Dan Haiem is the founder and CEO of AppMakers USA, helping business leaders design, build and scale apps that deliver real-world impact.
Using AI tools isn't new in app development. At AppMakers USA, we've used automation for years, and more recently, we've used AI to speed up the boring parts, like drafting boilerplate, generating test cases and exploring implementation options.
What’s new is the question I’m hearing more often: “Was this actually built by a person, or did you have AI generate the code?” As a small company that takes real pride in our team’s skills, that can feel a bit like an accusation, not because we’re hiding anything, but because it shows what customers are buying now: accountability.
When I read about the U.K. Society of Authors launching a “Human Authored” logo, I related immediately. It’s the same instinct. People want to know there’s a real person behind the work, and they want a simple way to tell. That's because when customers can’t see the origin at a glance, trust becomes a feature.
Why This Question Is Spreading
My guess is that generative AI lowered the cost of producing “finished-looking” output. I can see it everywhere, including copy, designs and code. Personally, that creates oversupply and confusion. Buyers can’t easily tell what's original, what's templated and what's machine-produced, so they default to skepticism.
And skepticism changes buying behavior. Premium work becomes harder to defend, reviews get harsher and disputes get messier because everyone is arguing from assumptions. Across industries, the response tends to mature in stages: starting at informal signals (reputation), then labels, then receipts.
What A Creation Receipt Means For Software
Most provenance talk focuses on media. Software is headed the same way because the risk is similar. If you can’t tell how something was produced, you can’t judge reliability.
To make this practical, I like to break it down into a simple product layer you can actually build. Think of it as a provenance layer—a lightweight set of disclosures, logs and proofs that travel with your work.
A creation receipt is the output of that layer. It answers three questions in plain language: what was delivered, how it was produced and reviewed and who stands behind it.
When you can do this well, it can make the origin legible when it matters.
The Provenance Layer
The provenance layer has three parts: disclosure, trace and verification.
1. Disclosure: This is what the customer sees. Start with a simple statement that matches how clients think, such as “human-built with AI assistance,” then define what “assistance” includes in your process. Put it where customers already look for confidence, like proposals, handoff docs, release notes and security summaries.
2. Trace: This is what your team can prove. It's the minimum audit trail that shows ownership, so these are tickets tied to changes, meaningful pull requests, testing evidence and approvals. The goal is to provide clarity.
If a dispute happens, you want facts, not vibes. A small reason-code vocabulary for AI use (e.g., boilerplates, tests, documentation or summarization) keeps disclosure consistent across teams.
3. Verification: This last part is what makes the receipt believable. On the media side, standards like C2PA Content Credentials use cryptographic signing so provenance can be validated. On the software side, the parallel is build and artifact provenance like signed releases, software bills of materials and attestations that connect the source to the build through to deployment.
You don’t need to become a standards expert to adopt the principle of "provenance that's portable and signed is harder to counterfeit than provenance that's merely claimed."
To keep this usable, make provenance legible. The fastest way to make it fail is to turn it into a compliance theater. So, keep it skimmable, available on demand and generated by the workflow. In a nutshell, provide one clean “build receipt” page they can read in under a minute, with optional detail if they want it.
When we built Echo Journal, an AI-based voice journaling app, we were explicit about what is user-created (entries) versus system-generated (insights). That same boundary thinking applies to code delivery to ensure that clients know that real engineers owned and reviewed the work.
What I'm Trying To Say Is…
In a world where output is cheap, trust is expensive. The teams that win won’t argue about authenticity in comment threads. They’ll build provenance into the product and delivery process, so customers can stop guessing.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?