April 19, 2025
In a world increasingly influenced by AI, how do we preserve and celebrate human thought?
Human Authored in an AI World
We're currently working on a project for one of our nonprofit clients called Human Authored. The goal is to distinguish and honor human creativity in an increasingly AI world. The premise is simple: people don't want to be fooled into reading works auto-generated from a remix of previously published works, courtesy of our robot friends.
I'm all in when it comes to transparency, in every area of life, and it makes sense to me to provide a certification program to help readers easily distinguish between human creativity versus AI. I'm genuinely excited we're able to collaborate on this project.
For nonprofits especially, trust is the currency. Donors, funders, and the communities you serve need to know your voice is yours. A certification like this isn't just a nice-to-have. It's a trust signal.
One thing that's been interesting technically. Once my client decided to open up the certification process to non-members, the question became: How do we verify the author is actually human? Their existing vetting process handles this for members, but for anonymous non-members, something new was needed. Enter AI-powered Identity Verification. Is this ironic?
A Case Study: How AI Enhances (or Doesn't) My Day-to-Day Work
Here's how AI showed up, or didn't, as I worked on this feature:
Client Use Case and Needs
We spent time understanding my client's specific use case and needs. That means conversations, smart questions, and leveraging our long-standing relationship. I didn't use AI for this part. But if I were new to this client, I would use AI's research capabilities to more quickly and deeply understand my client and their business niche.
This is the most important step and the one most often shortchanged. Good technology decisions start with a clear understanding of the problem, not a search for a solution.
Surveying the Marketplace
I vetted many Identity Verification tools on the market and created a matrix to help compare them. I used AI's research capabilities to help me frame the questions and gather data, but I found that I needed to verify sources myself, and in many cases schedule sales calls to gather deeper information. The annoying part of this is when organizations force me to go through the sales and marketing funnel of doom to get simple answers. I tried using AI to avoid this and couldn't. My AI would point me to their online pages but not answer such questions for the vendor. Sigh!
Beyond features and pricing, we were also evaluating data handling, compliance posture, and how each vendor manages client data. For nonprofits working with members, donors, or vulnerable populations, that due diligence is not optional.
Product Decision
After refining the matrix with client feedback, we decided on Veriff.com. When we were down to our final 3 choices my client said "go with your gut, Lisa". That... Well, AIs don't have guts do they?
Experienced judgment, shaped by years of decisions and outcomes, is exactly what a strategic advisor brings. It doesn't show up in a spreadsheet.
Signing Up
We signed up the client and started a subscription. It took some time as Veriff is an international company and didn't like the corporate credit card. AI tools won't do such things for you. You have to work through setting up the account and subscription yourself.
UI/UX
We're integrating Veriff into multiple user onboarding flows. This requires UI/UX thought and business requirements conversations about timing. Do we verify users before or after purchase? What if someone changes their name? Do we need the court documentation and to re-verify? What balance between usability and security should be taken? How much fraud detection? This required writing up options for my client and meeting to make decisions. My AI edited my drafts, but wasn't my final editor. That was my business partner.
These aren't just UX questions. They're policy questions. Getting them wrong has legal, ethical, and operational consequences. This is where strategic oversight pays for itself.
Technical Architecture
This step I think about as "directing my developers". It consists of deeply reading their documentation and planning the steps. Do I understand everything about this product and their API? Am I at the point where I can specify exactly how I want this implementation to be coded? I didn't think about incorporating AI into this process. But this is what I do for a living and I'm super good at it. Perhaps a less experienced architect could use AI planning capabilities to help them through this task.
A well-specified architecture protects the client's investment. Ambiguity here is expensive. It leads to rework, scope creep, and integrations that are painful to maintain.
Coding
The code-completion and suggestion AI built into the IDE can be mostly helpful, sometimes annoying. The same with code linters. Though I wouldn't characterize them as AI, we are implementing custom code linters for our company to keep our coding standards consistent across projects.
I use Vibe Coding. Let me share where I've landed on this more broadly, because I think it's worth a clear-eyed take.
Tools like Replit are genuinely useful for prototyping, small standalone apps, and building dashboards. If you need to get a quick visual in front of a client, sketch out a workflow, build a visual app to embed in a website, or test an idea before committing resources, Vibe Coding can be a solid accelerator. Use it for that.
Where I'd pump the brakes: if you're building anything with a backend and database, stop and think carefully about the data privacy implications. What data is flowing through this? Who can access it? Where is it stored? Is it covered by your privacy policy and your clients' expectations? These are not small questions.
Also, when business logic lives in AI-generated code that no one fully reviewed, accountability goes out the window. Who owns the decision when something goes wrong? Your developers need to understand the code they're shipping, especially when it governs anything important: access control, data validation, financial logic, workflow rules. "The AI wrote it" is not an explanation your client, your auditor, or your users will accept.
At Coat Rack, we have a clear line on this. We do not vibe code backends. Only frontends. Backend systems handle data logic, integrations, authentication, and often sensitive information. That code needs to be deliberate, reviewed, and fully understood by the humans maintaining it. Using AI-generated code for backend systems without that rigor is a risk we are not willing to take for our nonprofit clients.
User Testing
We've got robust backend automation for testing, but for User Testing and Acceptance Testing of the UI, I rely on humans. I've experimented in the past with front-end testing automation and found it pretty lacking, and expensive. Why? Because automation never does things or makes mistakes like humans do. It doesn't get confused like humans do. One of the most desirable pieces of feedback I look for from a human is "this worked but it was kind of weird?" That makes me sit up in my seat and work to understand why immediately.
For nonprofits especially, your users span a wide range of tech comfort levels, devices, and contexts. Human testers catch things automated tools simply cannot. That's not a cost. That's quality assurance.
I do think there's potential for AI to improve in this space. Imagine if AIs could model the human natural propensity for error and weirdness?
Deployment
We already have CI/CD set up, with agile sprints and releases managed by Jira.
Bug Fixing
We're perfect and have zero bugs. Haha... just making sure you're still reading! We are implementing more operational automation around bugs, support tickets, and production notifications. My AI has been really helpful in researching and implementing upgraded tools like Zapier AI.
Client and Development Documentation
AI can really help with code documentation. Plus my AI writes good "How To" documentation and files them neatly in Google Drive for me. I love that. And I know I keep saying "my AI". If anyone is curious, I've been using Sintra.
Final Thoughts and Takeaways
I always keep in mind my client and their needs. What my client wants is to implement a product where we can celebrate human creativity while acknowledging that AI is here to stay. I feel like my workflow already reflects that balance. I use AI tools for research, automation ideas, and documentation support, and AI tools have become embedded into many aspects of my job. The research capabilities are scarily fast and good (but always have to be verified). The automation suggestions I've received from AI helpers alone can keep me busy for months implementing them. Vibe Coding has found a real place in my toolkit, for the right scenarios.
The through line here is judgment. AI accelerates the work you know how to do. It doesn't replace the wisdom to know what to do, when to do it, and what guardrails matter.
Ultimately, I value sitting and thinking, reading and absorbing, communicating and refining direction. That's not fast. It can't be automated. It's intentional and truly, humanly, creative.
I'm curious to know what others think. Please reach out to me if you'd like. Have a great day!
