Why your meeting recordings, your grant pipeline, and your photo library are the same kind of work.

A stack of meeting recordings nobody has done anything with. A grant pipeline nobody applies to. A photo library nobody can find anything in. Founders describe these as three different problems. They’re describing one.

This is the shape of small business operations more often than people expect. Three things that look unrelated turn out to be the same shape of work underneath, and seeing the shape changes what you build first.

The short version

Every version of this kind of work has the same three parts. Something comes in (a recording, an organizational profile, a stack of photos). Repetitive shaping happens in the middle (transcribing, scanning, tagging). Something else has to come out (a writeup, a drafted application, a tagged record). The middle is almost always work a person on staff has been doing or should be doing, with judgment scattered through it but not concentrated.

Once you can see those three parts inside a piece of work, you can build the version where AI absorbs the middle and the person on staff keeps the judgment that mattered. The three problems above are all this shape. Most of the problems in your operation that look like “we should be doing this but we don’t have time” are this shape too. I wrote a broader piece called the pattern I keep finding. This one walks through three specific examples.

One: the meeting recording pile

The input: recordings of meetings the business runs. Team standups, sales calls, board sessions, client intakes, customer interviews. The format varies. The length varies. The recording exists.

The middle: somebody used to listen back and take notes, pull the action items, draft the recap, schedule the follow-up. Three hours of work per recording. Eventually they stopped doing it for everything except the most critical meeting.

The artifact: a recap that captures the actual conversation, an action-item list with owners and dates, a draft of any follow-up that should go out, a tag in your CRM about what happened with the person on the other side.

What gets absorbed: the listening, the note-taking, the action-item extraction, the draft of the recap, the first cut of any follow-up message.

What stays human: the editorial call on what to forward and to whom, whether a sensitive section gets included, the relationship context for follow-ups that need to feel like the human relationship the meeting actually was. Nothing goes out without a person signing off.

What gets shipped: the team goes from “we should write up that meeting” to “a draft recap is in your inbox by the time the meeting ends, you spend ten minutes editing and sending.” Recordings that used to pile up unprocessed produce artifacts the same day.

Where I built one of these: the $5K Method takes a 60-minute structured career interview as input and produces a 50-page personalized report as output. Different domain. Same shape.

Two: the grant pipeline

The input: the organization’s profile (mission, budget size, demographics served, geography, certifications, prior award history). Plus the public databases of available funders, RFPs, and pitch competitions.

The middle: somebody is supposed to scan funder databases every week, read each RFP carefully against eligibility, score the fit, draft applications that match the funder’s narrative, attach the required documents, track deadlines. The board agrees this should be happening. Nobody has time.

The artifact: a short list of well-matched opportunities each week, with a drafted application for each one, ready for board review and signoff.

What gets absorbed: the database scanning, the eligibility scoring, the first draft of the application narrative, the document attachment list, the deadline tracking.

What stays human: the strategic call about which opportunities to actually pursue, the relationship context for funders the board already knows, the final language in the application that needs to sound like a person and not a template. Nothing goes out without a board member’s signoff.

What gets shipped: the board goes from “we should be applying for grants” to “the system surfaces three to five matched opportunities each week with drafts attached, the board reviews on its monthly call, the right ones get submitted.” Money the org was eligible for, that nobody had time to chase, starts arriving.

Where I built one of these: On Purpose Oregon runs its grant pipeline this way, against a 16-person volunteer board with no dedicated grant staff.

Three: the photo library

The input: every photo the business has ever taken or commissioned. Product shots, lifestyle shots, event photos, founder portraits, work-in-progress documentation. Spread across phones, drives, Dropbox folders, archive hard drives.

The middle: somebody is the human search engine for the photo library. Find me the picture from the trade show last year. Find me the headshot of so-and-so on the team. Find me a photo of the product on a blue background. Every time. Manual scrolling, opening folders, asking around.

The artifact: a searchable library where anyone on the team can search “founder portraits on dark backgrounds” or “trade show 2025” or “products in lifestyle settings” and get back a curated set in seconds. The same library can feed an automated social workflow that posts a tagged image with caption-ready text in a couple of clicks.

What gets absorbed: the tagging, the visual description, the searching, the routing of the right asset to the right post.

What stays human: the editorial choice of which photo represents the brand in any given moment, the call about whether a sensitive image should ever go out, the final approval on every post that names the business publicly.

What gets shipped: the team goes from “where’s that picture” to “find me X” working as a real query against a real searchable library. The library also starts feeding social posts, lookbook builds, pricing assessments, anything visual the business does.

Where I built one of these: the first version is in production for NailFile, a stylist whose visual library was growing faster than her ability to use it.

The pattern across the three

All three have the same three parts. All three have a person on staff (or no person, because there was never time) doing the same shape of work over and over. All three have judgment concentrated at one or two decision points. All three produce something with the business’s name on it.

If your operation has any of these three shapes, you almost certainly have at least one more that’s the same pattern in a different domain. Member directories, sales call follow-ups, inspection photo workflows, vendor invoice processing, CRM enrichment, customer success Q&As, hiring intake. The list of operations with this kind of work hiding in them is longer than the list without.

What this isn’t

• It isn’t a claim that every operations problem is this shape. Some operations problems are about people, culture, strategy, or pricing, and this pattern does nothing for those. The pattern works for repetitive shaping work specifically.
• It isn’t a “build everything at once” strategy. Most engagements identify three to five pieces of work like this and ship the highest-leverage one first.
• It isn’t an API project. OpenAI or Anthropic APIs are some of the tools that get used, but the work is operations recognition, not engineering.
• It isn’t a content factory. These systems absorb work humans were already doing or should have been doing. They aren’t built to mass-produce content for its own sake.

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 with three pieces of work worth building, each with an investment range and a timeline. If you want to know which versions of this pattern are hiding in your operation, start a conversation.