There was a time when IT could get away with being the department that kept the lights blinking, the mainframe humming, and the help desk queue from catching fire. That world is gone. Technology has never been more powerful, but power alone does not create value. In many enterprises, it creates expense, fragmentation, risk, and a very expensive illusion of progress.
That gap between technology investment and business value was the center of a recent episode of the Amplitude of Tech podcast, in conversation with Bill Fandrich, Chief Healthcare Officer at Pellera, former CIO of Blue Cross Blue Shield of Michigan, and founder of M1 Nexus Advisors. Across 40 years spanning mainframes, ERP, cloud, and now AI, Fandrich has landed on one conclusion that has outlasted every hype cycle, and he walked host Shawn Cordner through the governance frameworks, value analysis, and three-tier investment model he used to put it into practice.
| Quick Answer AI investments stall when organizations buy tools before they fix integration, data, and governance, according to Bill Fandrich, a 40-year healthcare technology leader and former Blue Cross Blue Shield of Michigan CIO. His core position, formed early in his career, is that “it’s not about the technology.” The practical fix is a value analysis rather than a standard ROI calculation, a responsible AI framework that defines guardrails before tools get deployed, and an honest accounting of the integration and data work that has to happen first. |
The Old IT Problem Never Really Went Away
The classic complaint about IT was that it was too far removed from the business. The business asked for outcomes. IT asked for requirements. Somewhere between those two conversations, ambition got translated into tickets, projects, delays, change orders, and budget overruns.
Fandrich entered the workforce before PCs existed, when enterprises ran on mainframes and mid-range systems and every system had to be designed from scratch. What struck him was not the technology. “I was awed at how much, quite frankly, how challenging it was for companies to get value out of those investments and how many of those investments failed,” he said. The gap between business strategy and technology was enormous, and the professionals who closed it were the ones who could wear a business hat.
Watching peers implement the same ERP platform at different companies with wildly different results sealed it. “I kind of coined my own phrase back then, it’s not about the technology,” he said. The vocabulary has changed since. Leaders now debate AI models, cloud architectures, CCaaS platforms, and data fabrics. The leadership challenge has not.
This is where strategic technology advisory matters. Organizations need more than procurement support or implementation horsepower. They need a way to connect architecture, vendor decisions, operating models, and executive priorities into one coherent plan. That is the space where Amplix helps enterprises plan, evaluate, implement, and manage technology more effectively.
Cloud Was Supposed to Simplify Everything
Cloud may be the best example of the gap between promise and operating reality. Fandrich saw infrastructure-as-utility coming in 2008, before AWS had a formal brand, after a friend told him to go look at how Amazon managed its own infrastructure. “The light bulb went off,” he said. “The world is gonna be shaped, infrastructure by services, but if you want to treat it as a service, you better define it as a service, not as a tech stack.”
By the time cloud was something you could buy, he had already built the governance around it. Others did not. “Those who said, just turn on the cloud, we see now the consequences,” he said. Cloud required new architecture, new financial controls, new security models, new skills, and new operating discipline. Many organizations treated it as a destination when they should have treated it as a design challenge.
The same thing happened with packaged software. Fandrich argues the shift to packages cost enterprises something real: “We lost the value of engineering and architecture. We lost the value of people on site who really understood how the business works.” Teams became package installers while the ecosystem grew more complex, more fragmented, and more expensive. His standing skepticism: “I’ve never seen yet this promise of new technology is going to take out 50% of your IT costs. Every new technology comes with trade-offs.”
Pilot the Strategy Before You Pilot the Technology
AI feels bigger than previous waves because it does not sit neatly inside the technology department. It shows up in customer experience, claims processing, security operations, finance, procurement, software development, HR, and the contact center. Fandrich frames it accordingly: “to me more not technology, it’s a business capability. It’s everywhere. It’s not a separate thing.”
That distinction has consequences. A technology initiative can live in a roadmap. A business capability has to live in workflows, operating models, controls, incentives, and measurable outcomes. “If you just think you can bolt AI onto what you have and magic’s gonna happen, you’re gonna fail,” he said. “It’s a massive change management.”
His sharpest warning is aimed at how enterprises build AI strategy in the first place. “If your first step is to look at tech vendors when you’re building a strategy, you already failed,” he said. Too many organizations are letting vendors drive strategy instead of building internal capability. The pilot problem follows from the same mistake. “Even if you do a pilot, you should be piloting all these other things, not the solution,” he said. What you are proving out is the engine: governance, process redesign, data strategy, infrastructure, tech debt.
He ran cloud exactly that way. When his team brought him a multimillion-dollar procurement list for the first cloud workload, he refused it. “How about this? You get a hundred thousand dollars and you got three months to do something,” he said. “You’re not buying anything. We’re trying to prove out the model.” Ten years later, the organization had made tens of millions in cloud investment and could quantify the return, because the principles came before the spend.
The Integration Work That Made AI Possible
The clearest proof point in the conversation is not an AI project at all. It is the API transformation Fandrich’s team ran at Blue Cross Blue Shield of Michigan, a year of planning followed by three years of execution across a sprawling legacy estate.
The result: a rebuilt integration stack spanning 6,000 applications and hundreds of thousands of external connections, now running roughly five billion API calls a month. When AI arrived, the connective tissue was already there. “If we didn’t do that, we would never got one piece of AI done of any strategic value,” he said.
His conclusion is blunt. “Your integration, your data, at a minimum, those have to be dealt with. And you better get good at that because that is your business.” On AI as a shortcut around that work, he allows that the tools genuinely accelerate discovery and modernization, then adds the caveat: “There’s no skipping steps here.”
A Three-Tier Model for AI Investment
To keep executive conversations out of the weeds, Fandrich separates AI investment into three tiers.
- Productivity tools. Workforce enablement, run as a formal program with training and certification. Choosing between one assistant and another “does not solve your strategic goal,” he said, but an enabled workforce still matters.
- Business-led initiatives. AI applied to existing functions: SDLC, DevSecOps, QA, deployment, product innovation, and orchestration across third parties. Important work, measurable the way projects have always been measured.
- Big bets. Reimagining how a process works end to end. His example: a member walks into a hospital and relevant clinical information, coverage, cost, and care options surface for the physician immediately, with no one picking up a phone to argue about what is covered. “That is what AI is intended to do,” he said. These require looking at the business from the outside in and bringing stakeholders into the redesign.
His warning for organizations that only ever fund the first two tiers: “You’re probably going to not like the economic answer over time if that’s all you ever done.”
Responsible AI as the Thing That Enables Speed
Three and a half years ago, before most boards were asking about AI, Fandrich authored a responsible AI framework with his executive team. The logic was pragmatic. The train was not stopping, so the question became what guardrails to build.
The framework ran to roughly 12 principles tied directly to the organization’s mission, with trust at the center. Bias, ethics, and transparency had to be addressed proactively before anything deployed. A human stayed in the loop on care decisions. The board debated the principles directly, which he counted as a win in itself.
That framework is what made broad experimentation possible. The organization went from no AI tooling to more than 1,500 employees using AI tools regularly, all inside the guardrails. “Yeah, we can give you tools, but you’re not gonna pick any tool,” he said. The same pattern had worked earlier in analytics, where 60 to 70 concurrent sandboxes gave business teams 90 days to prove out a vendor without touching sensitive data or production systems.
Fandrich pushes back hard on the idea that governance slows innovation. Having spent time with hyperscalers, his read is the opposite: “They are the most disciplined organizations you’ll ever see in your life.” His summary line: “Speed and innovation comes with a higher degree of discipline.” And a clarification worth keeping: “Discipline doesn’t mean control, by the way.”
Value Analysis, Not an ROI Calculation
ROI is a measurement. Fandrich argues what AI business cases actually need is a value analysis, which is a different exercise.
A value analysis starts from strategy. Where does the organization need to be in five years, and how does it win there? From that, you map every function you perform today against the functions you will need then, including ones that do not exist yet. Some will run at a fraction of current cost. Others will cost ten times more, or appear on the books for the first time. “If you’re doing true value analysis, you’ll be able to answer those questions,” he said.
He is direct about where that leaves budgets. If an organization genuinely transforms, the equivalent of today’s IT spend will likely double or triple, and the offsetting savings have to come from somewhere identifiable. Pretending otherwise is how AI programs lose credibility with finance.
He also draws a line between running the business and setting strategy. Driving out efficiency should be continuous and built into the culture. “Admin savings is not a strategy,” he said. “That’s called staying in business.” And on economics as a driver: “Economics is a byproduct of everything else you’ve done. It’s a barometer to say, am I in line with what I expected? It’s not the driver.”
On vendor lock-in, his advice is to optimize for reversibility rather than for picking the eventual winner. Asked by his board which AI vendors to commit to, he told them that whatever he recommended would be wrong, and that some of the right partners probably did not exist yet. The posture he recommends: “Good enough for now, but not good enough to make a strategic commitment to.” The decision that matters is which dependencies would be painful and expensive to unwind.
The CIO as Translator
The best technology leaders translate. They translate business strategy into architecture, architectural constraints into executive risk, vendor promises into operational tradeoffs, technical debt into financial exposure, and innovation into adoption and measurable value.
Fandrich learned that early at EDS, where business leaders began treating him as a peer once he understood their work. “Business people accepted me because they actually saw me as a peer and as someone who had knowledge to give on the business versus a technology just trying to get requirements,” he said. The CIO who can explain how cloud governance affects margin, how AI affects labor strategy, and how security posture affects enterprise value earns a different seat.
That does not mean watering down the technical detail. Sometimes the truth is that a migration will take longer because the integration layer is a wreck, that a promising AI use case is not ready because the data underneath it is untrustworthy, or that the company is overspending because nobody has cleaned up contracts, circuits, SaaS licenses, or cloud consumption in years. Good technology leadership is often disciplined cleanup, and disciplined cleanup funds the next strategic move.
He has one more warning for leaders handing capability to outside partners, a line he has used with boards for two decades: “Don’t ever outsource your brain.” Vendors should enable a vision the organization owns.
On the labor question, he does not buy the apocalypse. “I don’t believe the hype that says fifty percent of jobs are gonna disappear.” His own track record at scale: functions eliminated through automation, five new functions created in their place. “At the end of the day, we’re not smaller, we’re different. Is that value? The question is value.”
What Healthcare’s Margin Pressure Reveals
Fandrich’s industry view sharpens the general argument. Healthcare margins are compressed from every direction: affordability pressure, a shift from commercial business toward government programs with different pricing policy, and the end of subsidies that propped up parts of the system. Margins that were already thin at one to two percent have not been met in a long time.
The technology diagnosis is familiar. “We bolted on so many things in the healthcare ecosystem with quadruple the number of vendors, quadruple the number of systems, and it’s more disjointed than ever,” he said. “Each individual item, I can promise you, had value, but when you assemble them, it has a negative value to the outcome.”
The opportunity sits in the rework, the back and forth, and the fraud. Streamlining administration will not fix healthcare policy, but it can slow cost growth enough to fund innovation that reaches patients. On interoperability, he is realistic: it gets solved when payers and providers decide together to change how reimbursement works, and the current chassis was designed 40 years ago.
For organizations thinking about the customer-facing side of that equation, it is worth understanding how CCaaS fits into the broader communications and customer engagement stack. The platform decision matters. The operating model matters more.
How Amplix Can Help
The enterprises that win will not be the ones that chase every new wave first. They will be the ones that can absorb new capabilities without losing control of cost, risk, architecture, or trust. That requires the unglamorous work Fandrich keeps returning to: integration, data, governance, value analysis, and leadership that can hold a strategy conversation before a vendor conversation.
Amplix’s capabilities across AI, CX, network and infrastructure, cloud, security, unified communications, and mobility reflect the reality that technology decisions rarely stay inside one lane. Contact Amplix today to separate what is trendy from what your organization actually needs.
Frequently Asked Questions
Why do so many AI investments fail to deliver business value?
Organizations buy AI tools before resolving integration, data quality, workflow, and governance. Without that foundation, AI projects hit a wall regardless of the model or vendor selected. The failure is rarely the technology itself.
What is the difference between a value analysis and an ROI calculation?
An ROI calculation measures whether an investment paid off. A value analysis starts from strategy and maps the functions an organization performs today against the functions it will need in five years, including costs that do not exist yet. ROI is a barometer. Value analysis is the plan.
What is a responsible AI framework, and what should it include?
A set of guiding principles, tied to organizational mission, that define how AI gets deployed before any tools are selected. Core components include bias and ethics review, transparency requirements, human-in-the-loop rules for consequential decisions, data exposure limits, and approved tooling. Fandrich’s framework ran to roughly 12 principles and was debated directly with the board.
How should enterprises categorize AI investments?
Fandrich uses three tiers: productivity tools for workforce enablement, business-led initiatives that apply AI to existing functions, and big bets that reimagine a process end to end. Only the third tier reshapes the value chain, and organizations that fund only the first two tend to struggle with the economics over time.
How do you avoid AI vendor lock-in?
Optimize for reversibility rather than for picking the eventual market leader. Build variability into the architecture and operating model, and assess each dependency by how painful and expensive it would be to replace. Treat current tools as good enough for now rather than as a strategic commitment.
Does AI governance slow down innovation?
The opposite, in Fandrich’s experience. Guardrails are what allowed his organization to move from no AI tooling to more than 1,500 regular users without exposing sensitive data. Sandboxes with defined controls and time limits let business teams experiment safely, and speed comes from the enablers being in place.