Absence AI

Reporting a student absence sounds small until you think about how many people and systems a single call can touch.

One parent message can mean notifying teachers, updating attendance, flagging the nurse, and making sure the details are clear enough that no one has to follow up later to clean up confusion.

That’s exactly the kind of workflow I think software should be good at: repetitive, detail-sensitive, and important enough that getting it wrong creates more work for everyone involved.

Absence AI was my way of exploring what it looks like to remove that administrative drag without asking parents or staff to learn a new system.

Instead of pushing people into a rigid form, the experience is driven by just saying what’s going on.

Designing for a calmer intake experience

If a parent needs to report an absence, speed matters, but so does emotional tone.

This isn’t a moment where someone wants to navigate a complex dashboard or decode school-specific language just to communicate something simple.

That’s what made voice feel like the right interface for this concept.

A parent can speak naturally, mention one child or multiple children, explain whether the absence is partial or full-day, and move through the process without having to translate their situation into the structure of a form first.

Because the interaction is imagined as part of an authenticated parent-facing system, the experience can start from context the school already has. It can know which children belong to that parent, use that as a constraint, and reduce the amount of repetitive back-and-forth that usually makes these workflows feel slower than they need to be.

But the important product decision wasn’t just “use AI” or “use voice.”

It was making sure the conversation still behaves like a responsible workflow.

So the experience is designed to confirm details before anything’s submitted. The goal is to preserve the ease of a conversation while still creating the confidence of a well-run process.

If a product feels easy but produces messy records, it isn’t actually reducing friction. It’s just moving the burden downstream.

Turning a conversation into useful school actions

What interested me most in this project was the translation layer between natural language and structured action.

Parents aren’t thinking in terms of data models. They’re thinking, “Emma will be out tomorrow,” or “Luca needs to leave after lunch,” or “they both have symptoms and I want the school to know.”

The system’s job is to take that loose, human input and turn it into something operationally clean.

That means identifying which student is affected, whether the absence is full-day or partial-day, which periods are impacted, and when health-related details should be routed into nurse notes rather than buried in a generic reason field.

It also means handling confirmation as part of the product, not as an afterthought. Repeating the details back before submission helps the parent feel heard, but it also protects the workflow’s integrity.

I think that’s where a lot of AI product thinking either gets shallow or gets interesting.

The interesting part isn’t the model doing something impressive in a vacuum. The interesting part is designing an experience where the output becomes trustworthy enough to plug into a real process.

Keeping the intelligence in the background

I didn’t want this concept to feel like “AI for schools.”

I wanted it to feel like a calmer front door to an existing operational task.

The product value isn’t that a parent gets to talk to a model. The value’s that a routine but time-consuming workflow can happen quickly, clearly, and with less manual handling from staff.

To me, that’s the more useful framing for AI in products generally.

It shouldn’t be there to add novelty. It should be there to absorb friction, create structure where people naturally speak in ambiguity, and let the actual user experience feel simpler than the process behind it really is.

That’s what Absence AI was meant to explore: not a flashy demo, but a more thoughtful way to turn conversation into action.

Feel free to visit the repo if you want to run it locally.