Bits & Bytes & Business

Can AI companies police themselves? The Great Fire of London, a $350 Chicago lab, and Upton Sinclair’s stockyards all point to the same answer. In Episode 2026.29 of Bits & Bytes & Business, we trace how fire insurers created the first independent inspectors, why the electrical industry got one before disaster while the meatpackers got one after a scandal, and what that means for the voluntary White House AI accord signed September 29, 2026.

The accord, as reported, promises outside auditors but carries no penalties, no government enforcement role, and no required publication of audit results. We argue that letting an industry police itself is a bad idea, and that a better system makes doing the right thing the profitable thing: publish audit findings, put money behind the risk with AI liability insurance, and add a legal floor of mandatory testing and incident reporting. We also give the strongest case for the voluntary approach a fair hearing.

📖 Read the companion post: https://www.linkedin.com/pulse/inspector-skin-game-joseph-h-schuessler-k93jc/
🎬 Watch the video overview: https://youtu.be/BsgrYhZeQR4?si=WIGrSvHxMygBNdVe

Chapters
0:00 Welcome back and the Episode 28 recap
1:45 Required disclosure and the position we take
2:39 The mystery: 13,000 houses gone in 1666
4:09 Nicholas Barbon and the first fire insurance office
5:43 Chicago, 1893: the World’s Columbian Exposition
7:15 William Henry Merrill and the $350 lab that became UL
8:32 Upton Sinclair, The Jungle, and the Meat Inspection Act
10:19 Inspector before the disaster vs. inspector after the scandal
11:07 The September 29, 2026 voluntary AI accord
12:16 What is missing: no penalties, no published audits
14:04 The FTC probe and early warning signs
15:27 Gartner Hype Cycle for AI Governance 2026: AI Liability Insurance
17:04 Why AI liability insurance is under 1% penetration
18:40 AI washing and the Gartner governance survey
20:17 The case for voluntary self-policing
22:02 Why the UL analogy does not rescue the accord
22:31 Bipartisan skepticism of self-policing
23:42 Full circle: back to London, 1666
24:36 The closing question and next week’s topic

The question for the comments: If no insurer will cover an AI system, should it be legal to deploy it?

👍 If this connected some dots for you, please like, follow, subscribe, and share it with someone in risk, compliance, insurance, or IT governance.

Disclosure: Anthropic, the maker of Claude, is among the reported signatories of the accord discussed in this episode. Claude helped produce the research materials for this series. This episode was developed with the assistance of AI tools, including Claude, Gemini, and NotebookLM, guided by detailed human-authored prompts, editorial review, and external sources including Gartner Hype Cycle reports, NIST, and peer-reviewed literature.

Sources: Gartner, Hype Cycle for AI Governance, 2026 (G00854164); UL Research Institutes, Our History; Al Jazeera and NPR coverage of the September 29, 2026 AI accord; Upton Sinclair, The Jungle (1906).

Next week: With the midterm elections just around the corner, how do you know that what you are seeing and hearing is real, and what happens when a lie is more profitable than the truth?

#AIGovernance #AIRegulation #AIInsurance #CorporateSocialResponsibility #RiskManagement #TechPolicy #UnderwritersLaboratories #BitsBytesBusiness

Who pays when an AI data center moves in next door? In this episode of Bits & Bytes & Business, we trace the electric bill, the water draw, and the noise back to a flooded 1760s mine, a French engineer, and an IBM physicist, and find out why every AI model has to dump its heat somewhere.

The chain runs from James Watt's repair of a mine-pump model to Sadi Carnot's rule that engines must reject heat to Rolf Landauer's 1961 finding that erasing information releases heat to the AI facilities now being built near homes. Then we get practical.

What we cover:

• Why AI data centers need so much electricity, water, and cooling
• Lawrence Berkeley National Laboratory's estimate that U.S. data centers used about 4.4% of national electricity in 2023, and could reach 6.7% to 12% by 2028
• A grid market monitor's estimate of about $29 billion in added costs for ratepayers, and the industry-commissioned study that found no clear national effect on household rates
• Water use and drought-area siting, and the Manassas, Virginia noise case
• Why Gartner's Hype Cycle places space-based data centers at Low benefit, under 1% penetration, and Embryonic, and why heat is hard to reject in a vacuum
• Why governors in both parties, in New York and Texas, have paused data center permits
• Three proposals: full-cost tariffs, independent measurement of power, water, and noise, and community say before the deal

This episode takes a clear position, and we give the other side a fair hearing. Figures marked as estimates are estimates, and sources are in the companion blog post.

Today's question: if the neighbors of a data center carry the noise, the water draw, and the rate increases, should they hold a veto over the project, or does the national race for AI outweigh the town? Tell us in the comments.

Next week: a novelist who aimed at the public's heart and hit its stomach by accident, and what happened the last time an industry promised to police itself.

Like, follow, subscribe, and share with someone who has a data center proposal in their county.

This episode was developed with the assistance of AI tools, including Claude, Gemini, and NotebookLM, guided by human-authored prompts, editorial review, and external research sources.

At 5:16 and 11 seconds on the evening of November 9, 1965, a protective relay near Niagara Falls detected power flow slightly above a preset threshold and did exactly what it was built to do. It tripped the line. Within ten seconds, that single correct decision cascaded into one of the largest blackouts in American history, leaving thirty million people across the northeastern U.S. and Ontario in the dark.

This episode picks up right where last week's shipping container left off — once AI agents are connected and speaking a shared language, what happens the moment that language has to stop, immediately, on command? We trace the answer from the 1965 blackout, through the 1968 formation of NERC, through Netflix's 2011 "circuit breaker" pattern in distributed software, landing on the real technology behind what everyone casually calls the "AI kill switch": AI guardrails.

We dig into why AI guardrails already sit at 5–20% market penetration on Gartner's Hype Cycle, further along than most of the governance technologies we've covered this season, why a safety mechanism working exactly as designed is not the same as a system working safely, the real tension between guardrails too permissive and guardrails too rigid, and the genuinely unresolved question of who's liable when a correctly functioning guardrail still causes harm. We also unpack what this means for the next generation of IT and business leaders training in the Dr. Sam Pack College of Business classroom right now.

This episode sets up next week's topic: the environmental and infrastructure cost of AI at scale — the power, water, and cooling demands straining the same grid that failed in 1965.

🔔 Like, follow, and subscribe so you don't miss next week's episode.
💬 Tell us in the comments: when an AI guardrail trips exactly as designed and the resulting action still causes real harm, who bears responsibility — the engineers who set the threshold, the organization that deployed the system, or is that simply an acceptable cost of operating complex, interconnected systems at scale?

#AIGuardrails #AIKillSwitch #AgenticAI #ITGovernance #EnterpriseAI #AIStrategy #TarletonCOB #HypeCycle

In 1956, a trucking executive who hated loading docks solved global cargo chaos with one absurdly simple idea: standardize the box, not what's inside it. That same principle is quietly solving the biggest problem in enterprise AI right now — how do agents built by different vendors actually talk to each other?

This episode picks up right where last week's telephone switchboard left off: Strowger's switch determined who gets connected to whom, but said nothing about what language the two ends of that connection actually speak once connected. We trace that gap from Malcolm McLean's 1956 shipping container, through ISO's global standardization of container dimensions, through 1970s TCP/IP, landing on the standard quietly solving the same problem for AI agents: the Model Context Protocol.

We dig into why MCP already has OpenAI, Google, and Microsoft on board barely a year after Anthropic published it, why "standardize the interface, not the content" is the same principle that built the modern internet, the real security risks Gartner flags in poorly governed MCP implementations, and why nobody's yet sure whether MCP becomes universal infrastructure or gets eclipsed by whatever comes next. We also unpack what this means for the next generation of IT and business leaders training in the Dr. Sam Pack College of Business classroom right now.

00:00 The 1950s waterfront: total chaos

00:21 Break-bulk cargo: a daily game of physical Tetris

01:43 The same chaos, now inside modern AI server rooms

02:51 Callback: Strowger's switchboard (Episode 2026.25)

03:56 The switchboard's limit: routing without language

04:58 The enterprise AI interoperability crisis

05:47 Why traditional API integrations are so brittle

06:44 Malcolm McLean and the loading dock problem

07:41 1956: the steel box solution

08:02 "Standardize the interface, not the contents"

09:37 ISO codifies the standard as global infrastructure

10:05 The next domino: computer networks

10:52 Cerf and Kahn and the birth of TCP/IP

12:28 The 2020s AI integration nightmare

13:25 November 2024: Anthropic publishes MCP

14:16 Inside MCP's client-server architecture

15:28 Runtime discovery vs. design-time integration

17:18 The security nightmare: automated malware delivery?

19:09 Gartner's fix: read-only tools and the MCP gateway

21:04 Continuous reauthentication and human-in-the-loop approval

22:00 Classroom deep dive: building and breaking the fortress

24:14 The economic trap: technical debt

26:07 The scope-narrowing trap

27:10 The existential risk: 1-5% market penetration

27:58 Enter the rival: Agent2Agent (A2A)

30:58 Who's liable when MCP fails?

32:22 The debate: should protocols require disclosure?

34:19 Next week's teaser: the AI kill switch

This episode sets up next week's topic: AI guardrails, the real technology behind what everyone casually calls the "AI kill switch" — anchored in a 1965 night in New York City when an automated safety mechanism worked exactly as designed and still triggered one of the largest blackouts in American history.

🔔 Like, follow, and subscribe so you don't miss next week's episode.💬 Tell us in the comments: should enterprises be legally required to disclose which AI communication protocol version they're running in any system handling sensitive data — the same way food packaging discloses ingredients — or is that a compliance burden nobody's actually asking for?

#ModelContextProtocol #AgenticAI #ITGovernance #EnterpriseAI #AIStrategy #TarletonCOB #HypeCycle #DataCommunications

In the 1880s, a Kansas City-area undertaker became convinced a telephone switchboard operator was deliberately misrouting his calls to her husband's rival funeral home. His response wasn't a lawsuit. It was an invention: the first automatic telephone exchange, nicknamed the "girl-less, cuss-less" system.

This episode picks up right where last week's tree-shaped org chart left off — McCallum's tree could tell you who reported to whom, but it couldn't coordinate dynamic, real-time connections between distant parts of an organization. We trace that gap from Almon Strowger's 1891 switchboard, through circuit switching, through 1960s-70s packet switching, through 2014's Kubernetes and container orchestration, landing on Gartner's next stage of agentic AI evolution: agent orchestration.

We dig into why agent orchestration sits at under 1% market penetration despite a High benefit rating, why "agent sprawl" is quietly creating technical debt and security risk across enterprises deploying agents without a coordinating layer, why most orchestration platforms today lock customers into a single vendor's walled garden, and why the market's fragmented terminology (management plane, agentic fabric, control tower, trust center) still needs unifying. We also unpack what this means for the next generation of IT and business leaders training in the Dr. Sam Pack College of Business classroom right now.

This episode sets up next week's topic: the Model Context Protocol — the language AI agents and tools actually speak to each other once orchestration connects them.

🔔 Like, follow, and subscribe so you don't miss next week's episode.
💬 Tell us in the comments: should enterprises hold off on scaling AI agent deployments until a genuine cross-vendor orchestration standard exists — or is waiting for standards just an excuse to fall behind competitors moving now?

#AgentOrchestration #AgenticAI #ITGovernance #EnterpriseAI #AIStrategy #TarletonCOB #HypeCycle #DataCommunications

In 1855, an American railroad executive faced a company too large for any one person to watch. His solution wasn't a memo — it was a diagram shaped like a tree, roots at the bottom, branches fanning upward through hundreds of station agents and conductors he'd never meet. It's considered the first formal org chart in American business history, and it solved a problem enterprises are quietly failing to solve again right now: how do you supervise a team when you can't watch every member directly?

This episode picks up where last week's Medici signet ring left off — proving one agent's identity is easy compared to supervising a hundred of them at once. We trace the throughline from Daniel McCallum's 1855 tree-shaped organizational chart for the New York and Erie Railroad, through 20th-century command-and-control dashboards, and land on Gartner's newest and least mature agentic AI category yet: human-agent collaboration workspaces — shared digital environments where people define outcomes, delegate to teams of AI agents, and approve results through structured handoffs and guardrails.

We dig into why this technology sits at under 1% market penetration despite a High benefit rating — the least mature category we've covered all season — why chat-based copilots don't provide a real "system of work" for supervising agent teams, what an "agent steward" role actually requires, and why cost attribution for agent-driven work is still an unsolved problem for most enterprises. We also unpack what this means for the next generation of IT and business leaders training in the Dr. Sam Pack College of Business classroom right now.

This episode sets up next week's topic: agent orchestration — what happens the moment one AI agent hands work directly to another, with no human in that specific loop at all.

🔔 Like, follow, and subscribe so you don't miss next week's episode.
💬 Tell us in the comments: as AI agents take on more of the "team," should the manager of a mixed human-agent workspace be evaluated and compensated differently than a manager of an all-human team — or is management just management, regardless of who's on the branch?

#HumanAgentCollaboration #AgenticAI #ITGovernance #EnterpriseAI #AIStrategy #TarletonCOB #HypeCycle #FutureOfWork

Before a Medici factor could transact business three hundred miles from Florence, he carried something more valuable than gold: a personal signet ring, pressed into wax, proving he was authorized to act in the bank's name. It's the same problem enterprises face today with AI agents — how do you trust something acting on your behalf when you can't watch it work?

This episode picks up right where last week's guild wardens left off, tracing identity and accountability forward from medieval merchant seals, through 14th-century bills of exchange, through modern public key infrastructure, and landing on Gartner's newest governance priority: AI agent identity — the unique digital credentials that let organizations know exactly which AI agent did what, on whose authority, and with what access.

We dig into why AI agent identity already sits at 5–20% market penetration on Gartner's Hype Cycle — ahead of last week's guardian agents — why "shadow AI" (agents quietly operating under borrowed human logins) is the identity failure hiding behind most rogue-agent headlines, what modern machine identity and access management (IAM) actually requires, and why this is a compliance and liability question your organization can't put off. We also unpack what this means for the next generation of IT and business leaders training in the Dr. Sam Pack College of Business classroom right now.

This episode sets up next week's topic: human-agent collaboration workspaces — once you know exactly which agents are operating in your enterprise, how do humans actually work alongside a whole team of them?

🔔 Like, follow, and subscribe so you don't miss next week's episode.
💬 Tell us in the comments: should an organization be held legally liable for actions its AI agents take under a shared or borrowed human credential — the same way a merchant house was once liable for a factor acting under its seal?

#AIAgentIdentity #AgenticAI #ITGovernance #IAM #CyberSecurity #EnterpriseAI #AIStrategy #TarletonCOB #HypeCycle

Before a bolt of silk could legally leave Lyon, an independent inspector had to check it, and forbidden from having any stake in the goods, sign off with an official mark. That guild warden, it turns out, is the real ancestor of this week's technology, not the automated looms we covered back in Episode 1.

This episode picks up right where the Vaucanson's-duck story left off: as AI agents proliferate across the enterprise, human oversight can't keep pace, so who, or what, watches the AI? We trace a four-century arc from guild wardens to Luca Pacioli's self-checking double-entry ledger, through 19th-century bank examiners, to Sarbanes-Oxley's mandated internal controls, landing on Gartner's newest and least-mature category: guardian agents, AI systems built specifically to monitor, constrain, and enforce policy on other AI agents.

We dig into where guardian agents sit on Gartner's Hype Cycle (still near the Innovation Trigger, under 1% market penetration, trailing the very technology they're meant to govern), the paradox that guardian agents must themselves be governed, why most can't yet do real-time blocking, and the vendor-independence question every CIO should be asking before their next AI procurement. We also unpack what this means for the next generation of IT and business leaders training in the Dr. Sam Pack College of Business classroom right now.

This episode sets up next week's topic: AI agent identity. How do you even know which agent you're watching?

🔔 Like, follow, and subscribe so you don't miss next week's episode.
💬 Tell us in the comments: should regulators require every guardian agent to be built and operated independently of the AI platform it oversees, or is that level of separation unrealistic for how enterprises actually buy AI?

#GuardianAgents #AgenticAI #ITGovernance #AITRiSM #EnterpriseAI #AIStrategy #TarletonCOB #HypeCycle

In 1739, a mechanical duck fooled all of Paris into thinking it could digest grain. It couldn't — the "autonomy" was a hidden compartment and a well-timed reveal. Three hundred years later, boardrooms are staring at a strikingly similar promise: AI agents that perceive, decide, and act on their own.

Welcome back to Bits & Bytes & Business for Fall 2026. In this season-opening episode, we trace an unlikely lineage — from Jacques de Vaucanson's fraudulent Digesting Duck and his very real automated loom, through Jacquard's punched cards, Babbage and Lovelace's Analytical Engine, and Hollerith's tabulators — straight into 2026's most hyped and least understood technology: agentic AI.

We dig into where agentic AI actually sits on Gartner's brand-new Hype Cycle (spoiler: right at the Peak of Inflated Expectations), why only 17% of organizations have deployed AI agents despite 42% planning to within a year, what "agent washing" means for your procurement team, and why CFOs should be worried about token costs the same way they once worried about cloud sprawl. We also unpack what this means for the next generation of IT and business leaders training in the Dr. Sam Pack College of Business classroom right now.

This episode sets up next week's topic: guardian agents — AI built to police other AI, and what it means when your organization needs a robot to watch the robot.

🔔 Like, follow, and subscribe so you don't miss next week's episode.
💬 Tell us in the comments: when your organization's first AI agent makes a costly autonomous mistake, who should be accountable — the vendor, the IT team, or the business leader who approved its goals?

#AgenticAI #ArtificialIntelligence #ITGovernance #EnterpriseAI #BusinessTechnology #AIStrategy #TarletonCOB #HypeCycle

Is teaching excellence a performance art, or is it a systems engineering problem?

In this episode of Bits & Bytes & Business, we explore how the educational paradigm has shifted away from charismatic live lectures toward the rigorous, evidence-based discipline of instructional architecture. Discover how Quality Matters (QM) and ACUE frameworks are fundamentally transforming the digital classroom, significantly reducing DFW (Drop, Fail, Withdraw) rates, and producing graduates who are immediately ready for the complex demands of the enterprise tech market. We break down the concept of the "Invisible Architect," analyze Cognitive Load Theory, and explain why building a highly aligned online course requires the exact same systems-thinking methodology as architecting an enterprise IT infrastructure.

Chapters

• 0:00 - Introduction: The End of "Sage on the Stage"
• 1:13 - The Mission: Defining Instructional Architecture
• 3:11 - The Invisible Architect: Systems vs. Performance
• 3:48 - The App UX Analogy: Why Disorganized Teaching Fails
• 5:18 - Academic Freedom vs. Standardization: The Great Debate
• 6:40 - Containerization for Educators: The Docker Analogy
• 8:39 - Scaling Quality Matters (QM): A Global Blueprint
• 10:07 - Designing Mental Models: QM Standards 1 & 2
• 12:10 - The Power of Alignment: QM Standards 3 & 4
• 14:04 - Pedagogy over Novelty: QM Standard 6 (AI & Tech)
• 15:13 - Universal Design for Learning (UDL) & Accessibility
• 17:18 - The Master Reviewer Gauntlet: Professional Credentialing
• 19:14 - Engineering Teacher Behavior: The ACUE Framework
• 20:16 - Impact Data: Reducing DFW Rates in Higher Ed
• 22:31 - Closing the Equity Gap: Dismantling the Hidden Curriculum
• 25:00 - Reclaiming Bandwidth: Structure as Faculty Liberation
• 26:30 - The Human Premium: Empathy in the AI Era
• 27:00 - Systemness & The Meauxmentum Framework
• 30:48 - Scaling Excellence: Texas A&M (TAMUS) Case Study
• 32:45 - Workforce Readiness: Course Design as Enterprise IT
• 35:24 - Digital Forensics Case Study: CPAP & Fitness Trackers
• 37:43 - The Subliminal ROI: Structuring the Strategic Mind
• 39:03 - Summary: Engineering Successful Knowledge Transfer
• 40:14 - Final Thought: Applying Design to Corporate Culture
• 41:50 - Conclusion & Next Week's Teaser

Resources & Tools

• Quality Matters (QM) Rubric & Standards
• Association of College and University Educators (ACUE) Effective Teaching Practices
• Cognitive Load During Problem Solving (Sweller, 1988)
• Tarleton State University MS-IS Program

Ready to future-proof your trajectory and master the fundamentals of the modern IT ecosystem? Pick up a copy of my textbook:

• Launching Your Tech Career

Interested in the arc of technology and what our future might look like? Check out my latest book:

• Emerging Technologies: History, Disruptions, and Future Paradigms

Professional Footer & BioI am Dr. Joseph H. Schuessler, a full-time Professor of Information Systems at Tarleton State University within the Dr. Sam Pack College of Business. As a veteran educator with over 15 years of experience in online instruction, a certified Quality Matters (QM) Master Reviewer, and an ACUE Advanced Certified instructor, my passion lies behind the scenes, building scalable, rigorous, and perfectly aligned digital learning environments.Subscribe to the channel for weekly deep dives into cybersecurity, information systems, and strategic career development.

AI Transparency Disclosure: To ensure the highest level of research and structural integrity, the audio and research materials for this podcast are developed collaboratively utilizing Generative AI tools, specifically NotebookLM and Gemini.