Field Notes

The only AI worth paying for solves a job you can name.

Paul Nykamp

Written by

Paul Nykamp

Founder of FieldKit

"We should be using AI somehow" is a sentence I hear in almost every first conversation. It's an honest feeling — the tools are genuinely good now, and it seems like everyone else is doing something with them. But as a starting point it's backwards, and it's how businesses end up paying monthly for tools nobody opened after week two.

"The goal isn't to add AI. It's to take one job that eats an hour a day and make it someone else's problem."

The nameable-job test

Before evaluating any tool, finish this sentence out loud: "I want it to ___." Not "help with marketing." Not "make us more efficient." Specific: draft the first reply to a new inquiry in our voice. Turn the shop's shared notes into something a new hire can search. Chase every unpaid invoice at day 14 and day 30 without anyone remembering.

If you can't name the job, the honest answer is that you're not ready to buy anything — and that's fine. It usually means the process around the job isn't settled yet, and no tool survives an unsettled process.

The three-month test

Here's how I evaluate whether something will stick:

Where AI genuinely earns its keep

In the businesses I work with, the wins are rarely glamorous. Drafting repetitive replies so a human only edits. Summarizing a week of messages into a Monday morning brief. Turning a binder of procedures into something searchable. First-draft quotes from a job checklist. Each one names a job, does it end to end, and gets used without a mandate.

And just as often, the right answer is a $0 one: a written procedure, a spreadsheet, or a feature inside software you already pay for. My job is to tell you which — honestly.

If you want a second opinion before you sign up for something, book a free 30-minute call with Paul.