Most of my working life has gone into helping businesses fix the things that slow them down, using whatever tools fit best. Lately that's meant a lot of AI, both to help me and to help them. A few of those stories are below, along with the small things I do the rest of the time.
Work stories
Card data
One fax from Alaska
One person spent 80% of their time pulling card data from FTP, email and a fax from Alaska into billing. Now they only handle the exceptions.
Read the storyShow less
Card data came in as spreadsheets over FTP, HTML tables in email and, yes, one fax from Alaska. One person pulled it all into Excel and ran many, many formulas to get it into billing, which took about 80% of their time. These were financial transactions, so every record had to be kept exactly as it arrived. I designed and built the replacement on Dataverse and a low-code automation platform. OCR and SharePoint's fax functionality handled the fax, and basic imports handled the rest. The system kept an untouched copy of every record, then consolidated the data and moved it through to billing automatically. Now that same person only handles the exceptions.
Payroll
The shared macro workbook
Payroll moved off a shared, macro-heavy workbook. Deductions got easier to add, and the treasury manager got the exchange-rate data he needed.
Read the storyShow less
Payroll ran on a very, very macro-heavy Excel workbook shared by several teams. As the company grew, people were editing the same things at once and the macros got confused. The company was also losing money on foreign exchange, with no way to track it. I designed the replacement on Dataverse and wrote the code myself. Every record was kept exactly as it came in, and so was every paystub, with the automation in between and reporting for each customer down to the employee. Adding a deduction like a phone stipend or a meal per diem used to mean resaving all those macros. Now it's added without touching them. The system also tracked exchange rates, which gave the treasury manager the data to fix the losses.
Retail
A truck full of clothes
Checking in a truck went from a day of paper for one person to a shared mobile app that settles overages on its own.
Read the storyShow less
A clothing retailer checked in every truck the same way: unload the boxes and compare them to an ERP printout that was supposed to match the driver's paperwork and the truck, and often didn't. On paper, nobody could tell what had already been checked, so either one person lost a whole day to it or everyone chipped in around their other work. I built the replacement on Dataverse. OCR read the driver's paperwork, and a blind-entry Power Platform app had staff record what was really in the boxes without seeing what was expected, so nobody could fudge a count. The really cool part: when a store got more than it should have, the system decided on its own whether to keep the extra or send it back, based on current inventory. Now one person or three could work the same truck without tripping over each other, and everyone could see what was left. As it rolled out across stores, each store or region adjusted the app to fit how it worked, down to who signs off: the manager at smaller stores, an assistant manager at bigger ones.
Invoices
Keeping it in Excel
Instead of moving an Excel expert to a database, I automated around her. Now every file she opens is already right.
Read the storyShow less
Vendor invoices all arrived as Excel files, and some vendors added extra columns as the months went by. Every line had to be traced to a general ledger account, given the right debit or credit, and reshaped for import into the ERP. The person who owned it was a master of Excel but didn't build automation, and moving her to a database would have meant changing how she worked and inventing a new way to handle exceptions. So I kept it in Excel. Power Automate picked up the invoice emails in Outlook and saved them to SharePoint, and Office Scripts I wrote with AI's help did the account lookups, turned columns into rows, and untangled each vendor's odd naming. Now when she opens a file, it's already right, and her job is managing the exceptions, not copying and pasting. A new GL account is one row in a config workbook. And because Office Scripts are well-documented code, she can use her own AI to troubleshoot them.
Currently
-
Espresso Working on good espresso and average latte art.
-
Bike From full-time bike commuting to just trying to keep up with the kids.
-
Lifting I pick up heavy things and put them down.
-
Pasta There's always room to improve, even on the simplest recipes.
-
Woodworking Sometimes all I make is sawdust, but it's good to create with my hands.
-
Guitar Irish pub songs, and slowly building up my technical ability.