Obviously, adopting artificial intelligence in managed services sounds great. It promises efficiency, faster responses, less burden on teams, and better decision-making (though that always depends on who’s making the decisions). But what people don’t usually talk about is what really happens in the day-to-day: the nuances, the challenges, the unexpected curves.

In recent months, I’ve worked on several projects where AI was supposed to be the great accelerator. And it has been… but not without its complexities. That’s why today, I’m not here to talk about theory or benefits (which do exist), but about what you actually encounter when you get your hands dirty: doubts, decisions, roadblocks, and real lessons learned.

Here are 5 real challenges I’ve experienced while implementing AI in managed services—straight from the trenches:

1. Lack of Strategy

Without a doubt, the first and most impactful issue I run into is the lack of a clear strategy. You can’t approach AI with a “we’ll figure it out as we go” mindset—but unfortunately, that’s more common than you’d think.

Many clients want to implement AI but skip key questions:

Why exactly are we using it? Which part of the process does it improve? Who’s leading the change? How will we measure impact?

Without these answers, frustration follows. Expectations go unmanaged, real problems remain unsolved, and AI ends up being just another experiment… one no one fully adopts.

Partnering with a company like Axazure shouldn’t just be seen as a tech implementation support—it’s a strategic collaboration. Someone to help you decide which steps to take, when to take them, and how to approach them. As I said in another article: you have to measure wisely, and define KPIs that actually track what matters.

2. Disorganized (or Nonexistent) Data

Now that the Tour de France is in full swing, all eyes are on the route, the cheering crowds… but above all, the cyclists: the best in the world, because without them, there’s no race.

AI works similarly. You might have the best route (your strategy) and the best teams (your technology), but without top-level “cyclists”—in this case, quality data—you’re going nowhere.

What usually happens is the data is siloed, mislabeled… or just doesn’t exist. 80% of the effort in implementing AI isn’t in the model; it’s in the prep work: cleaning, classifying, structuring—in short, setting the stage.

This becomes even more critical for services aiming to automatically resolve tickets, maintain natural language conversations with clients, operate 24/7, provide personalized experiences, or guide users through complex processes. It all sounds great… but it’s unfeasible without a solid data foundation.

I call this pre-adoption, and yes, it requires investment.

3. Uneven Team Capabilities

Another key issue—closely tied to adoption—is that not all teams are equally prepared to work alongside AI. And I’m not just talking about technical training (which is often the least relevant factor), but deeper issues: culture, mindset, trust.

Sometimes AI is seen as a threat, and there’s the problem. If you don’t support the change with a clear strategy for training, communication, and tangible value examples, resistance sets in fast. It’s not about bad intentions—it’s about fear. If people feel (even wrongly) that this might put their job, knowledge, or autonomy at risk… the implementation is doomed from the start.

That’s why aligning the project’s objective with internal messaging is so important. You need to explain not just what AI does, but why it does it, what for, and how it improves people’s work. Because if the teams don’t understand it or believe in it, they won’t adopt it—plain and simple.

With this in mind, at Axazure we propose an AI Adoption Center that combines training, workshops, and sessions tailored to different profiles—ensuring no one is left behind.

4. Unrealistic Expectations

The hype around AI has caused a lot of damage—and this isn’t the first time that’s happened.

The reality is that every AI project requires training, context, and adaptation. Good results come through iteration, not magic (which means—you guessed it—you have to invest). Implementing AI is more like training a new teammate than buying a superpower, and that’s likely the core problem.

A few weeks ago, my boss asked if AI was helping me, and I told him clearly: that mega proposal I used to spend nearly a month on now takes about a week. But I also made it clear—respecting confidentiality—that I’m still the one writing and thinking, and I know exactly what I want. I put the pizza in the oven, it comes out delicious—but I chose the ingredients, I arranged them, and of course, I’m the one who enjoys it.

In managed services, this matters even more, because it’s not just about speeding up tasks—it’s about doing so with rigor, consistency, and continuity.

5. Security and Ethics: The Elephant in the Room

Now we reach the topic no one wants to bring up—but it’s everywhere. The mother of dragons: security, privacy, bias, and accountability. The elephant in the room.

Once AI starts intervening in critical processes (responding to clients, prioritizing tickets, generating documentation, recommending actions), uncomfortable questions arise:

What data was used to reach that conclusion?

Could there be bias we’re unaware of?

Who’s responsible if the recommendation is wrong?

These questions often go unaddressed at the start—they’re avoided or downplayed—until the project advances… and sometimes, by then, it’s too late.

Once again, we return to Point #1: STRATEGY. And yes, also to #2, #3, and #4. If you haven’t sorted out permissions, decision traceability, and an ethical framework from the start, you’ll end up limiting AI usage out of caution—or worse, shutting it down completely after the first incident.

AI can completely transform how we deliver managed services. It can reduce time, anticipate issues, prioritize better, and offer more consistent experiences—but only if it’s implemented smartly… and soulfully.

CONCLUSION

Artificial intelligence can completely change the way we deliver managed services—reducing time, anticipating errors, prioritizing more effectively, and offering more consistent experiences—but only if it’s implemented with intelligence and heart.

A managed service is not a one-off project or a product. It’s a long-term relationship, with ongoing challenges and evolving needs. That’s why implementing AI in this context demands even more clarity: it’s not just about whether it works, but whether it’s sustainable, understandable, and adaptable over time.