The pattern I keep finding inside small business operations.

Across the systems I’ve shipped this year, the same shape of work keeps showing up. There’s something that goes in. Someone has to turn it into something else. Either someone on staff is buried doing it by hand, or nobody is doing it because there was never time.

This is the pattern I keep finding inside small business operations, and naming it has changed what I look for when I walk into a new engagement.

The short version

Every operation has work that hides in the same shape. Something comes in (a recording, a photo, a row of data, an applicant’s eligibility profile). Someone has to turn it into something else (a writeup, a searchable library, a public-facing page, a drafted application). The middle is almost always a person on staff doing the same repetitive work over and over with their judgment scattered through it, or it’s work that nobody is doing because there was never time.

AI Operations is the practice of finding that work, building the version where AI absorbs the repetitive middle, and leaving the judgment with the person who should have it. Once you can see the shape, you start finding it everywhere.

Five common versions

Most operations have at least two of these. Some have all five.

1. Recordings to writeups. A recording comes in (a podcast episode, a sales call, a board meeting, an interview, a voice memo). A written artifact has to come out (the show notes, the call recap, the meeting minutes, the personalized report). The middle is someone listening back, taking notes, and shaping the prose. The first version of this I built was the $5K Method, which turns a 60-minute career interview into a 50-page personalized report. The pattern applies to media businesses, sales operations, board secretaries, and any consultant who runs structured interviews as the input to deliverables they sell.
2. Photos to a searchable library. Visual assets come in (product photos, lifestyle photos, inspection photos, before-and-after shots, work portfolio images). A managed, searchable, postable library has to come out. The middle is someone manually tagging, scrolling, retitling, and copying assets across to social. I built the first version for NailFile. The same pattern applies to beauty practices (nails, hair, lashes, brows, tattoos), interior designers, photographers, real estate offices, and any business whose visual library sits unused because finding anything in it is a manual job.
3. Profile data to public-facing pages. Information about people or entities comes in (founder bios, member profiles, board rosters, vendor information, employee directories). A public-facing record has to come out, kept up to date by the people inside it instead of by the staff person who used to chase everyone for updates. The first one I shipped was the founder portal for Portland State Business Accelerator (PSBA), which lets each portfolio company update its own public profile in 30 seconds. The pattern applies to nonprofit member directories, B Corp certified-business listings, professional association rosters, and any directory where the people in it should maintain their own entries.
4. Eligibility data to drafted applications. An organization’s profile comes in (mission, budget size, demographics served, geography, certifications). A list of well-matched opportunities and a drafted application for each has to come out. The middle is the work nobody had time for: scanning grant databases weekly, reading RFP requirements, drafting attachments, tracking deadlines. The first version I shipped runs On Purpose Oregon’s grant pipeline against a 16-person volunteer board with no dedicated grant staff. The pattern applies to any small nonprofit, any family office without grants staff, any small business chasing public-sector contracts, and any volunteer-led membership organization eligible for funding nobody had time to apply for.
5. Milestones to public announcements. Internal milestones come in (a portfolio company hit a number, a member earned a credential, a new product launched, a partnership closed). External-facing publications have to come out (press releases, social posts, announcement emails, milestone-marker entries). The middle is somebody on staff manually crafting the same shape of announcement every time. I built the press release pipeline for PSBA, which takes milestone data in and produces a release the team reviews and ships. The pattern applies to investor relations, alumni offices, nonprofit communications, B Corp certification announcements, and any business with regular newsworthy events but no dedicated communications staff to publish them.

What this isn’t

A few things this pattern gets confused with.

• It isn’t generic workflow automation. Zapier, n8n, Make are some of the tools that build these systems once you’ve recognized what to build. The recognition is the hard part. Most workflow automation projects fail because they automated the wrong middle.
• It isn’t a content factory. These systems are designed to absorb work humans were already doing, or were trying to do and couldn’t. They aren’t built to mass-produce content for its own sake.
• It isn’t a chatbot. The artifacts these systems produce have a shape and an audience. They aren’t conversations.
• It isn’t replacement of the person who used to do this work. The pattern always has a human checkpoint before any artifact ships. The person who used to do the middle by hand now reviews and decides, instead of typing.

How to know if you’ve got one

Five signals tend to show up together when an operation is hiding this kind of work.

1. You can name a piece of work where the inputs are well-defined (a recording, a photo, a data row, a profile) and the output is well-defined (a writeup, a published page, an application, a post).
2. Someone on staff has been doing the middle of that work by hand and the volume has grown faster than the staff time available, or nobody is doing the work and everyone agrees it should be happening.
3. The middle is mostly repetitive shaping, with judgment concentrated at one or two decision points (which photo, which framing, which numbers to highlight) rather than spread evenly through the whole task.
4. The artifact has a person’s name or your business’s name on it. It needs to be right, which means a person signs off before it ships.
5. If you stopped doing this work entirely, the business would lose meaningful value. It isn’t optional, it’s just chronically incomplete.

If two or three of those are true for a piece of work in your operation, you have this kind of work hiding there.

Here’s what I do

The Scoping Engagement is the version of this question shaped for your operation. Two weeks of looking, then a written plan that names three pieces of work worth building the AI version of, each with an investment range and a timeline. If you want to know which versions of this pattern are hiding inside your operation, start a conversation.