The Digital Dilemma

You can introduce new technology, rewrite a process, or promote your best performer. Your team still needs clarity, trust, and leadership to make the change work.

In Episode 49 of The Digital Dilemma, Nick Dreyfus welcomes Jennifer Santoso of VAST Leadership Development for "People First: Leadership, Change Management & AI." They explore leading through uncertainty, strengthening communication, and helping teams grow.

Jennifer draws on 15 years of management consulting supporting the Department of Defense and experience developing supervisors at San Diego Gas & Electric. Her approach starts with understanding how people experience change, involving employees in the process, and helping them move forward.

Nick and Jennifer discuss AI as a thought partner and a tool for repetitive work while emphasizing human judgment, relationships, and accountability. They explore how leaders can address concerns about changing roles.

IN THIS EPISODE

Change management: Anticipating disruption, involving the people doing the work, and supporting adoption.

Communication and trust: Why leaders become "chief reminding officers," how consistent communication helps, and why authentic conversations matter more than checking a box.

AI and human judgment: Where routine assistance can help, and why context, tone, and expert review remain important.

Competence and character: The two sides of trust and what happens when a leader has one without the other.

Coaching and blind spots: How openness, feedback, and honest conversations help people develop as their responsibilities evolve.

The promotion trap: Why being excellent at a job does not automatically prepare someone to manage the people doing it.

Leadership development: Practicing interpersonal skills, applying lessons between training sessions, and revisiting them as you grow.

Career progression: Setting expectations, discussing added responsibilities, and defining milestones so advancement becomes a shared effort.

The conversation also tackles a difficult balance: leaders need confidence that someone can handle a bigger role, while employees need clarity and recognition when they take on more. Jennifer explains how regular development conversations can reduce uncertainty on both sides.

For CEOs, business owners, nonprofit leaders, managers, and emerging leaders, bring these questions to your next team meeting: What is changing for our people? Have we explained why? What support do they need? Are we helping our future leaders prepare?

CONNECT WITH JENNIFER SANTOSO

Explore leadership development, workshops, and coaching with VAST Leadership Development.

Website: https://vastleader.com

LinkedIn: https://www.linkedin.com/in/jenn-santoso-7427394/

CONNECT WITH NICK DREYFUS AND i-NETT

Managed IT and cybersecurity: https://www.i-nett.com

Fortify AI and AI readiness: https://i-nett.ai

Nonprofit technology resources: https://nonprofit.i-nett.com

Book a conversation: https://meetings-na2.hubspot.com/ndreyfus

Email: [email protected]

Phone: (858) 337-2866

Follow The Digital Dilemma for practical conversations about technology, leadership, and business. Share this episode with a colleague navigating change.

Data, Skills, Ownership: The 3 Fixes That Make AI Work in Business
The Digital Dilemma | Episode 48 | Nick Dreyfus

Do you know how your employees are actually using AI, or just what your policy says they’re allowed to do?

When a deadline is approaching and the work keeps piling up, people look for help. Without a useful, approved way to use AI, that can mean uploading company information into a personal account, trusting an answer nobody checked, or quietly building a process leadership knows nothing about.

In this episode, I break down three things every organization needs to address to make AI useful in everyday work:

• Data: What information can your team use, where can it go, and what might they be sharing without realizing it?
• Skills: How do you give people practical guidance and support without expecting everyone to become an AI expert?
• Ownership: Who tests the system, approves what it can do, and takes responsibility when it gets something wrong?

I also challenge the idea that a confident answer deserves our trust. You wouldn’t keep expanding an employee’s responsibilities if they were confidently giving customers incorrect prices or making promises they couldn’t keep. AI needs scrutiny, too. We discuss hallucinations, the limits of benchmark scores, and why clear permissions and human accountability matter long after the demonstration ends.

The goal is to give your team AI they actually want to use because it reduces the volume of boring, annoying, repetitive tasks taking over their day. That means more room for sound judgment, better customer service, and the work that helps people and businesses grow.

You’ll leave with a practical Monday checklist: understand what’s already in use, set clear data rules, name an owner, and choose one repetitive task to improve and measure.

At i-NETT, that’s the conversation we want to have with you about Fortify AI: your people, your work, and the tools and ongoing support needed to make a useful difference.

Bring one task your team is tired of doing the hard way. Let’s talk about it.

• Book a meeting with Nick: Schedule your AI consultation
• Explore Fortify AI: i-nett.ai
• Find i-NETT: i-nett.com
• Listen to more episodes: The Digital Dilemma

If this sounds like a conversation your leadership team needs to have, share this episode with them.

Copilot will hand your receptionist the expansion plan, if your folder permissions allow it.

Who can read your salaries, margins, client files and expansion plans? Think about everyone who changed roles, covered for someone or inherited access over twenty years.

In Episode 47 of The Digital Dilemma, I explain why "We've got Copilot, we're totally good" can create a false sense of security. I'm Nick Dreyfus. At twenty-three, two weeks into a new job, I opened the HR folder and could see what everyone earned.

The same problem showed up during a financial services firm's AI readiness review. The receptionist had access to the manager's expansion plans. She wasn't snooping. She didn't know the access existed. Neither did the people preparing to deploy Copilot.

Copilot respects the permissions you give it, including the ones your organization has forgotten. AI can make information much easier to find and connect. A folder that used to take hours of clicking to discover may now become part of an answer to a simple question.

For a CEO or CFO, that raises questions about confidential strategy, employee information, competitive advantage and client trust. Buying an AI license doesn't settle who should be able to see what.

We explore how this happens: IT handoffs, broad department groups, old archives and temporary access that stays forever. Somebody was trying to help. Years later, nobody remembers why access was granted or whether it still makes sense.

We examine two scenarios: an employee joining a competitor with years of accumulated access, and an attacker using a compromised employee account. Both show why access reviews, identity protection and activity visibility matter.

Here's what you'll take away:

• The "newspaper test" for separating everyday files from sensitive information.
• A green, yellow and red approach to reviewing what your team can access.
• Why "give her what Bob has" deserves a second look.
• How outdated folders and forgotten copies can complicate AI readiness.
• What to review when people change jobs, help another department or leave.
• Why an independent review can support your IT team and expose blind spots.
• Where assessment tools help, and where human judgment is still essential.

I also discuss the questions behind choosing an AI solution. Copilot and Fortify AI should be evaluated around the work you need done: connected business systems, model choices, guardrails, activity visibility and ongoing management. Check the capabilities and configuration of the solution you're buying. Every platform, including ours, depends on the permissions underneath it.

Have someone outside your organization review who can access your sensitive folders and ask why. Start with the files that would matter most if the wrong person read them. If you're already using AI, make that review part of the work you're doing now.

This episode is for CEOs, CFOs, owners and nonprofit leaders at organizations with 25 or more employees, especially teams of 50 to 1,000. Bring it to your next leadership meeting and ask: do we know who can read our expansion plan?

Want an outside set of eyes? Let's talk about AI readiness, access permissions and what your business needs.

Book a conversation with Nick Dreyfus:
https://meetings-na2.hubspot.com/ndreyfus

Fortify AI and AI readiness:
https://i-nett.ai

i-NETT managed IT and cybersecurity:
https://www.i-nett.com

Nonprofit technology resources:
https://nonprofit.i-nett.com

Follow The Digital Dilemma for practical AI and cybersecurity conversations for business leaders.

Your team focuses on growth. Our team makes that happen.

If an employee pasted your client list into a free AI account this morning, how long would it take you to find out?

An AI policy is not the same as knowing what actually happened.

In Episode 46 of The Digital Dilemma, "The Guardrails Were Off," I revisit the OpenAI and Hugging Face story from Episode 40 and bring it back to your business: who decides what your team can share with AI, and how do you know those decisions are being enforced?

For CEOs and CFOs, this is a conversation about client trust, financial information, contractual obligations and getting useful work done without creating unnecessary exposure.

Your employees should not have to become security experts every time they open a chat window.

In this episode:

- Green, yellow, red: a practical way to separate public information, data that needs protection, and information that should stay out of AI tools.

- What a controlled AI environment requires, including an organization-controlled workspace, written commitments about retention and model training, role-based access, query logs and approved system connections.

- Why AI governance requires ongoing configuration and oversight, not just a policy document or a software purchase.

I also give you three questions to bring to your leadership team this week:

Can you produce a record of your team's AI prompts from last month?

Do you know which models answered and whether the information was retained?

Who decided what employees could share, and how would you know if they got it wrong?

This is the business side of shadow AI and data protection: helping your people work productively without leaving them to guess about information they are responsible for protecting.

Designed for CEOs, CFOs and leadership teams at organizations with 25 or more full-time employees, especially those with 50 to 1,000.

Not sure where that line sits in your organization? Book a conversation with me at i-NETT. Let's discuss what your team is using, what information is involved, and where clearer rules and better controls could help.

CONNECT WITH NICK DREYFUS

Book a conversation:

https://meetings-na2.hubspot.com/ndreyfus

Fortify AI by i-NETT:

https://i-nett.ai

i-NETT managed IT and cybersecurity:

https://www.i-nett.com

Listen, then bring the three questions to your next leadership meeting.

Your team focuses on growth. Our team makes that happen.

For general educational purposes only. Not legal advice.

You Have An AI Policy. Your Vendors Have A Free Account.

An executive questioned one number in one ad. The marketing agency's answer stopped him cold: we got the numbers from you. Inside a shared marketing folder sat an accounting spreadsheet someone had dropped there by accident, and for two months the agency's AI had been reading the whole folder and building campaigns from it. The AI saw their real costs and priced an entire product suite at roughly a ten percent markup instead of their normal margin. One ad ran. Nobody lied. Nobody was hacked. Somebody dragged a folder, and an AI nobody approved made pricing decisions in public.

Host Nick Dreyfus, VP of Business Development at i-NETT, argues that if your own house is in decent shape, your biggest AI risk is not your team. It is the vendors you trust most. You give your bookkeeper one hundred percent of your data. Your accountant, your marketing agency, your billing company, all of them are holding everything, and most businesses have never asked one question about what those vendors do with it in the age of AI.

In this episode: the clean kitchen problem, why a spotless AI policy tells you nothing about your supply chain. The two question email to send every vendor Monday morning, free. The move almost nobody makes: putting vendor AI work inside your environment, behind your own security, and offsetting the license against their bill. Why we've got Copilot, we're totally good is not the finish line when twenty years of file permissions say otherwise. And the money nobody discusses: in client environments we have audited, do it yourself AI agents commonly cost three to five thousand dollars a month to run, usually because they are doing way too much thinking on a model that is too expensive for the task. Built right, the same agent often runs at a fraction of that.

Questions answered: Which of my vendors are running my data through AI right now? What am I allowed to ask them? What does a real AI policy cover? Why do most AI projects show no return? What should an agent actually cost to run, and who is watching it after it ships?

Start here, and it costs nothing. Send every vendor who touches your data two questions: are you using AI on any of our data, and if so, what guardrails and security do you have around it? Everything in this episode was preventable. All they did was not ask a couple of questions.

Nick is not an attorney and this is not legal advice. Vendor obligations depend on your contracts, so bring them to your counsel.

AI readiness: https://i-nett.ai | Managed services: https://i-nett.com | Technology Fund: https://nonprofit.i-nett.com | Book time: https://meetings-na2.hubspot.com/ndreyfus | [email protected] | (858) 337-2866

Tags: AI policy, shadow AI, vendor risk, AI agents, token costs, Copilot, AI readiness, managed IT, MSP, cybersecurity, CEO, CFO, small business, San Diego, The Digital Dilemma, Nick Dreyfus, i-NETT

The Guy You Fired Still Has The Keys

He was not fired. He resigned on good terms, and there was a party. That is worse, because when somebody leaves happy, nobody is watching. His login stayed active, since disabling it was a five minute job on a list of four hundred things. That account still had access to the company bank account, and the password on it was sitting in a public credential dump. Nobody was hacked. Somebody just did not do the job.

Host Nick Dreyfus, VP of Business Development at i-NETT, argues that nobody attacks your best system. They attack your cheapest one. And your cheapest one got there one of two ways: a job that never got done, or a budget set by someone who was never given the full picture. Confuse those two and you make the wrong call every time.

In this episode: why your IT director spends the day resetting passwords, and what falls off the list when they do. Which layer of support is worth a senior salary and which two are not. Why the person who knows how bad it is has every reason not to tell you. The technology leader who did not know what business continuity meant, and the one to two week recovery nobody upstairs knew about. How attackers reached one of America's largest retailers through an air conditioning contractor running free consumer antivirus. What happens inside an IT company when margins get squeezed, and which tool quietly gets swapped. And a company that paid zero ransom and still lost two months, hundreds of thousands of documents, and a quarter of its revenue.

Questions answered: How do I know if my IT team is telling me everything? What should I ask before approving another technology budget? Are we running anything that stops getting security updates this year? What happens when someone leaves, and who verifies it? If we were encrypted tonight, how many days until we operate again?

Start here, and it costs nothing. Tell whoever runs your technology that nothing they report will be held against them, and mean it. Ask what happens when somebody leaves, then compare enabled accounts against headcount. Then ask your leadership which three customers would leave if you went dark for six weeks, and what those three are worth as a share of profit. That number is your real security budget.

The full list of fifteen questions, plus every source cited, is at thedigitaldilemma.riverside.com

Nick is not an attorney and this is not legal advice. Breach disclosure duties vary by state and industry. Insurance policies differ, so read yours and get your broker's answer in writing.

AI readiness: https://i-nett.ai | Managed services: https://i-nett.com | Technology Fund: https://nonprofit.i-nett.com | Book time: https://meetings-na2.hubspot.com/ndreyfus | [email protected] | (858) 337-2866

00:00 He was not fired, and that is worse 03:45 Why the five minute job never gets done 08:00 The word that scared me 09:15 The air conditioning guy 13:30 Why the cheap thing exists 16:30 What happens when margins get thin 19:00 Two months down, zero ransom paid 24:30 The seven questions 30:30 The ones with the least 32:00 What to do Monday

Tags: cybersecurity, offboarding, business continuity, ransomware, Server 2012 end of life, cyber insurance, managed IT, MSP, CEO, CFO, small business, San Diego, The Digital Dilemma, Nick Dreyfus, i-NETT

Your Marketing Agency Is Running Your Client List Through AI

The meeting ended. The guests dropped off. The team kept talking about costs and relationships. The AI note taker was still in the room, and it emailed the transcript to everyone on that call.

Nobody was hacked. Nobody broke a policy, because there was not one. And there was no undo.

Host Nick Dreyfus, VP of Business Development at i-NETT, argues your company already has an AI program. You did not approve it, you did not budget for it, and you cannot see it. Half of it is your team. The bigger half is everyone outside your building who already has your data.

In this episode:

• The one document almost no company has, and why "everybody knows the rule" falls apart when it matters
• Why shadow AI is a procurement failure, not a discipline problem
• Two risks everyone mashes together: a vendor reusing your work, and your data landing where you cannot see it
• The free version is free because you are the payment
• Why an AI tool wired into your CRM is an authorized skeleton key
• The four questions to ask any vendor touching your data, and what to do when they get vague
• Why Uber burned its 2026 AI budget in four months with nobody misusing anything
• Why we are at 1998, and why committing to one model now is a bet you do not have to make

This episode answers:

• Which of my vendors is running AI on my company's information?
• Do I need employees to sign something naming approved AI tools?
• Can a vendor reuse the work they did for me and sell it to my competitor?
• How does a small company push back on a much larger vendor?
• Who should pay when a vendor wants to use AI on my account?

Start here, and it costs nothing. List every outside party that touches your information. Send one email asking whether they use AI on your account, which tools, what tier, and whether your material trains anything. Ask for it in writing. The answers, including the vague ones, tell you where you stand.

Nick is not an attorney and this is not legal advice. Have your counsel review any policy or vendor document before anyone signs it.

AI readiness assessment: https://i-nett.ai

Managed services and cybersecurity: https://i-nett.com

Community Technology Fund: https://nonprofit.i-nett.com

Book a time with Nick: https://meetings-na2.hubspot.com/ndreyfus

[email protected] | (858) 337-2866

Chapters: 00:00 The note taker that stayed in the room

01:40 We deploy AI for a living

03:20 The paperwork almost nobody has

08:30 The part nobody talks about

09:40 Your work walking out the door

12:00 You are the payment

14:30 The skeleton key in your CRM

16:30 The four questions to ask every vendor

20:30 Uber burned a year of budget in four months

22:30 What to do this week for nothing

25:00 The decision that had to happen first

Tags: shadow AI, third party risk, vendor management, AI governance, AI usage policy, data privacy, cybersecurity, managed IT services, MSP, small business, San Diego, intellectual property, data leakage, vendor agreements, The Digital Dilemma, Nick Dreyfus, i-NETT

One in eight things it tells you is made up, and it says every one of them with the same confidence. You would fire that employee. So why did you hand it your invoicing?

AI is not coming for your team. It is coming for the worst two hours of their day. But it is not an employee, and treating it like one is how companies end up with a pile of spend and nothing on the P&L.

Host Nick Dreyfus, VP of Business Development at i-NETT, argues the replacement narrative is wrong and the real risk is the opposite one. Not that AI takes the job, but that nobody is checking its work.

In this episode:

• The one in eight test: apply your hiring standard to a system that is confidently wrong, and your deployment plan writes itself
• The invoice that hit ten thousand percent: what a 5% price increase becomes with no guardrails, and why your client catches it before you do
• Why MIT found roughly 95% of AI deployments show no return, and why it is not the technology's fault
• What an agentic engineer does all day, for a CEO with no interest in becoming technical
• The rule that changes everything: do not let AI think, let it choose
• Why five hours saved in accounting is worth almost nothing, and five hours saved in sales takes a 90 day close to 45

This episode answers:

• Is AI going to replace my employees?
• Why do most AI projects fail to produce any ROI?
• What is an AI agent, and who should be building ours?
• Do we hire AI engineers or bring in a partner?
• What AI is already running in my company that I do not know about?

You have three choices and one is not real. Do nothing, which is not neutral, because your team is already using AI on personal accounts. Build your own agentic engineering team. Or bring in a partner who has one.

Start here. Ask your team what AI they use and what they put into it. Then ask who would catch it if it got something wrong. If nobody has an answer, you found your starting point.

Free AI usage policy template you can hand your team Monday: https://i-nett.ai/ai-usage-policy A starting point, not legal advice. Have your attorney review it.

Book a time with Nick: https://meetings-na2.hubspot.com/ndreyfus Our position on AI: https://i-nett.ai Managed services and cybersecurity: https://i-nett.com Community Technology Fund: https://nonprofit.i-nett.com LinkedIn: https://www.linkedin.com/in/nicholas-dreyfus/ [email protected] | (858) 337-2866

Chapters: 00:00 The employee who makes things up one in eight times 03:00 Why confidence is the danger, not error 08:00 The invoice that hit ten thousand percent 13:00 Why 95% of AI deployments fail 18:00 What an agentic engineer actually does 23:00 Not all saved hours are worth the same 28:00 Your three choices, and one is not real

Your team focuses on growth. Our team makes that happen. New episodes weekly.

Educational only. Not legal or financial advice. Statistics cited are third party research and results vary.

#AI #AIAgents #ArtificialIntelligence #SmallBusiness #CEO #BusinessStrategy #ManagedServices #MSP #AIStrategy #DigitalDilemma

98% of executives say their AI investment is paying off. MIT found that 95% of AI pilots cannot prove any return at all. Both numbers are real.

Today we show you where the money actually goes.

Tesla sells a feature called Full Self-Driving Capability. On their own website, in the same paragraph, they tell you "the currently enabled features do not make the vehicle autonomous." Nobody lied. The name just did the work. "Artificial intelligence" is doing the exact same thing to your budget.

Host Nick Dreyfus breaks down why most companies cannot answer the one question that matters: did the AI you bought ever hit your P&L?

No hype, no vendor spin, no Silicon Valley bias.

In this episode:

• Why AI is not artificially intelligent, it is a prediction engine, and why that distinction explains every hallucination and every failed pilot
• Why "we saved 10 hours a week" is not a financial result, and the exact math a CFO should run before believing it
• Why not all saved hours are worth the same: free up accounting and you get accounting, free up sales and you get pipeline
• The one sentence to put in your approval document before you spend a dollar on any AI project
• The two questions to ask any AI vendor before you sign anything
• A free exercise you can run with your team Monday morning that costs nothing and tells you more than any vendor demo

This episode answers:

• Is AI actually artificial intelligence, or something else entirely?
• Why do 95% of AI pilots fail to show measurable ROI?
• When should a business kill an AI project instead of doubling down?
• What belongs on a CFO's dashboard for AI spend?

The line to remember: you cannot have AI thinking, you can have AI flagging. The humans do the thinking. The machine handles what is repetitious and flags what is out of the ordinary. Companies that respect that line are the ones whose numbers eventually show up.

Book a time with Nick: https://meetings-na2.hubspot.com/ndreyfus

Learn more:

• Managed services and cybersecurity: https://i-nett.com
• Our position on AI: https://i-nett.ai
• Community Technology Fund: https://nonprofit.i-nett.com

Connect with Nick Dreyfus:

• LinkedIn: https://www.linkedin.com/in/nicholas-dreyfus/
• [email protected] | (858) 337-2866

Chapters:

00:00 Two numbers that cannot both be true

02:30 What AI actually is, and is not

07:00 Why saved hours are not a return

13:00 Where to point AI first

18:30 When to kill an AI project

22:30 Two questions before you sign

25:30 The Monday morning test

Your team focuses on growth. Our team makes that happen.

New episodes weekly.

Educational only. Not financial or investment advice. Results vary by environment.

#AI #ROI #CFO #SmallBusiness #ArtificialIntelligence #ManagedServices #MSP #BusinessStrategy #DigitalDilemma #AIStrategy

An advanced OpenAI model was placed inside a restricted cybersecurity test. It found a way out, reached the internet, and compromised Hugging Face infrastructure while pursuing answers to the test.

It was not told to attack Hugging Face.

It was told to complete an objective.

That distinction should get the attention of every CEO and CFO.

In Episode 40 of The Digital Dilemma, Nick Dreyfus explains why this incident is not a science-fiction story about an evil machine. It is a business warning about speed, persistence, access, and what happens when a powerful system pursues a goal without enough boundaries.

The next generation of cyberattacks will not need to get everything right on the first attempt. AI can test one path, fail, change its approach, and try again. One firewall, one antivirus product, or one green backup checkbox is no longer a business-continuity strategy.

Nick connects the OpenAI and Hugging Face incident to the 2010 Flash Crash and Knight Capital's $460 million automated trading failure. He also shares firsthand stories from the field:

• A respected IT provider reported successful backups, but the latest recoverable copy was more than six months old.

• A manufacturer automated a process, shipped approximately 100,000 affected products, stopped taking new orders, and spent nearly seven months recovering.

• Mid-sized companies believed enterprise equipment made them secure while unsupported servers and excessive permissions left openings attackers could exploit.

This episode is built for leaders running companies where payroll, production, billing, accounting, sales, and customer relationships all depend on technology.

CHAPTERS

00:00 It was a matter of time

01:35 What OpenAI says happened

04:30 Why the model's objective matters

06:20 The Flash Crash and Knight Capital lesson

08:45 What i-NETT finds after an attack

11:05 A green backup checkbox is not recovery

14:10 How one automated mistake reached 100,000 products

16:45 Why security is the last place to buy the cheapest checkbox

19:10 Why a good MSP may still be unprepared for AI

21:15 The new AI responsibility inside every business

22:45 Why cyber insurance is not the strategy

24:20 Three questions every CEO and CFO should ask this week

26:20 The leadership decision that cannot wait

THE THREE QUESTIONS

1. What are we protecting, and where are our unsupported systems or excessive permissions?

2. Can you prove that our recovery works, and how long would it actually take?

3. What authority are we giving AI, what data can it access, and who is watching it?

Policy without testing is hope.

Security without monitoring is hope.

Backups without recovery are hope.

Hope is not a business-continuity plan.

Learn how i-NETT approaches responsible AI adoption:

https://i-nett.ai

Explore strategic managed IT, cybersecurity, and business continuity:

https://i-nett.com

Connect with Nick Dreyfus on LinkedIn for practical conversations about AI, cybersecurity, managed services, and the decisions facing business leaders right now.

Follow The Digital Dilemma on Apple Podcasts, Spotify, or wherever you listen to podcasts. If this episode makes you think differently, share it with a CEO or CFO who needs to hear it.