Spend a week in San Francisco and you'll hear that every company needs an "AI strategy," a Chief AI Officer, and a $40k/month model budget. Spend a week in Cedar Rapids, Des Moines, or Madison and you'll hear something different: "We just need this one thing to stop eating our Tuesdays."
Both things are real. But only one of them is true for the businesses we work with. The Midwest doesn't need an AI revolution. It needs an AI Tuesday — small, durable wins that compound quietly while the coasts argue about AGI.
The problem with most "AI advice"
The default advice — "pick a model, write a prompt, deploy an agent" — assumes infinite engineering capacity and a high tolerance for things that almost work. That's a fine assumption inside a Series B startup with twelve engineers. It is a terrible assumption inside a 40-person logistics company, a regional credit union, or a family-owned manufacturer running on a 12-year-old ERP.
What those teams actually need is the boring part: a clear map of where AI helps, where it doesn't, and what to build first so the rest of the year gets easier instead of harder.
The stack, as we actually deploy it
Below is the stack we keep returning to across our embedded engagements. It is opinionated, deliberately small, and chosen to survive the moment when the founder asks "wait, what is this costing us per month?"
- A capable model, accessed simply. One frontier model behind one gateway. Not seven providers, not a vector database before you need one. The point is leverage, not architecture cosplay.
- An internal tool layer. Most of the wins aren't a chatbot. They're a quiet form, an internal dashboard, a Slack command — something that lives where the work already happens and makes the next click obvious.
- Workflow plumbing. The unglamorous middle: a job queue, a few webhooks, structured logging. This is the difference between "demo" and "Tuesday."
- A human in the loop, by design. Not as a safety net bolted on later. As a first-class part of the flow, with a clear UI for approving, correcting, and teaching the system.
Where it actually pays off
The highest-ROI work we ship rarely looks impressive in a demo. It looks like:
- A 6-hour weekly intake process compressed into 20 minutes, with a human reviewer left in charge of the calls that actually matter.
- A quoting workflow that drafts the first pass from a messy email, so the estimator spends their day pricing, not parsing.
- A back-office tool that turns "ask Marcia, she knows" into a searchable, durable institutional memory — without making Marcia feel replaced.
None of this is a moonshot. All of it ships in weeks, not quarters, and pays for itself before the next budget cycle.
The Midwest advantage
Coastal AI culture is optimized for narrative. Midwest business culture is optimized for outcomes. That mismatch is usually framed as a disadvantage — "they're behind." We think it's the opposite. The companies we work with don't need to unlearn five years of AI theater. They get to walk in fresh, skip the hype, and aim straight at the boring, compounding work.
That's the whole bet behind Space Pirate Labs, and behind this newsletter: pragmatic AI, built embedded with your team, evaluated on whether Tuesday got easier — not whether it tweeted well.
Written in Iowa City. New issues go out when there is something worth sending.