> how i think about the hard part

How I work.

How I take AI from "we should try this" to something a regulated shop can actually stand behind: adoption that sticks, governance that isn't theater, and launches that don't slip. Same rule everywhere - make the hard part feel possible, and leave the place better than I found it.

AI work drifts. I keep yours pointed at one outcome, not the newest shiny tool. That's Stay in Orbit: every prompt earns its place, or it gets cut.

// start here

Find your Moon.

Most AI projects fail before a prompt gets written: nobody named the outcome first. I do, every time. That's finding your Moon.

Too many people start with ChatGPT. I start with the Moon.

The method isn't a slide, it runs live: this site and the operation behind it ship through it every day. See the architecture →

// how i think about it

Four hard parts,
and how I work them.

Most AI problems are one of these four - and they blend: adoption usually needs a little governance, and a stalled rollout usually ends in adoption done right. Here's how I think about each.

01

AI adoption & enablement

curious to fluent

The licenses are bought; usage is low. The way I move "we should try AI" to "we use it every day and love it" is rollout strategy, hands-on prompt training, the change management that makes it stick, and starter automations built with the team, not handed down to it.

The tell: seats are paid for but half the team still hasn't logged in.

↳ Someone who watched me do exactly this →

02

AI governance & guardrails

adopt fast, leave no mess

Most "AI adoption" quietly ships tech debt: unverified docs, hallucinations nobody caught, automations no one maintains. I build the opposite - someone owns every output, it's verified, maintained, and retired when it's stale. It's real governance, mapped to the NIST AI RMF (govern, map, measure, manage) that examiners and the NCUA point institutions to - with model risk (the 2026 interagency guidance that replaced SR 11-7), vendor and partner-ecosystem risk, explainability, and fair lending covered in plain English. Lightweight enough that people actually follow it.

The tell: the team is already pasting into AI, and the board will eventually ask who owns the results.
03

AI rescue & dig-out

stalled rollouts

Tools bought, pilot run, and it stalled - or it's a tangle no one trusts anymore. The way I unstick it is to audit what's actually there, find what's salvageable, and turn a stuck rollout into something that delivers. No rip-and-replace reflex; make the existing investment earn its keep.

The tell: "we already did AI" is said with a sigh, not a smile.
04

Fintech implementation & delivery

the launch whisperer

Fifteen years leading digital banking and fintech rollouts across Visa, Fiserv, and Lumin Digital - nine launches in five years, every one on time, on budget, and with teams who enjoyed the ride enough to send the recommendation later. That's the delivery discipline I bring to a launch that cannot slip.

The tell: a launch that cannot slip and a team that needs a steady hand.
// the short version

AI dos & don'ts,
on the house.

A taste of how I think about adoption. Consider this the napkin version.

Do
  • Start with one painful, repetitive task, not a moonshot.
  • Name a real internal champion and give them time, not just a title.
  • Write the dos and don'ts down where people can find them.
  • Measure adoption and outcomes, not vibes.
  • Let people watch a skeptic become a believer. It converts the room.
Don't
  • Buy a pile of seats before you have a use case.
  • Ban it in a memo and hope it goes away. It won't.
  • Let "AI policy" mean one Slack message no one reread.
  • Put sensitive data into tools you haven't vetted.
  • Assume the loudest skeptic won't convert. They convert hardest.
// the method

Leave It Better.

My framework for moving fast on AI without dumping a mess on whoever comes next. It's the NIST AI RMF, translated out of compliance-speak and into four things a real team can actually do. Governance as manners: own your outputs, and leave the place better than you found it.

// step 01 · own

Name an owner

Every automation, prompt, and AI-written doc gets a human whose name is on it - because "the AI did it" is not an owner.

// step 02 · map

Map where it lives

We find every place AI has quietly crept into the work and rank it by how much it'd hurt if it were wrong.

// step 03 · verify

Verify & maintain

We check it for accuracy, bias, and drift on a real cadence - not once at launch and then never again.

// step 04 · retire

Retire the stale

When something's outdated, we decommission it on purpose, so your knowledge base gets cleaner over time instead of muckier.

// don't just take my word for it

Working with me,
in their words.

“I worked with Elliott for three years, and honestly, she’s one of the best project managers I’ve teamed up with. She’s got a real knack for getting things done, keeping projects moving and handling whatever curveballs come her way without missing a beat! She was also always pushing the team to think about how AI fits into our workflows, always looking for smarter, more efficient ways to deliver for clients without losing the human touch that makes her so good at what she does. On top of all that, she’s just a great person to work with: thoughtful, resourceful, and someone you can always count on. Any team would be lucky to have her.”
Tom Hockenbery, PMP
Project Management Professional · worked with Elliott for three years
More recommendations →
// let's build something

Building an AI or
delivery team? Let's talk.

Standing up AI adoption, governance, or a launch that can't slip is exactly the work I want to be doing next. If that's what you're building, I'd love to be part of it.

Elliott Storms