AI projects that earned their keep.
The AI enablement work I actually use: automations at Chapman, a couple of tools I built for my own life, and two projects I’m still figuring out.
AI enablement in practice
Personal Productivity
-
LIVE
Digital Twin
I wanted to know whether AI could learn how I actually write, rather than making me sound like AI.
Copilot analyzed my Outlook and Teams history and helped me build a style guide from the way I communicate. That became the foundation for a custom GPT trained only on emails I wrote myself, with a little context about the things I care about: educational technology, higher ed, canoeing, gardening, and history.
Two tools, one voice.
-
LIVE
Morning Brief
Every morning, a single hand-drawn page tells me what I need to know: what’s on my calendar, what’s waiting in my inbox, what’s happening in Teams, what’s already handled, and how heavy the day looks.
It runs on a schedule. I don’t have to remember to ask.
That’s the point.
Community Projects
-
ADOPTED
Accessibility Guidelines for an Elite Paradragon Team
Our paradragon team needed formal accessibility guidelines. Nothing off the shelf really fit a mixed-ability paddling crew, so I used GPT to help with the structure, language, and supporting materials.
The team adopted the guidelines.
The interesting part wasn’t having AI write something for us. It was having a tool that could help turn a very specific problem into something usable by the whole team.
-
PUBLISHED
Coach’s Guide for Amputee Paddlers
Coaches working with amputee paddlers had almost nothing written specifically for that population.
So I wrote the guide.
It covers what can be different about coaching an amputee athlete on the water, what coaches should think about before the first practice, and the questions worth asking rather than assuming you already know the answers.
AI helped with the work, but the expertise and the experience came from the athletes and coaches who actually live it.
Workplace Skills & Automations
-
IN USE
Vendor Scorecard Automation
Evaluating a vendor renewal used to mean digging through contract PDFs, old email threads, and Teams conversations, then pulling it all together by hand.
Now Claude researches all three, scores the vendor against a weighted rubric, drafts a scorecard with a renewal recommendation, flags risks, and puts time on my calendar ahead of the renewal with links to what it found.
I show up already briefed.
That’s a pretty good use of AI.
-
IN USE
Concur Expense Reconciliation
Monthly PCard reconciliation in Concur is mostly UI busywork: classify the charge, write the justification, attach the receipt, repeat.
Claude runs through the report while I watch and review what it does.
What used to take about an hour now takes a few minutes of review.
I am very happy to never again spend an hour classifying conference parking.
-
IN USE
Staff Coaching Files
I wanted a better way to prepare for staff coaching conversations without relying entirely on whatever I happened to remember about the last few months.
Claude looks through my own email and Teams messages for our interactions with that person and organizes a draft coaching file: communication patterns, strengths, possible areas for growth, and things worth discussing.
It is built from actual interactions rather than my memory alone, but it is still a draft. The judgment is mine, and so is the conversation with the person.
Development Projects
-
NEARING
COMPLETIONFINCH
Faculty rebuild the same professional record from scratch at every reporting milestone: annual review, tenure review, promotion.
The information already exists. It’s just scattered across half a dozen systems.
FINCH keeps one continuously updated record instead, pulling from institutional systems, public sources, and personal documents, then producing whatever format the moment calls for.
The goal is simple: nobody should have to re-enter information the university already has.
-
EARLY
DEVELOPMENTYaso
My health data lives across five-plus sources, and none of them can see the whole picture.
Yaso pulls the data into one place and looks for patterns within a specific window of interest rather than throwing everything at everything and hoping something interesting falls out.
I can ask it a question in plain language instead of staring at five different charts trying to figure out whether anything is connected.
It’s still early. I’m interested in whether the idea works before I decide what it should become.
More in progress
Tools I work with
Claude and Cowork run most of the automations above. Claude Code and the BMAD Method are helping structure the two projects still in development. Gemini, NotebookLM, and Microsoft Copilot are part of the broader toolkit.
Still curious? If AI in higher ed is what brought you here, I’d love to hear what you’re trying to figure out.