Where to Begin Your AI Implementation: Build a Task-Level Inventory First

Where do you start with planned AI implementation? One practical starting point: document what your positions actually do.

Your office’s job descriptions probably describe roles in general terms—“prepares monthly journal entries,” “assists with rate case support,” “reconciles subsidiary ledgers to the general ledger.” That’s accurate. It isn’t specific enough to tell you what’s exposed to automation and what isn’t.

Build a Task-Level Inventory, Not a Job Description Rewrite

For each position, list the actual tasks performed in a given month, and roughly how many hours each one takes. Not “accounts payable”—instead, “keys invoices into the ERP, averaging 6 hours a week.” Not “financial reporting”—instead, “pulls consumption data from the billing system, builds the variance schedule in Excel, and drafts the narrative explanation for board packets, averaging 10 hours a month.”

That level of detail is the point. A job description tells you what a position covers. A task inventory tells you where the hours actually go—and which of those hours are the kind AI can touch today.

Sort the Tasks Into Three Categories

Once you have that list, sort the tasks into three categories.

The Inventory Is the Deliverable

Once you have hours attached to each task, you can see where the automation conversation should start—not with the newest tool on the market, but with the tasks eating the most hours in category one.

That’s the starting point. Next week: what to do with that inventory once you have it—how to redesign the role around categories two and three, and what that means for training.