Strategic Data Products · Chicago

The AI Pod Jumpstart

Give us a problem. Let AI solve it.

Two weeks. One problem. One agent that ships.

Your hardest work doesn’t live in a process document. It lives in the head of the person who has been doing it for nine years — the exceptions, the workarounds, the four systems they check before they hit send. You can’t automate that from a flowchart. You have to sit next to them.

Give us a problem Book a Call

I build AI applications for businesses.

Twenty years building data systems for Oracle and InvescoProduction AI systems shipping todayAnthropic Solutions Partner

§ 01 The playbook

The playbook isn’t mine. It’s Uber’s.

In two months Uber ran sixteen of these sprints across sixteen business functions — finance, legal, operations, marketing, support. Publicly reported results:

15 hrs 30 min

Capital allocation across 150 cities

2 days 10 min

Financial pacing reports

2 weeks 50 min

Marketing web quality assurance

9,000 self-serve

Manual support workflows

Source: Uber engineering leadership, public remarks, 2026.

Uber could run sixteen of them because it had thirty AI-fluent engineers to spare. Most companies don’t — the people who understand your systems are already booked solid. That is the gap I fill. You get the pod for two weeks. What it builds stays with you.

§ 02 A two-week sprint

Ten working days

First

Find the opportunity.

I look across the operation for where AI actually pays. The test is not the department — it is the shape of the work: repetitive, high-volume, judgment-heavy, spread across systems nobody wants to open at once.

Days 1–2

Shadow.

I sit with your expert and watch the real job — every click, every exception, every “well, usually, except when…”

Day 3

Prioritize.

I rank what I saw by volume, repetition, dollar impact, and whether the data is actually reachable. You approve the target.

Days 4–5

Build.

I build next to the person who does the work. They see it wrong, and I fix it the same day.

Days 6–9

Validate.

I put it in front of others doing the same job. Does it generalize past one person’s habits? Does it make the job better or just different?

Day 10

Ship.

It runs in production. You own it.

10 working days · one problem or task · one expert from your team, part-time.

§ 03 The deliverable

What you keep

A working agent, in production. Not a proof of concept. Something a real person uses on Monday.

The workflow, written down. The job as it is actually performed, exceptions included.

A skills library you own. Prompts, tools and connections in your systems, your account, your data.

Governance that survives an audit. Human approval gates, role-based access, an audit trail, and a plan for sensitive data.

The next three candidates, ranked. So the second pod is faster than the first.

§ 04 Where it works

FinanceLegalOperationsMarketingSupportHRProcurement

The test isn’t the department. It’s the shape of the work: repetitive, high-volume, judgment-heavy, and spread across three or four systems nobody wants to open at once. If a good employee is doing it by hand every week, it’s a candidate.

§ 05 Receipts

Two I have already built.

Both in production today. Neither company hired anyone to run them.

Village of Bellwood, Illinois

An AI chief of staff for the Mayor’s office, running across six village departments. FOIA-compliant records handling, role-based access, tokenized personal information, and a human approval gate on every outbound action. Built on Next.js, Supabase, and pgvector.

Brown Sugar Bakery · 75th Street, Chicago

175,000+ customer emails triaged, classified and drafted by agent, with an 88% cost reduction by routing routine mail to a cheaper model and hard mail to a stronger one.

§ 06 Start one

Pick the task everybody complains about.

A short call. We look at the shape of the work, name a target, and start the ten days. If it isn’t a fit for your business, you will know that too.

The engagement

$1,500. The complete POD assessment. One expert from your team, part-time.

That is the whole price of the assessment. You get the spec and the decision. If you want the agent built after that, we scope the build separately — and if it is not right for your business, you have lost two weeks and $1,500, not a year.