Function-first AI solutions
Purpose-built tools that live inside one team's workflow and speak its language — not a general-purpose assistant hoping someone finds a use for it.
Point solutions, built to the shape of one business function.
Not a platform. Not a year-long program. We build focused AI solutions around a single workflow — mixing the right proportions of people, process, and tools — and we measure success the only way that counts: whether your team still uses it six months later.
One conversation. No deck, no discovery fee.
What we build
Small solutions, aimed at the work that actually hurts.
Most AI initiatives stall because they were scoped to a technology instead of a job. We start from a business function — quality, supply chain, finance ops, field service, HR — and build the smallest thing that removes real friction inside it.
Agents and copilots with guardrails
AI that drafts, checks, routes, and summarizes — with the human decision points left exactly where your risk tolerance says they belong.
The plumbing underneath
Document pipelines, extraction, and integration into the ERP, CRM, or PLM you already run. Most AI projects die here — so we start here.
Adoption you can see
Enablement, change support, and instrumentation so usage and value are visible after go-live — not just impressive on demo day.
The recipe
Three ingredients. Never the same proportions twice.
Every bourbon starts with a mash bill — corn, rye, barley. Shift the percentages a few points and you get a completely different whiskey from the same three grains. Business solutions behave exactly the same way. We weigh the same three ingredients on every engagement, and the blend is different every time.
People
Who does this work today, what they trust, and what they'll actually pick up on a busy Tuesday. Roles, incentives, and the change that has to land for any of it to matter.
Process
The workflow as it really runs — including the spreadsheets and workarounds nobody documented. We fix what's broken before we automate it, because automating a bad process just makes it faster.
Tools
The systems you already own, plus the smallest amount of new AI needed to close the gap. New technology is an ingredient, not the recipe.
A sample mash bill
Illustrative only — the point isn't the numbers, it's that the blend changes. Two companies with the same problem rarely need the same solution, and the difference is usually people and process, not the model.
How we work
Weeks, not quarters. Then a real decision.
A small batch is deliberately small. You should know whether this works — and whether your people will use it — long before anyone asks you to sign a transformation program.
Step 01
Discovery
A short, focused look at one function. We map the workflow, sit with the people doing it, and find where AI genuinely helps — and where it plainly doesn't.
Step 02
The recipe
A written plan: what we'd build, the blend of people, process, and tools it needs, what it costs, what it should return, and how we'll know if it worked.
Step 03
The batch
We build and pilot small, with real users and real data. Tight loops, visible progress, and something in your team's hands early enough to change our minds.
Step 04
Proof
Measured adoption and measured value, then an honest decision: scale it, adjust the recipe, or stop. We'll tell you when the answer is stop.
Already built
A few batches already pouring.
Each of these started as one function that was drowning in manual work, and ended as something small enough to hand to real users and see whether they came back.
Search and rescue
Wayfound
Runs a missing-person search from one live map. The command post opens the incident, draws grids over satellite imagery, and checks searchers in by QR code; every phone reports its position so cleared ground, clues, and team messages update as the search unfolds.
wayfound.smallbatch.solutions
Equestrian
Gate Call
Show bills run once in a print magazine and are easy to miss. Gate Call photographs each one, transcribes the dates, classes, fees, and contacts, and lets riders search by distance, date, and discipline — with the original scan kept alongside.
gatecall.smallbatch.solutions
Legal service of process
Civil Process
A process server photographs the day's paperwork; vision models pull out every service address. The app confirms the list, optimizes the route, hands off to turn-by-turn navigation, and keeps the stops in sync across devices on a short share code.
www.civilprocess.app
Severe weather intelligence
STORMR
Three free feeds on one dark map: NOAA MRMS rotation tracks, live NEXRAD reflectivity, and active NWS warning polygons — each layer reordered and faded independently. Not an official warning source; the NWS remains the source of truth.
stormr.smallbatch.solutions
County public safety
County Livestock Registry
A voluntary directory where residents register the animals they keep. When cattle or horses turn up loose, law enforcement searches an address and radius on the map, filters by animal type, and reaches the owner directly instead of guessing.
livestock-registry.smallbatch.solutions
Veterinary compliance
DispenseLog
Veterinary clinics still track controlled substances on hand-written sheets. DispenseLog keeps the same columns, adds a per-lot running balance that can't drift, converts the dose from mg to mL on its own, logs every edit immutably, and exports the report an inspector expects.
dispenselog.smallbatch.solutionsWhy small batch
Most AI projects don't fail on the model. They fail on adoption.
Small by design
Senior people, small teams, direct access. The person you talk to is the person building it — there's no layer between the conversation and the work.
Scoped to a function
We don't sell enterprise-wide transformation. One function, one workflow, one solution that earns the right to the next one.
Adoption is the deliverable
A pilot nobody opens is a failed project, however good the demo was. That's why the tagline isn't a slogan — it's the acceptance criteria.
Start a conversation
Have a function that's drowning in manual work?
Tell me what the workflow looks like and where it hurts. If there's a small batch worth building, I'll tell you what it would take. If there isn't, I'll tell you that too.
michael@smallbatch.solutions