AI can be a great tool for gathering requirements, creating specifications, and turning project discussions into actionable tickets. But when AI-generated content moves from one person or process to another without careful review, complexity can quickly snowball.
What starts as a simple request can become an over-engineered set of requirements filled with unnecessary acceptance criteria, jargon, and scope. The solution isn’t to stop using AI; rather, it’s to use it thoughtfully, review what it produces, and remember that someone else – that you probably have some respect for – has to understand what you send them.
AI plays a great game of telephone
Recently, our dev team ran headfirst into this problem. We realized some workflows had accidentally turned into a high-tech game of telephone. An initial meeting was recorded and summarized by AI. Then, someone used those notes to generate a spec document. Finally, a project manager fed that document into another AI to churn out Jira tickets.
By the time the developers were reading the tickets, the AI had hallucinated requirements big time.
The “graceful” red flag
One of our developers, Lon, got bit by this pretty badly. During a group review of some upcoming work, the team noticed a bizarre pattern: the exact same pie-in-the-sky sentences were copy-pasted across multiple tickets. Everything supposedly needed to be “reversible,” “auditable,” and – the real giveaway – “graceful.”
When the team finally stopped to ask what the original requester actually wanted, they realized none of those requirements were real. It was just filler text the AI added to sound professional. But before they caught the mistake, Lon had already started over-engineering the project, building out five new database tables when the simple, original request only needed one.
Handing a burden to our teammates
This situation highlights an interesting side effect of our new tools: the empathy gap. Pasting a raw AI output might save you ten minutes of typing, but it shifts a massive burden onto the person receiving your “notes.” You’re easily handing them two to three times the amount of text they would have received otherwise, forcing them to sift through the fluff just to figure out what they need to do. It’s an unfair trade. You save yourself a little time, but you make your coworker’s life much harder.
Scope creep in disguise
Sometimes, the AI is genuinely trying to help by inserting what it sees as “best practices,” like suddenly adding heavy security requirements even though you’re working with a basic internal tool. Those ideas aren’t always bad, but they can cause massive scope creep if you aren’t paying attention. You have to stop and ask yourself questions like: Do we actually want to build all of this right now? Is this level of security overkill for the situation?
Don’t use a hammer on every dang thing
Ultimately, this isn’t an AI problem; it’s a how-we-use-it problem. We aren’t going to ditch AI – it’s clearly incredibly useful for saving time and organizing our thoughts. But we have to be mindful about how we wield this seemingly miraculous tool.
There’s a great scene in the 2016 sci-fi film Arrival that perfectly captures this idea. Dr. Banks and Colonel Weber are realizing the profound consequences of how differently each nation is communicating with the aliens.
Dr. Banks: Let’s say that I taught them Chess instead of English. Every conversation would be a game. Every idea expressed through opposition, victory, defeat. You see the problem? If all I ever gave you was a hammer…
Colonel Weber: Everything’s a nail.
When we use AI to polish our communications, we can’t just blindly swing the hammer. It’s on you to verify that any automated tool is actually doing what you intend. Read what the AI writes. Edit out the fluff. Confirm you agree with its assumptions and recommendations.
And, just as importantly, think about the person on the other end of your message and how their day will be affected by what you’re about to send.
Need help with your FileMaker system?
Whether you’re trying to untangle an over-engineered FileMaker solution or just want to make your systems work better for the actual humans using them, we can help. Schedule time with us so we can chat and have a real conversation.
This piece represents a collaboration between the human authors and AI technologies, which assisted in both drafting and refinement. The authors maintain full responsibility for the final content.