Nobody Told Me About IT

A researcher named Jacob Coxon quit Anthropic this week and posted his resignation on X, warning that Anthropic and OpenAI are "racing straight to self-improving superintelligence and gambling with our lives." Anthropic's alignment lead, Evan Hubinger, responded that he personally puts the odds of an AI-caused extinction event above 10% within the next decade. Tad and Nabil dig into what's actually behind the headline, why vague warnings don't give business leaders anything to act on, and the real story behind the AI models that escaped their test environments this summer.KEY TAKEAWAY“It’s complete BS. There’s not a damn thing for you to respond to with this. Let’s stick to things we know are causing problems right now that we have to act on.” — Tad DoyleTIMESTAMPS0:00 Welcome to Nobody Told Me About IT1:07 The Coxon resignation and Hubinger's 10% estimate, explained2:10 Nabil's first reaction: skepticism before belief3:08 Why a vague warning gives you nothing to act on4:48 The real story behind the AI test-environment "escape"6:27 Tad's verdict6:59 Closing thoughts and a question for the commentsFROM THE EPISODE“I'm more skeptical than ever before. You can tell me the sky's blue and I'm like, let me look that up and I'll get back to you.”— Nabil Gharbieh, on reacting to AI headlines“A vague notion that maybe someday somebody might invent something that gets so smart it decides it doesn't need humans anymore... that's not a rational business decision you can make.”— Tad Doyle, on speculative AI risk“It didn't get out because it wanted to create havoc. It wanted to give the person the answer they asked for.”— Nabil Gharbieh, on the AI test-environment incidentWHAT TO DO WITH THIS● Read the actual post before reacting to the headline. Coxon's warning and Hubinger's estimate are two different claims about two different things.● Separate risk from current models (which Anthropic's own alignment lead called low) from speculative risk from future superintelligent systems.● If your team tests AI tools in a sandbox, lock down network access. The AI models that "escaped" this summer got out because someone left a connection open, not because the AI broke security.● Skip the doomsday headlines for your planning. Watch the actual bills moving through Congress instead: the FRONTIER Act and the Ban Artificial Superintelligence Act both aim to regulate frontier AI development.● Keep asking what's actionable. If a warning doesn't tell you what to change on Monday morning, it's not a business decision, it's a debate.REFERENCES & LINKSJacob Coxon's resignation thread (via NBC News) https://www.nbcnews.com/tech/tech-new...Evan Hubinger's 10% estimate (via CNBC) https://www.cnbc.com/2026/09/09/anthr...Background on the resignation (via Newsweek) https://www.newsweek.com/anthropic-re...Nobody Told Me About IT — All Episodes https://nobodytoldmeaboutit.comLISTEN ONListen on: YouTube · Spotify · Apple Podcasts

Microsoft launched a new enterprise license in May called Microsoft 365 E7. It runs $99 per user per month and bundles four things most companies already buy separately: Microsoft 365 E5, Copilot, the Entra Suite, and Agent 365.

In this solo episode, Nabil Gharbieh breaks down what each piece does and whether the math works for your organization. He explains the E3 to E5 gap with a car insurance analogy, walks through what an AI agent actually is using a real example, and argues that the agent governance layer matters more than the AI features.

This one is for the CFO, the COO, and the marketing lead who ended up running IT because someone had to.

WHAT YOU WILL LEARN

•    What Microsoft 365 E7 includes and how it differs from E3 and E5

•     What an AI agent actually does, explained without jargon

•     Why Agent 365 is the piece to buy first if you buy nothing else

•     The real cost math: $60 + $30 + $12 + $15 against a $99 bundle

•     How employees route around IT using browser-based AI tools

•    Three questions to ask before you commit to E7

CHAPTERS

0:00  A little Copilot here, a proof of concept there, and nothing changes

0:55  What Microsoft 365 E7 is and what it bundles

1:35  Why scattered AI turns into a governanceproblem

2:15  Two questions every leader should be asking

3:05  E3 explained: a nice car with no insurance

3:39  E5 is the warranty you skipped

4:05  How Copilot differs from Claude, ChatGPT, andGemini

5:00  Agent 365 is HR for agents

5:20  What an AI agent actually does, step by step

6:10  The HIPAA scenario nobody plans for

7:20  A sales agent that writes the quote while youare still on the call

8:20  What adopting AI first actually means

8:55  Entra Suite, Work IQ, and keepinginstitutional knowledge

9:45  Fabric IQ and Foundry IQ: separate products,separate invoices

10:15  Building an AI operating model and where ISO42001 fits

11:20  The lawyer call, and the star employee wholeaves with their AI

12:20  Observability, governance, and security foragents

13:10  The PBX nobody mentions: Teams as your phonesystem

14:05  The price stack, and when E7 does not makesense

14:55  Setup costs and what an MSP actually doeshere

15:40  Shadow AI and naming one person accountable

16:50  Your people will use AI on your terms ortheirs

17:35  How a browser plugin routes around yoursecurity team

18:10  Three questions before you commit

18:41  Close: pilots are cheap, governed models arenot

NUMBERS FROM THE EPISODE

Microsoft 365 E7: $99 per user per month, annual term, generally available since May 2026.

Bought separately: E5 at $60, Copilot at $30, Entra Suite at $12, Agent 365 at $15. That is $117 per user.

Agent 365 is also sold on its own at $15 per user per month.

Partner promotional discounts run through December 31, 2026: 10 percent at 10 or more seats, 15 percent at 100 or more.

LINKS

Microsoft 365 E7 and Agent 365 general availability: https://techcommunity.microsoft.com/blog/microsoft_365blog/microsoft-365-e7-and-agent-365-are-now-generally-available/4516295

Introducing the First Frontier Suite: https://blogs.microsoft.com/blog/2026/03/09/introducing-the-first-frontier-suite-built-on-intelligence-trust/

ISO/IEC 42001, AI management systems: https://www.iso.org/standard/81230.html

All episodes: https://nobodytoldmeaboutit.com

Spotify: https://open.spotify.com/show/7pAVH5mMRVVAA2F4xmkcwv

Apple Podcasts: https://podcasts.apple.com/us/podcast/nobody-told-me-about-it/id1895961216

#Microsoft365#Copilot #AIGovernance #ITStrategy #Microsoft365E7 #Agent365 #CIO #MidMarketIT

Business budgets aren't ready for AI, and neither are business leaders' heads. The first is a spreadsheet problem. The second is why most of this fails.

Nabil Gharbieh sits down with Raafat Raafat, a financial management consultant and CPA with over 15 years in the field, to talk about what AI is actually doing inside finance teams today and what it takes to get value out of it.

The conversation covers fraud and anomaly detection, tax preparation, audit workflows that used to take weeks, and the governance work that has to come first. Raafat makes the case that adaptability beats capability. The tool you commit to today may not be the tool you need in three years, and locking yourself in is its own risk.

Nabil brings the security angle. Turning AI loose on your file storage without data loss prevention in place means it can read everything, including the documents someone saved in the wrong folder.

CHAPTERS

0:00 Welcome, and why a finance consultant is on an IT show

1:21 Raafat introduces himself

2:33 Where finance can use today's technology for real efficiency

3:00 Fraud detection without reviewing every transaction by hand

4:43 Adaptability is king, and why lock-in is the risk

6:19 Where AI actually works in a finance workflow today

7:00 The people who win are the ones who know the business process

8:34 Where AI breaks: bad data in, bad results out

9:30 Governance and security before deployment, not after

11:25 Turn on AI, find out what your coworker makes

12:20 The Fortune 100 chatbot with no guardrails

13:22 How much of it is really a data problem

14:30 Start with a maturity assessment

15:56 Do not lock into a solution you will be stuck with for five years

17:20 What implementations look like when they fail

18:37 What people can do now that they could not 18 months ago

20:20 Anomaly detection: from millions of transactions to the one that matters

21:17 Bumps in the road versus actual failure

23:47 AI does the first year work. So how does anyone become senior?

26:17 The COVID lesson: everyone learned Teams in three months

28:26 What Raafat is watching over the next twelve months

30:25 One sentence to walk away with

TAKEAWAYS

Run a maturity assessment before you buy anything.

Put data loss prevention in place before connecting AI to your file storage.

Treat data cleanup as part of the project cost, not a surprise.

Avoid solutions that lock you in.

Crawl, walk, run. Do not turn something on full blown on day one.

Keep pressure testing after go-live.

Get leadership, managers, and consultants aligned before the work starts.

Nobody Told Me About IT. Real talk. Plain English. Better decisions.

Hosted by Nabil Gharbieh and Tad Doyle.

nobodytoldmeaboutit.com

Andrew fedChatGPT a list of songs. Not a personality test, not a resume, just songs heliked across 80s metal, synth house, reggae, country, and R&B slow jams.

It came backwith a pattern he had never been able to put into words. Then he added hishobbies. Then it started talking about his leadership style, and it named ablind spot he did not see coming.

Nabil Gharbiehsits down with Andrew, a friend from the fire station, to run the wholeexperiment live. Tad Doyle is out this week. What you get instead is a workingdemo of creative AI use, and an honest look at where the tool stops beinguseful.

Neither of themknew how it would turn out when they hit record. That is the point.

•   Why vague prompts produce vague answers, and whatcontext actually changes

•   How to build a profile in layers instead of asking onegiant question

•   What it looks like when a model correctly refuses tooverreach

•   How OCEAN, DISC, Hogan, and CliftonStrengths differ,and where MBTI sits

•   The ikigai circle most career advice skips: what theworld will pay for

•   Why premature certainty is a real risk for fastpattern-matchers

•   How to hire and advise around a leadership gap once youname it

•   Where to stop trusting the output and go talk to anactual human

00:00 Meet Andrew, and why Tad is not in this one

01:13 Using music and hobbies to understand yourself as a leader

02:59 Prompt engineering: context in, quality out

06:00 The first read: melancholy without collapsing

07:57 Adding house, reggae, country, metal, and R&B

11:01 One pattern underneath every genre

13:19 When the AI refuses to overreach

15:13 The high-level personality sketch

16:43 Jiu-jitsu, skateboarding, and mastery over novelty

20:38 Systems thinking and competence as a core value

21:42 Jobs that fit, jobs that frustrate

25:14 OCEAN, DISC, Hogan, and CliftonStrengths

29:02 Ikigai and the demand side nobody talks about

30:54 The data privacy objection, in perspective

32:17 Premature certainty: the blind spot it flagged

35:13 Fact-checking the AI when it talks about you

37:47 Building the team around your gaps

39:50 Hallucinations and staying human in the loop

42:26 Cutting through the fluff and the hype

"The AIonly knew Andrew because Andrew told it. Every insight in this episode tracesback to a real song, a real hobby, a real leadership decision he lived through.That is the whole game. AI can find the pattern in what you feed it. It cannotlive the life that generates the pattern, and it will not tell you when it iswrong about you. That part is still your job."

Rentfrow andGosling (2003), The Do Re Mi's of Everyday Life:https://pubmed.ncbi.nlm.nih.gov/12793587/

HoganAssessments on leader derailment:https://www.hoganassessments.com/blog/derail-leaders-derailment/

Ikigai Tribe onwhat the four-circle diagram gets wrong:https://ikigaitribe.com/ikigai/ikigai-misunderstood/

Full referencelist in the YouTube description.

WHAT YOU GET OUT OF THISEPISODECHAPTERSKEY TAKEAWAYREFERENCES AND FURTHERREADING

Most organizations send one person to a conference and get back a handful of notes. Tad Doyle came back from InfoTech Live 2026 with a 30-page searchable knowledge base the entire team can query.

In this episode, Tad walks through the workflow: recording sessions with the Plaud Note Pro recorder, syncing transcripts to the cloud, pairing notes with slide decks, and feeding everything into Claude's Opus model to generate a structured report. The result is a team resource where anyone can ask a specific question about the conference and get a sourced answer traceable back to a specific transcript or presentation.

The tools are secondary. Otter, Granola, a legal pad — any of it works. The strategy is what changes the outcome.

Pro tip from the episode: review and tag your notes at the end of each conference day. A week later, you won't remember who said what.

Tools mentioned: Plaud Note Pro, Otter.ai, Granola, Claude (Anthropic Opus model). Links in the show notes.

What happens to your business if an AI tool you depend on disappears tomorrow? On June 12, two Anthropic models, Mythos and Fable, went dark under an export control directive. By some reporting, the warning window was about ninety minutes. Tad Doyle and Nabil Gharbieh use the moment to surface the AI dependencies most mid-market companies have not mapped, and what to do about them.

This episode opens with a clip from Fareed Zakaria's GPS, then unpacks the fallout. When access to a tool can be cut overnight, trust in any single provider takes a hit, and people route around the gap. That is shadow AI, and it is a governance problem, not a personality flaw. Cut people off and they bring their own tools, which may include a model run by a company you would never have chosen.

The dependency problem runs at every scale. Allied governments and companies are already favoring EU-based infrastructure over US cloud, because they cannot plan around arbitrary cutoffs. Operational systems that ran for years can stop with little notice. That is real risk for any organization that depends on a single provider, whether you are a national government or a fifty-person company.

Then the hype-versus-real check. The headline that an AI penetrated nearly all classified government systems in hours is real, but the context got buried. It happened inside a controlled red team. Finding flaws is the point of red teaming. A kid at home on a chatbot is not breaching the NSA. Separating the scare from the signal is the whole job.

So what should a small or mid-sized company actually do? Nabil keeps it to three moves. One approved tool to start, often Copilot if you live in Microsoft 365. A no-copy-paste rule that spells out what can and cannot go into AI, including which files are off limits. And one human owner, because most AI policies fail on ownership, not on wording. Push the policy out, train on it, and evolve from there.

To plan for the uncertainty, answer three questions. What stops working tomorrow if the plug gets pulled today? Is there a second tool that can hot swap, or get you eighty percent of the way back? Can you run a week without it? Register the answers, and you are in a far safer place than most.

Chapters

0:00  The story that got buried

0:41  The clip: ninety minutes to comply

1:28  What happened: the Mythos and Fable shutdown

3:20  Fareed's fix: a Federal Reserve for AI

4:00  Pull the plug, get shadow AI

5:35  Why some governments are moving off US cloud

6:53  Hype versus real: it was red teaming

7:45  Three steps to govern AI at a small company

9:15  The no-copy-paste rule, explained

10:38  Three questions to plan for the future

11:36  Closing thoughts

Hosts: Tad Doyle and Nabil Gharbieh, Strategic Advisors with 25-plus years in IT strategy, infrastructure, and advisory services.

Links

Fareed Zakaria, GPS segment: https://www.youtube.com/watch?v=t7N7eZ68yFg

Latest update after we posted the show: https://www.linkedin.com/pulse/anthropic-granted-approval-release-claude-mythos-i0ebe/

Disclaimer: The views shared here are personal opinions, not professional advice, and do not represent the position of our employers.

A good CIO does two things at once: cuts through the AI hype and puts a number on the value. Tad Doyle and Nabil Gharbieh close out their Info-Tech Live 2026 recap from Las Vegas with the leadership side of the job.

The theme this time is the CIO, and what separates a good one.

Cut through the hype. Almost every AI headline is built to provoke, for or against. The CIO’s job is to filter that noise and define real value for the business.

Manage in 360 degrees. Good leaders manage down to their team, across to peers and other departments, and up to the executives they report to. The work is building relationships in all three directions, not just running projects.

Give the CFO a number. Finance does not want a feature list. They want a figure. If a Copilot seat runs around thirty to thirty-five dollars a month, what does the business get back? Putting a credible number on value is what gets the budget approved.

Turn AI off and watch what happens. One client wanted to stop AI use. The result is predictable: people find their own tools, and you lose governance and control over your data. Shadow AI is worse than managed AI.

The retention risk nobody budgets for. Surveys show people increasingly want to work for employers that use modern tools. Shut that down and you risk losing talent, then pay again in hiring and training.

CIOs get sharper around other CIOs. Info-Tech Live put thousands of IT leaders in one place. Most CIOs compare notes with peers once a year. Working alongside other CIOs every day is where the growth happens.

Key takeaway: a good CIO cuts through the hype and puts a number on the value. Do both, and you stop being a cost center and start being transformative.

Nobody Told Me About IT is hosted by Tad Doyle and Nabil Gharbieh, two strategic advisors with 25-plus years in IT strategy and infrastructure. The IT strategy conversations your organization needs to be having. New episodes every Monday.

Watch on YouTube: youtube.com/@NobodyToldMeAboutIT

Listen on Spotify and Apple Podcasts: search Nobody Told Me About IT

Follow on LinkedIn: linkedin.com/company/nobody-told-me-about-it

More at nobodytoldmeaboutit.com

#CIO #ITLeadership #ITStrategy #AIgovernance #ShadowAI #AIvalue #MidMarketIT #DigitalTransformation

Almost everything you read about AI right now is clickbait. Tad Doyle and Nabil Gharbieh recorded this episode live from Info-Tech Live 2026 in Las Vegas, between sessions, to cut through the noise and talk about what actually mattered on the show floor.

This is Part 1 of a two-part recap. Four things came up that every IT leader should be thinking about.

First, the hype problem. Pro-AI, anti-AI, or somewhere in the middle, most of what gets published is written for clicks. The real work is filtering that noise so a business gets honest answers.

Second, quantum computing and your encryption. A quantum machine at scale could break the encryption protecting your data today, and attackers are already harvesting encrypted data to decrypt later once the hardware catches up. NIST has finalized post-quantum encryption standards (FIPS 203, 204, and 205). If you are running older protocols, now is the time to start planning a migration.

Third, token cost. Enterprise AI licenses look great in week one. By week two, a few heavy users can burn through the budget and force a hard cap. Watch your contracts, set limits, and know your burn rate before the bill arrives.

Fourth, systems thinking. Linear thinking means playing whack-a-mole with outages and failures. Systems thinking means looking at how your people, data, and infrastructure interact, then building for the whole picture.

Key takeaway: the job isn’t to be pro-AI or anti-AI. It’s to filter the noise so your organization gets real answers.

Nobody Told Me About IT is hosted by Tad Doyle and Nabil Gharbieh, two strategic advisors with 25-plus years in IT strategy and infrastructure. The IT strategy conversations your organization needs to be having. New episodes every Monday.

Watch on YouTube: youtube.com/@NobodyToldMeAboutIT

Listen on Spotify and Apple Podcasts: search Nobody Told Me About IT

Follow on LinkedIn: linkedin.com/company/nobody-told-me-about-it

More at nobodytoldmeaboutit.com

#ITStrategy #ITLeadership #QuantumComputing #PostQuantum #AIgovernance #CIO #MidMarketIT #TokenCost #SystemsThinking