things, and that's it:

• A Mac running macOS 12 Monterey or later — click the Apple menu, then About This Mac if you're not sure
• A free Anthropic account — create one at claude.ai
• A stable internet connection and about five minutes

No IT ticket is required for most installs. That said, if your credit union manages employee Macs through a device management platform like Jamf or Kandji, check with your IT team first. Those environments sometimes require admin approval before new apps can be installed.

How to Install Claude Desktop on a Mac

Follow these steps in order.

1. Go to claude.ai/download in your browser. This is Anthropic's official download page. Make sure you're on the Mac tab, not Windows.

2. Click "Download for macOS." A .dmg disk image file will land in your Downloads folder, typically named Claude.dmg.

3. Open the .dmg file. Double-click it from your Downloads folder or the browser's download bar. A window will appear showing the Claude app icon and a shortcut to your Applications folder.

4. Drag the Claude icon into the Applications folder. Click and hold the Claude icon, drag it onto the Applications folder shortcut in that same window, and release. That's the install.

5. Eject the disk image. Right-click the mounted Claude disk in Finder's sidebar or on your desktop and choose Eject. This is just housekeeping — the app is already in Applications.

6. Open Claude from Applications. Press Command + Space to open Spotlight, type "Claude," and hit Enter. Or open Finder, navigate to Applications, and double-click Claude.

7. Handle the Gatekeeper prompt. The first time you open any app downloaded from the internet, macOS shows a security dialog: "Claude is an app downloaded from the internet. Are you sure you want to open it?" Click Open. This is normal — macOS is doing exactly what it should.

8. Sign in to your Anthropic account. Claude Desktop will show a sign-in screen. Enter your claude.ai email and password, or use "Continue with Google" if that's how you set up your account.

9. Grant any requested permissions. macOS may ask for access to specific folders or other features depending on how you use the app. You can allow or deny these now and adjust them any time in System Settings → Privacy & Security.

10. Pin Claude to your dock. Right-click the Claude icon in your dock while the app is open, go to Options, and select Keep in Dock. One click from anywhere, any time. This small step is what turns a one-time download into a daily habit.

What Are Those macOS Permission Prompts?

Mac users sometimes pause when permission dialogs appear. Here's the short version: macOS requires every app to ask your permission before touching sensitive areas of your computer. Claude Desktop is no exception.

Two prompts you're likely to see:

• "Claude wants access to your Documents folder" — This appears when you try to open or reference a file stored there. You can allow it for that folder specifically or choose a different location.
• "Claude wants to use the microphone" — This only comes up if you use a voice input feature. You can deny it without losing any core functionality.

To review what you've granted at any time, go to System Settings → Privacy & Security and scroll through the categories. Claude will appear anywhere it has access.

What to Do Once You're Signed In

Don't overthink the first session. Pick something real from your week.

Paste in the executive summary of a vendor proposal and ask Claude to flag anything your team should look at more closely. Or draft a short member-facing communication about a rate change and see how the first version reads.

These low-stakes tests give you a feel for where Claude adds speed and where you'll want a human to review before anything goes out. That instinct—knowing what to hand off and what to keep close—is the skill that matters most as you build this into your workflow.

What's the Risk?

Claude Desktop is a low-risk install for most credit union executives, with a few things worth keeping in mind. The app itself doesn't require any sensitive data to function. For everyday tasks—drafting internal communications, reviewing vendor summaries, preparing board materials from non-confidential sources—the risk profile is low and you can move freely.

The area that warrants more caution is data connectivity. Claude Desktop can be configured through MCP integrations to read files on your local machine. Before you point it at anything containing member NPI, account data, or other regulated information, talk to your compliance team. Whether that use is appropriate depends on your data governance policy, your vendor agreements with Anthropic, and applicable state and federal regulations. Consult your compliance team before connecting Claude Desktop to any member-identifiable or sensitive data. On the accuracy side, Claude is capable and fast, but it can be confidently wrong. Anything going to your board, your members, or your examiners needs a human read before it goes out.

This post was drafted with AI assistance and reviewed by a human at CU 2.0. AI makes mistakes; verify any specific claim before acting on it.

Frequently Asked Questions

This post was drafted with AI assistance and reviewed by a human at CU 2.0. AI makes mistakes; verify any specific claim before acting on it.

The post How to Install Claude Desktop on a Mac: A Step-by-Step Guide for Credit Union Executives appeared first on CU 2.0.

Meet Storyfi – the NACUSO New CUSO of the Year for 2026.  

When I saw the press release I had to get Nick Olexa, CEO of Storyfi, a Michigan based CUSO owned by six Michigan credit unions.

What Storyfi is doing is important – it’s harnessing video to help credit unions tell their stories and it’s working within credit union budgets.

That matters because, nowadays,  video is the globe’s communication tool of choice.  Whether it’s a 7 second video on Facebook or a 3 hour video on Netflix, video genuinely is how the world tells and communicates its stories and, frankly, credit unions had been lagging at this.

An upshot is that, in the main, credit union marketing has been blah.

Storyfi’s purpose is to reverse that reality.  To really get credit union stories told well.

Understand: every credit union has a story.  So does every CUSO.  Most of the stories revolve around solving problems and, often, working through adversities.  So these are good, compelling, human stories.  And in video they shine.

Since its 2023 founding, Storyfi has worked with around 140 organizations to tell their stories.

On the show, Olexa tells Storyfi’s origin story, the kinds of videos it creates for customers, and why all this really matters in thinking of the credit union future.

The post CU 2.0 Podcast Episode 413 Making Video Magic with Storyfi, the New CUSO of the Year appeared first on CU 2.0.

What Happens When Your GPT Doesn't Know Your Credit Union?

You built a Custom GPT. You gave it instructions. You asked it to help draft a member response about your overdraft policy—and it handed you something polished, competent, and completely generic. Something that could've come from any bank in the country.

That's the gap a knowledge file closes. Upload the right document, ask the right question, and your GPT stops guessing and starts quoting your actual policy.

What Is a Knowledge File in a Custom GPT?

When you build a Custom GPT inside ChatGPT, you can upload documents directly into the GPT's configuration. Those documents become a private reference library. When someone on your team asks the GPT a question, it searches those files and pulls relevant context before it answers.

Think of it like handing a new employee your policy manual, your rate sheet, and your member handbook on day one—except the GPT actually reads all of it before responding.

A knowledge file doesn't replace the model's general intelligence. It gives the model something specific to work with—your words, your policies, your voice.

This is called Retrieval-Augmented Generation, or RAG. The GPT retrieves relevant passages from your uploaded documents, then uses those passages to shape a more accurate, more specific answer. You don't need to understand the technical plumbing. You just need to know what it does—and what trips it up.

What File Types Can You Upload to a Custom GPT?

As of 2025, ChatGPT's Custom GPT builder accepts a range of common file types. The ones most useful for credit unions include:

• PDF — your best option for policy documents, member disclosures, loan guidelines, and anything that already exists as a formatted document
• Word (.docx) — drafts, procedures, internal memos
• Plain text (.txt) — clean and simple, no formatting overhead
• Markdown (.md) — useful if your team works in tools like Notion
• PowerPoint (.pptx) — works, but text-heavy slides pull better than visual ones
• Spreadsheets (.csv, .xlsx) — usable for rate tables or product comparison data, though the GPT handles narrative prose more reliably than rows of numbers

Always verify current file-type support at platform.openai.com—this list changes as OpenAI updates the product.

Each file can be up to 512 MB, and you can upload up to 20 files per Custom GPT (verify current limits at openai.com, as these have shifted over time). In practice, clean and focused files outperform large, sprawling ones. A tightly written 10-page lending FAQ will outperform a 90-page all-hands policy dump every time.

How Do I Add a Knowledge File to a Custom GPT?

You'll need a ChatGPT Plus, Team, or Enterprise account to build Custom GPTs—verify current plan requirements and pricing at openai.com.

1. Open the GPT builder. Go to chatgpt.com and click your profile icon in the top-right corner. Select "My GPTs," then click "Create a GPT" or open an existing one to edit it.

2. Switch to the Configure tab. The builder has two tabs: Create (a conversational setup tool) and Configure (where the real controls live). Click Configure.

3. Scroll to the Knowledge section. Partway down the Configure page, you'll see a section labeled "Knowledge." Click "Upload files."

4. Select your file. Choose a PDF, Word doc, or other supported file from your computer. Give it a moment to upload and process. It will appear as a thumbnail below the upload button.

5. Enable "Code Interpreter & Data Analysis" if needed. If you're uploading spreadsheets or want the GPT to do any math with your data, check that toggle. For most policy documents and narrative text, you can leave it off.

6. Save and test immediately. Click "Update" (or "Create" for a new GPT), then open a conversation and ask a specific question that your document should answer. Don't assume it works—confirm it.

Before and After: What a Knowledge File Actually Changes

Say your credit union has a specific policy: first-time homebuyers who qualify for your Community Champions mortgage product get a 0.25% rate reduction if they complete a homebuyer education course.

Without a knowledge file, you ask your GPT: "What should I tell a first-time homebuyer about our rate discounts?"

It says something like: "Many lenders offer rate discounts for first-time homebuyers. You may want to mention any applicable programs, such as down payment assistance or rate reductions tied to education requirements."

Accurate? Sure. Useful to your loan officer? Not really.

With your lending FAQ uploaded, you ask the same question:

The GPT says: "First-time homebuyers who qualify for the Community Champions mortgage product can receive a 0.25% rate reduction by completing an approved homebuyer education course. Confirm they meet the income and property eligibility criteria first."

Same model. Same question. Entirely different usefulness.

The Mistake Most People Make

Here's what goes wrong more often than not: someone uploads a file and walks away expecting magic. They ask the GPT a question, get a mediocre answer, and conclude the knowledge file "didn't work."

Usually, the file worked fine. The question was too vague.

The GPT retrieves content based on how closely your question matches the language in the document. If your policy doc calls it a "rate adjustment" and you ask about "discounts," the retrieval may miss. If you ask something broad like "tell me about our mortgage products," the GPT has no anchor—it doesn't know which part of a 40-page document to pull from.

The fix is straightforward: ask specific questions that use the same terminology your document uses. Coach your member services team or loan ops team the same way. "What is the income limit for the Community Champions product?" will pull a cleaner answer than "what are our mortgage options?"

A second common mistake: uploading scanned PDFs that are images, not text. The GPT cannot read a picture of a document. Run scanned files through OCR software first, or copy the key content into a clean Word doc before uploading.

Why This Matters for Your Credit Union

Your policies, products, and rates are not generic. Your members deserve answers that reflect your actual terms—not a best-guess composite of what the industry typically does.

A well-configured Custom GPT with a solid knowledge file is one of the most practical AI moves your team can make right now. Your member services team gets answers grounded in your actual disclosures. Your loan ops team can check guidelines without digging through shared drives. Your marketing team can write product descriptions that match what you actually offer.

It's not complicated. It takes the right file, the right questions, and about twenty minutes of setup.

This post was drafted with AI assistance and reviewed by a human at CU 2.0. AI makes mistakes; verify any specific claim before acting on it.

What's the risk?

The biggest risk here is data exposure. When you upload a file to a Custom GPT, that content is processed and stored by OpenAI on their infrastructure. For internal training documents or general product FAQs, that's a manageable risk most credit unions can live with. For documents containing member NPI, nonpublic financial data, or sensitive regulatory correspondence—talk to your compliance team before uploading anything. Review OpenAI's data handling and retention policies for your specific plan tier (Plus, Team, and Enterprise have different terms), and consult your information security officer if you're unsure where a document falls.

The accuracy risk is lower than many executives assume, but it's real. The GPT can misquote or partially retrieve a passage, especially from long or poorly structured documents. Never let a knowledge-file-powered GPT answer a member directly without a human review step in the workflow. For internal use—loan officers checking guidelines, your board prep team summarizing policy options, marketing drafting product copy for human review—the risk profile is low and the practical value is high. Go build it. Just don't upload your SAR log in the process.

Frequently Asked Questions

This post was drafted with AI assistance and reviewed by a human at CU 2.0. AI makes mistakes; verify any specific claim before acting on it.

The post How to Add a Knowledge File to a Custom GPT (And Actually Make It Work) appeared first on CU 2.0.

You built a custom GPT that drafts member-facing emails in your credit union's tone. Your loan ops team would save an hour a day using it. So why is it still sitting on your personal OpenAI account?

Sharing a custom GPT is easier than most executives expect. There are a few things worth getting right before you roll it out—but none of them require an IT project.

Personal GPTs vs. Team GPTs: What's the Difference?

When you build a custom GPT in ChatGPT, it lives on your personal account by default. Only you can see it, use it, and edit it. That works fine for personal productivity. It doesn't scale to a team.

A ChatGPT Team workspace (as of 2024—verify current pricing and plan details at openai.com) changes that. Inside a Team workspace, you can publish a custom GPT to everyone in the workspace. Your member services team, your loan ops team, your marketing coordinator—they all get access to the same GPT, with the same instructions, the same tone guidelines, and the same uploaded reference documents.

The distinction matters for three practical reasons:

• Personal GPTs are tied to one account. If that person leaves, the GPT goes with them.
• Team GPTs live in a shared workspace. The organization controls access, not the individual.
• Team workspaces don't use your conversations to train OpenAI's models by default—which matters if your team is discussing anything member-adjacent.

Consult your compliance team before uploading any member data or nonpublic personal information into any AI tool, even a private workspace.

How Do I Share a Custom GPT in a ChatGPT Team Workspace?

Once your Team workspace is set up, the publishing process takes about five minutes.

1. Open your GPT editor. Go to chatgpt.com, click your profile icon in the top right, and select "My GPTs." Find the GPT you want to share and click the pencil icon to edit it.
2. Go to the "Configure" tab. This is where your GPT's name, description, instructions, and uploaded files live. Make sure everything is finalized—your team will see whatever is here.
3. Set the access level. Scroll to the publishing settings. You'll see options like "Only me," "Anyone with a link," or "Everyone in [your workspace name]." Select the workspace option.
4. Save and publish. Click "Update" or "Save." The GPT is now visible to everyone in your Team workspace inside the GPT library.
5. Point your team to it. Team members can find shared GPTs by clicking "Explore GPTs" or browsing the workspace library. Send a quick Slack message or email with a direct link when you first launch—don't assume people will find it on their own.

Publishing a GPT to your Team workspace takes about five minutes. The harder work is the five minutes before that: making sure the instructions are actually ready for someone else to use.

Who Can Edit a Team GPT?

This is where a lot of credit unions trip up. By default, the person who created the GPT is the only one who can edit its instructions. If your CTO built it and then went on vacation, no one else can update the system prompt while they're gone.

A few things to know here. Workspace owners and admins have broader access controls, but editing a specific GPT's instructions typically still requires the original creator's account—or for the creator to transfer ownership. This area of the product evolves quickly, so verify with OpenAI's current documentation before you assume.

The practical fix: designate at least two people at your credit union who have the technical access to edit each GPT you deploy. Treat it like a shared credential in your password vault. And keep a copy of every GPT's system prompt in a shared Google Doc or SharePoint folder. That document is your source of truth regardless of what happens inside the tool.

How to Version Your GPT Instructions

AI tools don't come with version history the way Google Docs does. If you edit your GPT's instructions and something breaks, there's no "undo" button that shows you what changed.

A simple versioning habit solves this. Keep a running document—a basic table works fine—with columns for version number, date changed, what changed, and who approved. Before you touch the live instructions, paste the current version into the doc and increment the version number.

For anything that touches member-facing language or regulatory topics, have your compliance officer review the new version before you publish.

Treat your GPT instructions like a policy document: date-stamped, stored outside the tool, and reviewed before any major update goes live.

This isn't bureaucracy for its own sake. If an examiner ever asks how you're controlling AI-generated content, a version log shows intentional governance—not a wild west rollout.

What Do Team Members Actually See?

When a team member opens the GPT library, they'll see the GPT's name, description, and the chat interface. They do not see your system prompt or uploaded files—those stay hidden behind the scenes.

What they can do:

• Start conversations and get responses shaped by your instructions
• Upload their own files within a conversation (if you've enabled that capability)
• Reference any knowledge documents you've built into the GPT's configuration

What they cannot do, unless they're the creator or an admin with edit rights:

• See or edit the system prompt
• Add or remove uploaded knowledge files
• Change the GPT's name or description

This means you can publish one GPT for drafting board memos, another for member complaint responses, and a third for comparing vendor proposals—and your team just picks the right tool for the job without needing to understand what's running underneath.

What This Looks Like in Practice

Say your marketing team needs to produce consistent member newsletter copy. You build a custom GPT with your brand voice guide uploaded, a system prompt that enforces tone and required compliance disclaimers, and a clear description of when to use it. You publish it to the Team workspace. Now anyone on the marketing team—even a new hire—can draft newsletter copy that sounds like your credit union.

The compliance officer reviews the GPT instructions quarterly. You keep a version log in SharePoint. If OpenAI updates something that changes the GPT's behavior, you have the original instructions ready to restore from.

That's a real, manageable AI workflow. No IT project required.

What's the Risk?

For most internal use cases—marketing copy, member FAQ drafts, internal training materials—the risk profile of a ChatGPT Team workspace is relatively low. Conversations aren't used to train OpenAI's models by default, which is a meaningful step. But "low" isn't "zero." You're still sending data to a third-party cloud service, and your data governance policy and vendor agreements should reflect that. If your credit union hasn't formally assessed OpenAI as a vendor, that step belongs on your compliance team's list before you roll this out broadly.

The bigger risk is operational, not regulatory. Without a designated backup editor and an external version log, one person's departure or one accidental edit can break a GPT your team has come to rely on. If you're using a GPT in any workflow that touches member-facing language, lending decisions, or complaint handling, apply the same change-control discipline you'd use for any policy document. For pure internal productivity work—drafting agendas, summarizing meeting notes, structuring board presentations—you can move faster. Just keep the version log.

This post was drafted with AI assistance and reviewed by a human at CU 2.0. AI makes mistakes; verify any specific claim before acting on it.

Frequently Asked Questions

This post was drafted with AI assistance and reviewed by a human at CU 2.0. AI makes mistakes; verify any specific claim before acting on it.

The post How to Share a Custom GPT With Your Credit Union Team appeared first on CU 2.0.

Your Team Is Re-Explaining the Same Context Every Day, Isn't It?

Your loan ops manager opens Claude, types "you are a credit union compliance expert," pastes in your lending policy, and then finally asks the actual question. Every single session. Your member services team does the same thing with your fee schedule. Your marketing coordinator re-explains your brand voice from scratch every Monday morning.

That's not an AI problem. That's a setup problem. Claude Projects solves it—and you can have one running for your team in under an hour.

What Is a Claude Project?

A Claude Project is a persistent workspace inside Claude where you store instructions, upload knowledge documents, and keep conversation history. Your team never has to re-explain who you are, what your policies say, or what tone to use.

Think of it as a version of Claude tuned specifically for a task at your credit union. You build it once. Your team uses it over and over. The context is already there waiting.

A Claude Project is the difference between hiring a temp every Monday and working with a sharp colleague who already knows your policies, your tone, and your members.

If you've read our post on building custom GPTs for credit unions, the concept maps directly. Both tools let you embed instructions and knowledge so your team starts every conversation ahead of the curve. Claude Projects lives inside Claude (Anthropic's platform); custom GPTs live inside ChatGPT (OpenAI's platform). The workflow is nearly identical. Pick the tool your team already uses—or run both.

Why Does This Matter for Credit Unions Specifically?

Credit unions carry more context than most organizations. You have state-specific regulations. You have a unique field of membership. You have board-approved policies your examiners expect staff to know cold. You have a member-first tone that generic AI responses don't naturally produce.

Every time a staff member opens a blank AI session, all of that context lives only in their head—or gets copy-pasted from a shared doc that may or may not be current.

A Claude Project bakes that context in. Your lending team gets a project that already knows your HELOC guidelines. Your marketing team gets one that already speaks your brand voice. Your exec team gets one that already knows how to frame things for a board audience.

How Do I Set Up a Claude Project for My Credit Union?

You'll need a Claude account with Projects access. As of 2025, Projects are available on Claude's paid plans (Pro and Team)—verify current plan details and pricing at anthropic.com before you start.

Here's the setup from start to finish:

1. Create a new project. From the Claude home screen, click "Projects" in the left sidebar, then "New Project." Give it a clear name—"Lending Policy Assistant" or "Board Memo Writer"—so your team knows exactly what it's for.

2. Write your project instructions. This is the most important step. Claude calls this the "Project Instructions" field. Write it like you're onboarding a sharp new employee who knows nothing about your credit union yet. Cover who you are, who your members are, what tone to use, and what this project should help with. A few sentences won't cut it—aim for a solid paragraph or two. For a lending project, something like: "You are a lending assistant for [Credit Union Name], a federally chartered credit union serving [field of membership]. You help loan officers draft member communications, summarize applications, and check responses against our written lending policies. Always use plain language. Always flag anything that may need compliance review before it goes to a member."

3. Upload your knowledge documents. Click "Add Content" inside the project. Upload the documents Claude should reference—your lending policy, fee schedule, brand voice guide, member FAQ. Claude reads PDFs, Word docs, and plain text files. Keep documents current; an outdated policy doc is worse than no doc at all. Consult your compliance team before uploading anything that contains member PII or sensitive exam materials.

4. Test it before you share it. Ask a few questions your team actually asks. See if Claude pulls from your documents correctly and whether the tone is right. Tweak your instructions if the answers are off. Most projects need one or two rounds of refinement before they're ready for the full team.

5. Share it with your team. On Claude's Team plan, you can share a project with other users in your organization. They'll see it in their sidebar and can open new conversations inside it. Each conversation is separate, but the instructions and knowledge documents are shared across all of them.

Claude Projects vs. Custom GPTs: Which Should You Use?

Both are solid tools. Here's the honest comparison as of 2025:

• Claude Projects tends to handle longer documents well and often produces more nuanced prose—useful for drafting board memos or analyzing vendor proposals.
• Custom GPTs (in ChatGPT) offer more third-party integrations and sit inside OpenAI's broader ecosystem.
• Team sharing works similarly on both platforms, though the interface differs.
• Pricing varies—verify current plans for both Anthropic and OpenAI before committing your team to either.

If your team already uses Claude day-to-day, start with Projects. If they're on ChatGPT, start with custom GPTs. The concepts transfer almost one-to-one, so time spent learning one makes the other easy.

Four Starting Points Worth Building First

Not sure what to build? These four projects tend to deliver immediate value for credit union teams:

Examiner prep assistant. Upload your recent exam findings, your policies, and your CAMELS context. Use it to draft responses to examiner questions or stress-test your policy language before the visit. Have your compliance team review any output before it goes anywhere official.

Board report writer. Feed it your board report template and a few past reports. Use it to turn raw data into clean executive summaries your board will actually read.

Member communication drafts. Upload your brand voice guide and product descriptions. Use it to draft member-facing emails, letters, and FAQs in your voice—not generic AI voice.

Vendor proposal summarizer. When you're comparing core conversion vendors or digital banking platforms, drop in the proposals and ask for a structured comparison against your stated priorities.

None of these replace human judgment. Your team should always review AI-generated output before it reaches a member, a board member, or an examiner. But they can dramatically reduce the time it takes to get to a solid first draft.

What Should You Keep Out of Your Projects?

Avoid uploading documents that contain member account data, Social Security numbers, or other personally identifiable information. Claude processes data through Anthropic's servers, and while Anthropic has enterprise data-handling commitments, your member data belongs in systems built for it—not in an AI workspace. When in doubt, consult your compliance and information security teams before uploading anything sensitive.

What's the Risk?

The risk profile here is low-to-moderate, and where it lands depends almost entirely on what you put inside the project. If you're uploading brand voice guides, policy summaries written for staff training, and product descriptions, you're in low-risk territory. Go build. The workflow is no riskier than a shared Google Doc.

The risk rises if you start uploading documents that contain member NPI—account numbers, SSNs, loan details. Claude is not a core system or a document management platform with your credit union's data security controls around it. Before uploading anything that could touch member data, talk to your compliance officer and your information security team. On the regulatory side, no examiner has flagged credit union teams for using AI drafting tools on internal workflows, but the space is moving fast—consult your compliance team if you plan to use project outputs in any official exam response or member-facing regulatory disclosure. For internal productivity and first-draft work, the risk is low. Go.

This post was drafted with AI assistance and reviewed by a human at CU 2.0. AI makes mistakes; verify any specific claim before acting on it.

Frequently Asked Questions

This post was drafted with AI assistance and reviewed by a human at CU 2.0. AI makes mistakes; verify any specific claim before acting on it.

The post How to Set Up Claude Projects for Your Credit Union Team appeared first on CU 2.0.

Are Your Claude Prompts Actually Working for You?

You've opened Claude. You've typed something in. The output was… fine. Generic. A little flat. Not quite what you needed.

The tool isn't the problem. The prompt is.

Most executives type a question the way they'd Google something—short, vague, with no context. Claude is not a search engine. Think of it as a very capable new hire who knows a great deal but has no idea what your credit union does, who your members are, or what format your board expects. Give it that context, and the output changes completely.

Here's how to write prompts in Claude that actually produce something useful.

What Makes a Prompt "Good"?

A good prompt does three things: it tells Claude who you are, what you need, and how you want it delivered.

That's it. You don't need special syntax or a certification. You just need to stop assuming Claude already knows your world.

A prompt isn't a command. It's a briefing. The more context you give, the closer that first draft lands to what you actually need.

Think about how you'd onboard a sharp intern on day one. You wouldn't say "write something about our HELOC product." You'd say: "We're a $400M community credit union in Ohio. Our members skew 45 to 65. We want to explain our HELOC rate promotion in plain language for an email to existing mortgage holders. Keep it under 150 words and skip the jargon." That's a prompt.

The Three Moves That Improve Every Prompt

Move 1: Give Claude a Role and Context

Start by telling Claude what kind of expert to act as—and who you are. This single change improves output quality more than almost anything else.

Instead of: "Summarize this vendor contract."

Try: "You are a credit union operations consultant reviewing vendor contracts. I'm a COO at a $250M credit union. Summarize the key risks and renewal terms in plain English. Flag anything unusual."

Claude doesn't know you're in financial services unless you say so. It doesn't know your members, your core system, your exam cycle, or your board's communication preferences. Tell it.

Move 2: Specify the Format You Need

Claude will pick a default format if you don't ask for one—and it might not be the format you want. Need bullet points for a quick board update? Say so. Need a table comparing two vendors? Say so. Need a three-paragraph email in plain language for a member who just turned 70? Say that too.

Format instructions that work well in credit union contexts:

• "Write this as a short email, no more than 150 words."
• "Give me a table with three columns: feature, vendor A, vendor B."
• "Format this as talking points for a 10-minute presentation."
• "Write this at an 8th-grade reading level."
• "Use headers and bullet points. This is for an internal memo."

Move 3: Iterate—Don't Start Over

This is the move most executives miss. When the first output isn't right, the instinct is to close the window and start fresh. Don't.

Tell Claude what to fix, in the same conversation. Each follow-up builds on what came before. That's how the tool works best.

Try follow-ups like:

• "Good start. Make the tone warmer—this goes to members, not examiners."
• "The second paragraph is too long. Cut it in half."
• "Add a sentence explaining why this change benefits members."
• "Rewrite the subject line. The current one sounds like spam."

You're not starting over. You're editing with a very fast collaborator.

5 Copy-Paste Prompt Starters for Common Credit Union Tasks

These are ready to use. Drop in your specifics wherever you see brackets.

1. Member-facing email "You are a credit union communications specialist. Write a short member email (under 150 words) from [Credit Union Name] announcing [topic]. Our members skew [age range] and we want a warm, plain-language tone. Avoid financial jargon. End with a clear call to action."

2. Board memo summary "I'm a credit union CEO preparing a board update. Summarize the following information into a concise executive summary (3–4 short paragraphs). Use plain language. Structure it as: what happened, why it matters, what we're doing about it. Here is the source material: [paste text]."

3. Vendor proposal comparison "Act as a credit union operations consultant. I'm evaluating two vendors for [product/service]. Here are their proposals. Create a comparison table covering: pricing, contract length, key features, support terms, and any red flags. Note anything a credit union examiner might ask about."

4. Policy or procedure plain-language rewrite "Rewrite the following internal policy in plain English for a credit union staff member with no compliance background. Keep it under [X] words. Use short sentences and active voice. Here is the original: [paste policy text]."

5. Exam prep talking points "You are a credit union compliance consultant. Based on the following information, generate talking points I could use when presenting [topic] to examiners. Anticipate the top three questions they're likely to ask and suggest brief, factual responses. Here is the background: [paste text]."

Why This Matters More Than You Think

Many executives spend 10 to 15 minutes correcting AI output that a better prompt would have gotten right the first time. Multiply that across your lending team, your marketing coordinator, your member services leads—and across a full week—and the gap between okay prompts and good prompts becomes a real productivity story.

You don't need to become a prompt engineer. You just need to brief Claude the way you'd brief a smart person who's new to your credit union: explain the institution, the audience, and the desired outcome. Then use the conversation to refine, not restart.

That's the whole skill. Your entire team can learn it in an afternoon.

What's the Risk?

The good news: prompting itself carries very low risk. You're not connecting Claude to your core, your loan origination system, or your member database. You're typing instructions and reading responses. For marketing copy, board memo drafts, policy rewrites, and exam prep talking points, the risk profile is low—go ahead and experiment.

The place to be careful is what you paste in. If your prompt includes member names, account numbers, Social Security numbers, or any nonpublic personal information, you've moved into a different risk category. Anthropic offers enterprise privacy agreements (verify current terms at anthropic.com), but your credit union's data governance policy should drive what's appropriate—not the vendor's marketing materials. Talk to your compliance and IT teams before pasting anything you wouldn't want in a third-party system. For the five prompt starters above, used with placeholder or anonymized content, the risk is minimal. For workflows that touch real member data, get compliance in the room first.

This post was drafted with AI assistance and reviewed by a human at CU 2.0. AI makes mistakes; verify any specific claim before acting on it.

Frequently Asked Questions

This post was drafted with AI assistance and reviewed by a human at CU 2.0. AI makes mistakes; verify any specific claim before acting on it.

The post How to Write Better Prompts in Claude: A Practical Guide for Credit Union Executives appeared first on CU 2.0.

Title this show everything you need to know about fraud today but have been afraid to ask.

Start with this: Fraud is big and getting bigger.

That is the worrisome message delivered by Rene Perez, a fraud expert with Jack Henry, and seconded by Stephanie Ziegler, director of financial investigations at SELCO, a $2.8 billion credit union headquartered in Oregon.

Perez is on the show to discuss Jack Henry’s new Financial Crimes Defender tool and Ziegler and SELCO are early adopters and she tells why.

Much of the show is devoted to discussing trends in financial institution fraud – and the criminals just keep getting slicker and more sophisticated.

Credit unions can – and are – successfully fighting back but a key realization is accepting that the fraud tactics keep changing.

Nowadays, for instance, a huge problem, said Perez, is that criminal organizations are seeking to implant crooked employees in financial institutions and fintechs.  Perez and Ziegler discuss their organizations’ defenses against this.

Another mushrooming trend: account takeovers are epidemic.

The bottomline reality: fraud is an immense problem for all credit unions – and what you don’t know can indeed hurt you.

The post CU 2.0 Podcast Episode 412 Jack Henry and SELCO on Fraud Trends You Need to Know Today appeared first on CU 2.0.

Have You Heard Your Team Talking About Claude Desktop?

Maybe a colleague mentioned it at a conference. Maybe you've been using Claude in a browser tab and wondered if there's a better way. There is.

Claude Desktop is a free Windows app that puts Anthropic's Claude AI directly on your computer. No browser required, no tab-switching, no logging in every time. Download the installer from claude.ai/download, run it, sign in with a free Anthropic account, and you're done in under ten minutes. The rest of this post walks through each step in detail—and covers the Windows quirks most likely to trip up credit union teams.

What Is Claude Desktop, and Why Does It Beat the Browser?

Claude Desktop is the official Windows (and Mac) app from Anthropic, the company that makes Claude. It's free to download and runs on your machine rather than in a browser tab.

The practical difference is file access. Claude Desktop can connect to folders on your computer, so you can point it at a folder of vendor contracts, board documents, or exam prep notes and ask questions about what's actually in those files. The browser version doesn't offer that out of the box.

Claude Desktop turns a powerful general-purpose AI into something that can actually see your documents. That's the shift from "interesting tool" to "part of how I work."

For a credit union executive, that opens up real use cases fast: summarizing a lengthy vendor proposal before a board meeting, drafting a response to a member complaint, or pulling themes from a stack of policy documents before an examiner visit.

What You'll Need Before You Start

• A Windows 10 or Windows 11 PC (verify current system requirements at claude.ai)
• A free Anthropic account—you'll create one during install if you don't have one
• About ten minutes

That's it. You don't need IT to do this. That said, if your credit union manages its devices centrally and locks down software installs, check with your IT lead before starting. Some managed devices require admin rights to install new applications.

How to Install Claude Desktop on Windows

Step 1: Download the installer

1. Open your browser and go to claude.ai/download. Verify the exact URL is current when you visit—Anthropic occasionally updates download paths.
2. Click Download for Windows. The installer file (it ends in .exe) will land in your Downloads folder.

Step 2: Run the installer

1. Open your Downloads folder and find the file. It will be named something like Claude-Setup-x64.exe.
2. Double-click it to run the installer.
3. If Windows shows a security prompt asking whether you want to allow this app to make changes to your device, click Yes. This is normal for any new application.
4. The installer runs quietly in the background. Claude Desktop opens on its own when it finishes—no long setup wizard to click through.

Step 3: Sign in to your Anthropic account

1. When Claude Desktop opens for the first time, you'll see a sign-in screen.
2. Click Sign in and enter your Anthropic account email and password. If you don't have an account yet, click Create account and follow the prompts. It takes about two minutes.
3. Once signed in, you'll land on the main chat screen. Claude is ready to use.

Step 4: Pin Claude Desktop to your taskbar (optional but recommended)

1. Find the Claude icon in your taskbar or Start menu.
2. Right-click it and select Pin to taskbar.

This makes it a one-click open every morning. It's a small habit, but proximity drives adoption. The teams that use AI most consistently are the ones that don't have to go looking for it.

How to Connect a Folder to Claude Desktop

This is where Claude Desktop earns its keep for credit union work. Connecting a folder lets Claude read the files inside it and answer your questions based on what's actually in those documents—not just its general training.

A practical example: a folder of vendor RFP responses you're comparing, or a set of member-facing policy documents your marketing team is updating.

Here's how to connect a folder (as of 2025—verify steps in the current app):

1. Open Claude Desktop and click the settings or integrations icon. Depending on your version, it may look like a plug or a gear.
2. Look for a section labeled MCP or Filesystem integration. Anthropic has been expanding these features, so the exact label may vary. Check Anthropic's current documentation at docs.anthropic.com for the most up-to-date steps.
3. Select the folder you want Claude to access.
4. Return to your chat and ask Claude something like: "Summarize the key terms across the vendor contracts in my Documents folder."

A few things worth knowing: Claude reads what's in the folder but does not modify or delete your files. Your documents stay on your machine and are not sent to a public database. That said, consult your compliance team before pointing Claude at any folder containing member PII or data covered by your information security policy.

Common Windows Gotchas

The app won't install. Your PC may require admin rights. Right-click the installer file and choose Run as administrator. On a managed device, contact your IT team.

Claude Desktop won't open after install. Check your system tray—the small icons near the clock in the bottom-right corner of your screen. Claude Desktop sometimes minimizes there rather than opening a full window.

The sign-in page is blank or won't load. This is common on credit union networks that filter outbound web traffic. Your network's firewall or content filter may be blocking claude.ai. Ask your IT team to whitelist claude.ai and any related Anthropic endpoints. It's a routine firewall configuration change.

You see a message about an outdated version. Claude Desktop updates automatically in the background, but if it prompts you to update manually, download the latest installer from claude.ai and run it. The process is identical to the original install.

Why Credit Union Teams Stick With the Desktop App

Many executives try AI in a browser once or twice, then forget about it. The browser experience works, but it has friction—open a new tab, log in, start a fresh chat with no context.

Claude Desktop removes that friction. When it's pinned to your taskbar and connected to your working folders, it stops being a novelty. Your loan ops team, your marketing lead, your executive assistant—they can reach for it the same way they reach for Word or Outlook. Many credit union teams report they simply use AI more often and more thoughtfully once it lives on the desktop like any other business tool.

One thing to set expectations on: Claude Desktop does not connect directly to your core. It can work with files you export from your core—reports, spreadsheets, summaries saved to a local folder—but there is no native integration with core banking platforms.

What's the Risk?

Claude Desktop's risk profile is relatively low for general productivity use, but there are real considerations for credit unions before you roll it out broadly.

The main one is data handling. When you connect a folder, Claude reads those files locally. Anthropic's current data practices indicate that conversations may be used to improve their models unless you're on a paid plan with data controls—verify the current privacy settings at claude.ai and confirm what applies to your account tier before connecting any folder that contains member NPI, loan data, or anything covered by your information security program or GLBA obligations. Consult your compliance team first. For marketing copy drafts, internal training materials, or publicly available documents, the risk is low and you can move freely. For anything touching member data, get compliance sign-off.

On the examiner front, NCUA hasn't issued specific guidance on desktop AI tools as of this writing, but examiners are increasingly asking about AI governance. If you're deploying Claude Desktop across a team, document the use case, the data involved, and any guardrails you've put in place. That documentation will serve you well in any future exam conversation. Vendor concentration risk is worth noting too: Anthropic is a single AI provider, so factor that into your third-party vendor management thinking.

This post was drafted with AI assistance and reviewed by a human at CU 2.0. AI makes mistakes; verify any specific claim before acting on it.

Frequently Asked Questions

This post was drafted with AI assistance and reviewed by a human at CU 2.0. AI makes mistakes; verify any specific claim before acting on it.

The post How to Install Claude Desktop on a Windows PC (Step-by-Step for Credit Union Teams) appeared first on CU 2.0.

Can a banker successfully transition into a senior role in a credit union?  

Can a credit union deploy a financial wellness initiative that also works as a competitive edge in marketing?

Kevin Miller, CEO of $5.7 billion Travis Credit Union in Northern California, is on the show today and he is guilty of that banking background but he has logged 4+ years as Travis CEO.

And he also strongly believes that not only are financial wellness initiatives a key part of a credit union’s doing good in its communities, they also are a competitive edge against both mega banks and community banks.

As for Miller’s past work experience, he logged 4+ years at regional bank TCF as EVP of retail banking, 2 years at Santander, 4 years at BMO, and 6 years at Bank of America. A stint at a fintech rounded out his background.

In the show he explains why he jumped into a credit union and why hiring bankers is good business for a credit union.

The post CU 2.0 Podcast Episode 411 Travis Credit Union CEO Kevin Miller on Why Bankers Make Good CU CEOs appeared first on CU 2.0.