Guardians of AI: How AI Leaders Are Keeping AI Safe, Ethical, Responsible, and True

August 3, 2026 – Published by Authority Magazine

As AI technology rapidly advances, ensuring its responsible development and deployment has become more critical than ever. How are today’s AI leaders addressing safety, fairness, and accountability in AI systems? What practices are they implementing to maintain transparency and align AI with human values? The following interview, originally published by Authority Magazine, features Mike Ehlers’ insights on responsible AI and the future of ethical technology development.

Authority Magazine: Thank you so much for your time! I know that you are a very busy person. Before we dive in, our readers would love to “get to know you” a bit better. Can you tell us a bit about your ‘backstory’ and how you got started?

Mike Ehlers: It started with an Atari 800. My parents brought one home when I was young, and I just started messing around with it, figuring out how to get it to do things. I was hooked pretty quickly. I’ve always been a tinkerer and a problem solver at heart, and what drew me to computers was the ability to take on bigger and bigger problems over time. My career path wasn’t a straight line into HCM or SaaS, though. I started out working on some of the first fully digital mobile phone systems, then went through a couple of internet startups that didn’t make it. I learned a ton from those experiences, but they didn’t exactly work out. Eventually, I fell into benefits administration almost by accident, and once I was there, I was largely hooked. What kept me was the realization that the work I was doing could have a real impact on people, their lives, and their careers. That’s what still drives me today.

None of us can achieve success without some help along the way. Is there a particular person who you are grateful for, who helped get you to where you are? Can you share a story?

A number of people have invested in me along the way, but one who really stands out is Ted Gaty. I worked for Ted at two different points in my career, and one story has always stayed with me. I had been leading a team building a new analytics platform. After weeks of long hours, we released it and I took some time off. While I was out, I got pulled into a meeting with Ted, who was the CTO, and our CEO. I assumed it was a routine roadmap review. Instead, the CEO spent the entire meeting explaining how far off track the platform was and how disappointed he was with the result. Afterward, Ted just slapped me on the shoulder and said, “That was a great meeting.” I looked at him like he was crazy. His answer: “Because now we know exactly where his head is at.” That reframe changed the way I process tough feedback. Instead of reacting emotionally, I learned to look for the clarity in it. Three months later, we released a much stronger version of the platform. Ted taught me that honest, uncomfortable feedback is a gift because it gives you a clear path forward.

You are a successful business leader. Which three character traits do you think were most instrumental to your success? Can you please share a story or example for each?

• Open-mindedness — Technology changes quickly, especially with AI, and one quote I’ve always believed in is strong opinions, loosely held. I think good leaders need conviction, but they also need the humility to change their minds when new information shows up. Some of the best ideas I’ve seen have come from teams challenging long-standing assumptions and finding a better way forward.
• Decisiveness — I wouldn’t describe myself as a risk-taker. I’m actually pretty risk-aware. But one of my strengths is not getting bogged down by indecision. In leadership, you rarely have perfect information, so at some point you have to pick a direction, move, and adjust if needed. I’ve found that thoughtful action usually creates more momentum than waiting for certainty that never comes.
• Grit — Most meaningful progress is harder than it looks at the start. There are always setbacks, competing priorities, and moments when the original plan stops working. I’ve always had the mindset that if there’s an obstacle, you go around it, over it, or through it. It’s a bit like the Mike Tyson quote: everyone has a plan until they get punched in the mouth. What matters most is how you respond when that happens.

Thank you for all that. Let’s now turn to the main focus of our discussion about how AI leaders are keeping AI safe and responsible. To begin, can you list three things that most excite you about the current state of the AI industry?

One thing that really excites me is that AI is moving beyond the experimental phase and becoming a practical tool for real business value. Not that long ago, a lot of the attention was on novelty, whether that was generating clever content, funny memes, or showing what the technology might be able to do someday. Now we’re seeing organizations use AI in ways that actually drive growth, improve productivity, and make work more efficient. That shift from interesting to impactful is a big deal.

I’m also excited about the potential for hyper-personalization. In areas like benefits, payroll, and HCM, people don’t all need the same thing, and historically the experience has been one-size-fits-all. AI creates the opportunity to meet people where they are, understand their context, and deliver guidance that feels more relevant and useful. That has the potential to make technology feel much more personal and much more helpful.

The third is AI’s ability to reduce friction across complex systems. Most employees don’t think in terms of separate applications for benefits, payroll, HR, and other tools. They just want to get something done. AI has the potential to remove a lot of those seams by creating a more natural, conversational experience where people can simply ask for what they need instead of learning how to navigate a website or workflow. I think that will fundamentally change how people interact with enterprise software.

Conversely, can you tell us three things that most concern you about the industry? What must be done to alleviate those concerns?

Trust is a big one. People need to be able to understand not just what an AI system said or did, but also investigate why it said it or why it took a particular action. If a system feels like a black box, trust erodes quickly. That is one of the main reasons I believe so strongly in keeping humans in the loop, especially for higher-stakes decisions. Human oversight is what gives organizations the ability to review outcomes, challenge them when needed, and remain accountable for the final result.

Bias is another major concern, and it is not something you solve once and move on from. Models can drift over time, and every time you retrain a model, change the data it learns from, or upgrade the underlying LLM, you run the risk of introducing new and sometimes undesirable bias. That means organizations need disciplined testing, ongoing monitoring, and clear governance around how models are evaluated before and after changes are made. Responsible AI requires more than good intentions; it requires controls that help ensure fairness remains intact as the technology evolves.

The third concern is over-reliance. AI can make people and organizations dramatically more efficient, but one thing that keeps me up at night is what that means for the future pipeline of technical expertise. If AI increasingly displaces junior-level work, where do the next generation of engineers, analysts, and operators get the hands-on experience that helps them grow into real experts? Over time, that could create a gap in deep technical judgment. To alleviate that concern, organizations need to be intentional about how they develop talent, making sure AI accelerates learning rather than replacing the learning curve altogether.

As a CTO leading an AI-driven organization, how do you embed ethical principles into your company’s overall vision and long-term strategy? What specific executive-level decisions have you made to ensure your company stays ahead in developing safe, transparent, and responsible AI technologies?

We see AI as an important part of our strategy, but not the strategy itself. We’re a people-driven organization that uses AI to create better experiences and help people make more confident decisions. That distinction matters, especially in our space, where we’re helping individuals make decisions that impact their health, finances, and families.

One of the most important things we’ve done is stand up an AI Council, a cross-functional group that brings together technology, product, legal, and business stakeholders to evaluate how we’re using AI, where we’re introducing it next, and what guardrails need to be in place. It’s not a rubber stamp. It’s where we pressure-test use cases, discuss risk, and make sure we’re aligned on what responsible deployment looks like before anything goes live.

We’ve also embedded AI-specific checklists directly into our Architecture Review Board process. That means every time a new capability goes through architectural review, there’s a structured set of questions around data privacy, bias, transparency, and human oversight that have to be addressed. It’s not a separate process that teams have to remember to follow. It’s built into the way we already work, which makes it more sustainable and harder to skip.

At the end of the day, the goal is to make responsible AI part of how we operate, not something bolted on as an afterthought. When you build governance into your existing systems and routines, it scales with you instead of becoming something people work around.

Have you ever faced a challenging ethical dilemma related to AI development or deployment? How did you navigate the situation while balancing business goals and ethical responsibility?

Yes, and it came up in a very tangible way. As we continued advancing our platform and exploring how to make AI more helpful within the benefits experience, the team was excited about how helpful we could make it, and honestly, the technology could do a lot. But as we got deeper into development, it forced us to confront a fundamental question: where do you draw the line on AI’s impact on decision-making?

In benefits administration, there’s a clear distinction between providing information and providing advice. The moment a system starts telling someone which plan to choose or what coverage level is right for their family, you’re potentially stepping into fiduciary territory, and that carries serious legal and ethical implications. So we’ve had to be incredibly deliberate about what advancements we pursue, so we’re ensuring the AI is not recommend a specific plan or imply that one choice was better than another for that individual. Drawing that line has required close collaboration between our engineering, product, and legal teams, and there were real debates about where exactly that boundary sat.

What I took away from that experience is that the hardest part of responsible AI often isn’t what the technology can do — it’s knowing where to stop. The business case for going further is there. A more directive assistant could translate to a more impressive demo. But it would put us and our clients at risk. AI should help people make better decisions, not make the decisions for them, and sometimes holding that line means leaving capability on the table.

Many people are worried about the potential for AI to harm humans. What must be done to ensure that AI stays safe?

It takes a combination of clear standards, strong internal governance, and ongoing monitoring. This isn’t something you can set once and walk away from. AI systems evolve, and the way people use them evolves too, so you need to continuously evaluate how they’re performing and where risks might be emerging. I’ve mentioned it throughout this interview, but in ben admin, not being thoughtful about how AI is deployed is not a risk you should take. The decisions being supported often have real implications for people’s health, finances, and families, making trust, transparency, and human oversight essential.

Despite huge advances, AIs still confidently hallucinate, giving incorrect answers. In addition, AIs will produce incorrect results if they are trained on untrue or biased information. What can be done to ensure that AI produces accurate and transparent results?

It starts with acknowledging that AI is not infallible and being transparent about that with users. Problems happen when organizations position AI as something that always knows the answer. The reality is that the quality of the output depends heavily on the quality of the data behind it. If the underlying data is incomplete, outdated, or biased, the outputs will reflect that. Strong data governance and validation practices are table stakes.

Beyond data, AI systems need to be observable and auditable. Organizations need visibility into how outputs are generated, the ability to review and challenge decisions, and controls that allow humans to intervene when something doesn’t look right. In our space, where someone might be receiving guidance about healthcare or financial decisions, explainability isn’t optional. People should understand why a recommendation is being surfaced, and there should always be a path to review and correct it.

Here is the primary question of our discussion. Based on your experience and success, what are your “Five Things Needed to Keep AI Safe, Ethical, Responsible, and True”? Please share a story or an example for each.

1. Human accountability — AI can accelerate decision-making, but accountability should always remain with people. We experienced this firsthand when building a virtual assistant to help employees navigate their benefits. The technology was capable of doing a lot, and the team was excited about how helpful we could make it. But as we pushed further, we realized we were approaching the line of acting as a fiduciary — effectively advising people on which plans to choose. That’s a legal and ethical boundary we couldn’t cross. We had to be incredibly deliberate about what the assistant could and couldn’t say, and that required close collaboration between engineering, product, and legal. It was a powerful reminder that just because AI can do something doesn’t mean it should, and that a human needs to be accountable for where those lines are drawn.
2. Transparency — People need to understand how AI is influencing recommendations or outcomes. Trust disappears quickly when systems feel like a black box. A real example for us: we’re using AI to help configure clients on our platform, which involves a lot of complexity around plan rules, eligibility, and pricing. The AI does a good job, but when clients run into issues down the road, the first question they ask is, “Why was it configured this way?” If we can’t explain the reasoning behind a decision the AI made, we’ve got a trust problem. That’s driven us to think carefully about explainability — making sure that every AI-assisted configuration can be traced back to a clear rationale that a human can review and understand.
3. Responsible data practices — AI is only as good as the data behind it, and in our world, the data is incredibly sensitive. We handle Personally Identifiable Information, Protected Health Information, confidential client information about their benefits programs, and internal operational data. Understanding what data we have and what data we will allow AI to have access to is critical. For example, we would never allow personal client data — PII or PHI — to be used to train a model, even if it means compromising on what the AI capability can do. That’s a hard line for us. The temptation is always there to feed the model more data to get better results, but some boundaries aren’t negotiable. Strong data governance and clear policies about what AI can and cannot touch are foundational to doing this responsibly.
4. Continuous monitoring — AI systems need ongoing oversight. You can’t deploy them once and assume they’ll always perform the way you expect. We’ve built this into our process by maintaining a series of questions that serve as a litmus test for our models. We run these against every deployment, and we consider 90% accuracy or above to be our quality bar. Anything below that gets flagged and addressed before it goes further. More importantly, we track accuracy over time with each deployment at a minimum, because performance can shift as data changes, models are updated, or usage patterns evolve. Without that kind of disciplined, ongoing measurement, you’re flying blind — and in our space, that’s not a risk worth taking.
5. Empathy in product design — AI is fundamentally changing how we think about design. Historically, we designed systems for humans to do things — click through workflows, fill out forms, follow step-by-step processes. With AI, the design philosophy shifts. We’re now designing more for visibility and exception management — making sure humans have a clear view of what’s happening and that anything needing their attention gets surfaced and bubbled up automatically. That’s a more empathetic approach because it respects people’s time and cognitive load. Instead of asking someone to wade through every detail, we’re letting AI handle the routine and putting humans where they add the most value: making judgment calls on the things that actually need them.

Looking ahead, what changes do you hope to see in industry-wide AI governance over the next decade?

In traditional software engineering, we have well-established governance frameworks — SOC 2, ISO, HITRUST — that give organizations and their customers a shared baseline of trust. Those frameworks are only beginning to be updated for AI. ISO 42001 is a good example of progress, but adoption has been very limited so far. What I most want to see over the next decade is for these certification bodies to catch up with the pace of AI innovation.

When we have widely accepted, AI-specific standards that organizations can certify against, it will go a long way toward helping people, companies, and society build the additional layer of trust that’s needed right now. Right now, every organization is largely figuring out AI governance on its own. That’s not sustainable, and it creates inconsistency in how responsibly AI is being deployed across the industry.

What do you think will be the biggest challenge for AI over the next decade, and how should the industry prepare?

I think one of the biggest challenges will be learning how to work with something that is fundamentally nondeterministic. Traditional software is predictable. You give it the same input, you get the same output. AI doesn’t work that way. It learns, it changes, and the answer you get today might be different from the answer you get next week. That’s a significant shift for organizations that have spent decades building processes around consistency and repeatability.

Companies need to get comfortable with that reality and also be honest about where AI is the right tool and where it isn’t. If a process demands exact, repeatable results every time, AI may not be the best fit, at least not without additional controls. There’s also a growing challenge around information quality. As AI-generated content becomes more widespread, distinguishing between accurate, trustworthy information and misleading outputs is going to get harder for businesses and consumers alike.

The industry needs to prepare by investing in governance, validation, and education, while also helping leaders understand the nature of the technology itself. The organizations that will do this well are the ones that learn how to use AI’s strengths without expecting it to behave like traditional software.

You are a person of great influence. If you could inspire a movement that would bring the most good to the most people, what would that be? You never know what your idea can trigger.

Really? Mind telling my kids that? On a more serious note, if I could inspire a movement it would be this: help people lean into what makes them uniquely human while simultaneously investing in the governance and safety frameworks needed to keep AI aligned with humanity’s best interests. I was at my daughter’s graduation from DePauw University recently, and the keynote speaker, Nate Nichols, put it well. He told the graduates that if your career is built around simply following a process, AI will likely do that better. The real opportunity is in creativity, judgment, empathy, and even your wonderfully weird individuality. He also warned that we’re in something of an AI arms race, and the pace of the technology could outrun our ability to use it responsibly. I think both of those ideas belong in the same movement: becoming more fully ourselves in an AI-shaped world, while making sure this technology serves humanity rather than gets ahead of it.