Stacy Agobian, Good Faith EI

Stacy Agobian, Good Faith EI Contact information, map and directions, contact form, opening hours, services, ratings, photos, videos and announcements from Stacy Agobian, Good Faith EI, PO Box 12065, Irvine, CA.

I write a lot about who's accountable when an institution gets AI governance wrong.Here's the part I don't write about e...
08/14/2026

I write a lot about who's accountable when an institution gets AI governance wrong.

Here's the part I don't write about enough: the people on the other side of that failure.

When a system gets it wrong at scale, someone loses a job, a claim, a plan, a footing. They rarely get a hearing. They get a letter.

Today I'm launching Phoenix Rising Project by Good Faith EI: a California nonprofit public benefit corporation built for exactly those people.

Two groups to start:

โ†’ People displaced from their livelihood by institutional governance failures, and by AI and automation
โ†’ Women rebuilding after divorce, with children depending on how it goes

What we offer is education, not advice. The Phoenix Rising Method is a structured way to move from "I don't know what happens next" to a decision you can actually defend to yourself. It starts with seven questions most people have never been asked in order: what you're facing, what's actually urgent, what you're protecting, and what you want to be true in three years.

I've spent twenty years helping enterprises make defensible decisions with incomplete information. This is the same discipline, pointed at people instead of systems.

If you're in that spot right now, or you know someone who is

https://prei.global

Wintrust, a $71 billion bank, just told its investors that its AI could be biased, discriminatory, or wrong, in fraud de...
08/13/2026

Wintrust, a $71 billion bank, just told its investors that its AI could be biased, discriminatory, or wrong, in fraud detection and in products that handle customers' financial data.

Then, in the same breath, it named the control.

The control is a policy about which websites and AI apps employees are allowed to open on the company network.

Read that again. The risk Wintrust disclosed is an AI system producing an unfair or incorrect outcome that reaches a customer. The control it disclosed is an acceptable-use policy for staff. Both are real. But one does not govern the other. A rule about which tools employees may open is not a control over whether a model's decision is fair, accurate, or something you could defend to a regulator.

That's the gap I keep naming, and here it is in a company's own words, the risk is about the decision, and the control is about data hygiene. Between them sits the question no one answered. Who validates the model, who has the authority to overrule its output, and where is the dated evidence that the oversight actually runs?

To its credit, Wintrust was candid, it said it "may selectively" use AI, and admitted its own policies "may not be effective." That honesty is more than most companies offer. But naming a gap out loud is not the same as closing it.

So run the two-column test on your own disclosures. Column one, every AI risk you've named. Column two, the control you can actually attest to for each one. If the risk reads "biased or wrong decisions" and the matching control reads "we have a policy," those columns don't connect. That isn't oversight. It's a blind spot with a paper trail.

Which of your two columns is longer?

Commentary and opinion based on Wintrust's public SEC filings (2025 Form 10-K). No wrongdoing is alleged or implied. Not legal, financial, or compliance advice.

I found errors in my own Episode 7.So I did the thing I stand in front of boardrooms and tell them to do: I stopped, I w...
08/12/2026

I found errors in my own Episode 7.

So I did the thing I stand in front of boardrooms and tell them to do: I stopped, I went back to the source filings, and I corrected the record.

Call it my own Human Intervention Rate ticking up in real time.

Here's the irony I'm not going to pretend away. I build content with AI in the loop, and AI-speed confidence is seductive : fluent, fast, and perfectly happy to hand you a clean sentence that happens to be wrong. My whole thesis is that the human in the loop can't be ceremony. Episode 7 turned into a live test of whether I actually believe that. Turns out I do.

For the public record, three corrections:

๐Ÿญ. ๐—ก๐—ผ ๐—ด๐—ผ๐˜ƒ๐—ฒ๐—ฟ๐—ป๐—ฎ๐—ป๐—ฐ๐—ฒ ๐˜„๐—ฎ๐˜€ ๐˜๐—ผ๐—ผ ๐—ฏ๐—น๐˜‚๐—ป๐˜. Several companies I referenced do name a board committee with AI in scope : Centene, for one, names its Audit and Compliance Committee in its own 10-K. The precise (and stronger) critique was never "no oversight." It's the gap between naming a committee and being able to show an attestable control: a metric, a model inventory, a named framework, dated evidence.

๐Ÿฎ. ๐—Ÿ๐—ถ๐˜๐—ถ๐—ด๐—ฎ๐˜๐—ถ๐—ผ๐—ป ๐—ถ๐˜€ ๐—ฎ๐—น๐—น๐—ฒ๐—ด๐—ฒ๐—ฑ. Anywhere I referenced active cases, I've made it explicit : these are claims on the public docket being tested in court. Not findings. Not proof that any company's AI caused harm.

๐Ÿฏ. ๐—œ ๐—ธ๐—ฒ๐—ฝ๐˜ ๐—บ๐˜†๐˜€๐—ฒ๐—น๐—ณ ๐—ต๐—ผ๐—ป๐—ฒ๐˜€๐˜ ๐—ผ๐—ป ๐—”๐˜๐—ต๐—ฒ๐—ป๐—ฒ. It files a 10-K but no proxy, which means my own test literally can't be run on it. I left that in, out loud, instead of quietly dropping the inconvenient one.

None of this softens the point. It sharpens it. The gap between disclosed risk and attestable control is real, and it's exactly where the exposure lives. But if I'm going to hold companies to their own filings, I'd better hold myself to mine.

My co-host Aura spends the whole episode insisting human sign-off is "just ceremony." The joke lands on both of us: catching these errors was the ceremony : working exactly as designed.

Revised Episode 7 is live ๐Ÿ‘‰ https://youtu.be/SH5Loh35E8s

If you run AI-assisted anything : and you do : here's the free diagnostic: when your machine hands you something confident, who checks it before it ships? If the honest answer is "nobody," that isn't efficiency. That's the blind spot.

The human intervention isn't the slow part. It's the whole job.

๐Ÿ’ฌ So tell me: what's the last thing your AI handed you, confidently, that turned out to be wrong? Drop it in the comments : the good ones are worth more than any case study.

If this made you rethink who's holding the pen on your team, follow for more, and share it with the one person who needs to see it before their next deploy.

First Citizens' 10-K tells shareholders AI is a material risk.Its proxy says the board oversees market, credit, asset, l...
08/12/2026

First Citizens' 10-K tells shareholders AI is a material risk.

Its proxy says the board oversees market, credit, asset, liquidity, capital, operational, compliance, legal, strategic, and reputational risk.

Count them. That's ten. AI isn't one of them.

FCNCA, a top-20 U.S. bank, describes AI risk with specificity, including third-party models it can't see inside, generative systems that can produce wrong answers or leak confidential data, bias, and outputs no one can fully explain.

The proxy has a slot for cyber, technology, and every emerging risk a modern bank is expected to name.

No slot for AI.

Let me be precise. This isn't proof First Citizens does nothing about AI. It is proof that publicly inspectable governance treats AI as a risk to disclose, not a risk to govern.

They put an emerging technology before the board for cyber. They didn't for AI, whether by choice or because no one asked.

Disclosure is a sentence a lawyer writes once a year.

Governance is a paper trail with names, dates, and evidence attached.

Regulators are starting to ask for that paper trail. On the public record, First Citizens has none to point to.



Comment EI for a sample report. Book a no-charge 20-minute consultation with me at goodfaithei.com. This is a real consultation, not a sales pitch.

Disclosure does not equal an attestable control.UnitedHealth Group (UNH) did not get caught without AI governance.It had...
08/11/2026

Disclosure does not equal an attestable control.

UnitedHealth Group (UNH) did not get caught without AI governance.

It had an internal AI review board, oversight language, and a responsible-AI posture on paper.

Then, in Estate of Lokken v. UnitedHealth Group, a federal magistrate judge in Minnesota ordered discovery including records on how nH Predict works, AI oversight, and the identities of the internal review board.

UnitedHealth disputes the allegations and maintains nH Predict is a care-support tool, not a claims-decision tool. The claims are allegations and the case is ongoing.

The governance existed. The committee existed. But when a court asked for the control behind it, โ€œwe have a boardโ€ was not an answer.

This is the AI accountability gap.

The blast radius is operational, clinical, legal, and reputational. Liability exposure expands when no one can evidence who could intervene, when escalation fired, or how a wrong output was reversed.

This is the moral crumple zone : people nearest the automated system absorb blame for an outcome they could not inspect.

A policy is not a control.

An attestable control can evidence

Decision capacity
Human Intervention Rate
Escalation
Error correction
Human impact

Disclosure is present. Decision accountability is not.

Episode 7 : The Governance Gap
https://www.youtube.com/watch?v=09i77tTQuK4

Book a 20-minute consultation at https://goodfaithei.com.

Comment EI for the a sample report

First Citizens BancShares (FCNCA) runs automated credit decisioning and AML surveillance across a $200B+ balance sheet -...
08/10/2026

First Citizens BancShares (FCNCA) runs automated credit decisioning and AML surveillance across a $200B+ balance sheet - as every large bank now does.

To their credit, banks are the most model-governed institutions in the economy. First Citizens is supervised by the FDIC, the Federal Reserve, the CFPB, and North Carolina's Commissioner of Banks, and - at its size - maintains a board risk committee and a formal model-risk framework. Oversight isn't the question.

The question is narrower - does algorithmic fairness have a named owner? When automated credit models are governed as "model risk" and "compliance," disparate impact in who gets approved can fall between the technical validation of the model and the board's enterprise-risk lens - governed for accuracy, not for equity of outcome.

That's the gap worth asking about. Fair-lending law doesn't exempt automated systems - "the model decided" has never been a defense, and disparate impact is testable whether a human or an algorithm produced it. The banks that can evidence board-level ownership of algorithmic fairness, not just model accuracy, are the ones who answer calmly when asked.

Episode 7 - The Governance Gap
https://lnkd.in/eJbzMjqB

That's the layer I help build. 20-minute governance-architecture consult at https://goodfaithei.com. Comment EI for a sample report.

Most corporate AI governance is a warning label with no one standing next to the machine.In Episode 7 of Ethical Intelli...
08/07/2026

Most corporate AI governance is a warning label with no one standing next to the machine.

In Episode 7 of Ethical Intelligence, Aura and I break down the gap between AI disclosure and AI accountability across fourteen major issuers, featuring Intuit as our benchmark case study.

Watch Episode 7 at https://www.youtube.com/watch?v=09i77tTQuK4

Book a 20-minute consultation at https://goodfaithei.com

Comment SAMPLE for a free EI artifact.

06/19/2026

You aren't hiring operators. You're hiring human shields.
When the AI 'hallucinates' or fractures a community, the CEO doesn't go to jail. The 'human-in-the-loop' does.
Ron Bodkin calls it the Moral Crumple Zone. I call it professional cowardice.
If your system is designed to break a human before it breaks the company, you arenโ€™t a leader. Youโ€™re a predator.

In a low-stakes field, a crisis is a glitch in a video game. In legal, financial, or medical infrastructure, it is a sys...
06/18/2026

In a low-stakes field, a crisis is a glitch in a video game.

In legal, financial, or medical infrastructure, it is a systemic failure.

When a crisis hits high-stakes industries, the blast radius doesnโ€™t just affect the balance sheet. It hits the leaderโ€™s accountability directly.

Accountability is not a press release.
It is the weight of infrastructure built long before the storm.

Most leaders operate within a "Moral Crumple Zone": a space designed to absorb the shock of an ethical failure without actually protecting the humans at the center.

But when the system fails, the zone collapses.

What is the diameter of your blast radius?

The question isnโ€™t whether you have data.
Itโ€™s whether you have Decision Stewardship.

This is the technical evolution of compliance.

It is the bridge between a boardroom mandate and the reality of a high-stakes transition. It is the difference between a legacy and a liability.

Are you building a system, or just measuring the distance to the impact?

Letโ€™s look at the infrastructure.

06/17/2026

AI does not take the fall.
People do.

When an autonomous system fails, the blame gravitates toward the nearest human.
We call this the Moral Crumple Zone.

It is a structural flaw in how we deploy intelligence.
A human operator becomes the buffer for a system they did not build and cannot fully control.
They become the shield for corporate liability.

This is not a technical error.
It is a crisis of stewardship.

As leaders integrate AI into high-stakes transitions, the question is simple:
Are we building infrastructure that honors human judgment?
Or are we merely positioning people to take the hit when the algorithm fails?

In our latest conversation, Ron Bodkin explores the ethical reality of the Moral Crumple Zone and why disciplined risk stewardship is now a baseline requirement.

Clarity is not found in the data.
It is found in the standard we hold for human dignity.

Watch the full video with Ron Bodkin.

https://youtu.be/kXWO1b409w4

Address

PO Box 12065
Irvine, CA
92626

Opening Hours

Monday 8am - 5pm
Tuesday 8am - 5pm
Wednesday 8am - 5pm
Thursday 8am - 5pm
Friday 8am - 5pm

Website

Alerts

Be the first to know and let us send you an email when Stacy Agobian, Good Faith EI posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Shortcuts

Share