Choose the bets
Identify where AI can materially change revenue, cost, capital, risk, customer outcomes, or organizational capacity. Turn those opportunities into an accountable investment portfolio.
Michael Upchurch · AI Strategy & Value Creation Executive
I turn enterprise AI strategy into adoption, and adoption into measurable value.
I help enterprises decide where AI can create real economic value, build the operating system around those bets, and stay with the work until the result shows up in adoption, revenue, cost, customer outcomes, risk, or capacity.
Twenty-five years across financial services, enterprise technology, advanced analytics, and AI — from building ML portfolios inside large enterprises to owning the P&L of an analytics software company.
Working proof · The Eddington Factory
I built a governed AI factory and run it on live software.
The Factory is a governed decision-and-verification system for AI-written software: specialized agents build the work, independent graders test it against a frozen contract, a person approves release, and production verifies the result.
It gives me somewhere to test the same questions enterprises now face when capability outruns the operating model around it.
01IdeateIntent & acceptance
→
02BuildSpecialized agents
→
03ReviewIndependent challenge
→
04ApproveHuman authority
→
05DeployControlled release
→
06VerifyProduction evidence
Define the operating system. Separate responsibilities. Make authority explicit. Prove the controls can fail. Learn from production.
Enter the Factory$250M+Executive-approved annual value across an enterprise ML portfolio
72+Production ML use cases operating across nine businesses
$4.7MAnnual recurring revenue built from three commercial ML products
$11B → $22BLending portfolio growth during a contracting market
Executive mandate
The work gets interesting when AI has to become a business system.
The strongest fit is an organization that needs to make AI economically real — whether that means choosing the first serious bets or turning a growing portfolio into adopted, accountable systems.
Build the operating system
Align business, technology, finance, and risk around ownership, decision rights, funding, delivery standards, production criteria, and measures of success.
Create adoption and value
Turn technical capability into products people use, propositions customers buy, and operating changes the organization can absorb. Measure whether the promised value appears.
Manage the consequences
Match control to consequence, make evidence part of delivery, and anticipate the customer, workforce, risk, and operating effects that arrive as AI takes on more consequential work.
Selected experience
Strategy, operating models, products, and P&L.
Full experience →Capital One
Enterprise value at scale
Directed machine-learning strategy across nine businesses and built a portfolio of more than 72 production use cases. Business owners reported and executive-approved more than $250 million in annual value.
01
Fuzzy Logix
Products, revenue, and P&L
Co-founded and scaled an advanced-analytics software company, launched three commercial ML products, and built $4.7 million in annual recurring revenue while owning the commercial system around the technology.
02
Bank of America
Strategy translated into growth
Designed and executed mortgage strategy that helped expand the lending portfolio from $11 billion to $22 billion over four years while the broader market contracted.
03
Domino Data Lab
Industry strategy as commercial advantage
Led financial-services and insurance strategy for an enterprise AI platform, connecting model-validation modernization, platform adoption, governance, and sophisticated regulated buyers.
04
Eddington · Working lab for enterprise AI
Ideas get better when they have somewhere to fail.
Eddington is where I build systems, develop operating methods, and publish what the work is teaching me. It keeps my executive thinking close to current technical reality and creates practical methods that can travel into an enterprise.
01AIRO · Operating model
AI adoption operating system
A practical method for turning AI activity into one accountable portfolio, a repeatable path to production, measurable returns, and deliberate use of the capacity AI creates.
4 stages · 24 toolsExplore the method → 02Provance · Validation methodValidation built into delivery
A method for traditional, AI/ML, and agentic systems that moves requirements and independent challenge earlier and creates useful evidence as the work happens.
7 checkpoints · 12 toolsExplore the method → 03Field notesEssays
Writing on AI, work, value, governance, and what changes when productive capacity rises faster than the operating systems around it.
Current thinkingRead the notebook →Publishing every two weeksFifteen SecondsFirst issue · The moment AI put knowledge work on a clock.See what is next →
Two paths · One body of work
Bring the work inside your organization — or apply Eddington to a focused problem.
Senior executive mandate
I’m looking for the right senior mandate to turn AI strategy into an operating portfolio, adopted systems, and measurable business value.
Selected Eddington engagement
For a narrower problem, Eddington provides tested methods for AI operating models, validation, product strategy, commercialization, and agentic controls.