AI – super hot 🔥 topic. It is constantly changing, evolving faster than we thought. Every day we can be surprised at what new areas it can be used, and for what purposes. It’s hard enough to get a handle on what’s happening today - let alone think about tomorrow.
Many people struggle with how to combine AI - and learning in this field - with their ongoing, day-to-day work. The world is moving fast, and future require a high degree of workforce automation, advanced agent systems, and global regulatory frameworks.
Some DynaMights, who are part of this community decided to share their thoughts on this topic. If you want to add something as well feel free to leave a comment. We want to start the conversation, not to provide only some personal experience and point of view. 😉 Without further ago here is an extension of the introduction I outlined above.
Marina Pollehn @marina_pollehn:
I see that many companies are interested in the future vision of running everything IT-related in a fully automated way – referred to as AIOps. Often, this means that they are trying to jump to the desired result by, for example, feeding an AI tool with all their conversational or organizational data, hoping it will find patterns and turn them into something useful – namely autonomous operations.
I personally like to argue that this is probably not the most efficient way. Instead of letting AI be the end-goal, I think that it’s more relevant to think first about which use cases you would like to have automated. Where can you spot potentially “easy wins” or what do you spend a lot of time on as a time while it has repetitive components? For which use cases do you always want to be in control as a human, but could you be aided with forecasts or automated checks? Breaking the challenge down into small silo’s, which you later connect and integrate with other use cases or teams, builds trust and makes the giant AI wave coming at every organization less dooming.
Another critical component in using AI in operations is trust in decisions
of an automated process.
I am very much in favor of the human-in-the-loop principle. While one might argue that a process with human-in-the-loop (for example a person approving a ticket which is automatically generated) is not a fully automated process, this phase can create trust for employees. Once this trust is established, a team can move to the next phase, of integrating the use case with other teams and use cases, as well as letting it run fully autonomously. This human-in-the-loop approach can be implemented in many different ways.
You can, for example, let Dynatrace Workflows create tickets which are approved by a human before the next automated action is triggered, or you can let every Xth decision be verified by a human. Once this process is error-free, you can rely on less and less human verification, To summarize, I would prefer a gradual and targeted transition towards AIOps, instead of using AI just to claim you are using AI. In the end, it’s not the tool (AI) but your vision, design and clear goal of how and what you want to automate, which will help you win the game.”
Antón Pineiro @AntonPineiro:
I agree with what Marina says. I see the future from two perspectives, specifically from Dynatrace’s point of view.
- 1 -
How Dynatrace enables users to ask generic or specific questions and respond with data from their environment. For example: “Create a dashboard with the critical vulnerabilities from the last week,” “I want a workflow with a daily report of the issues for application X,” etc. The challenge here is how good the responses are and how much time they can save, without needing to understand Dynatrace internals, DQL, and so on. And of course, the cost of these queries, to avoid a single question consuming a large amount of resources due to the associated DQL.
- 2 -
The second aspect is how Dynatrace can monitor the consumption of AI services by customers, providing value to understand things like “which teams are consuming the most tokens,” “which prompts have consumed the most tokens,” etc. In other words, providing visibility into how AI is being used and delivering value to help optimize those resources
Gil Givati @gilgi
I completely agree that AI is not a goal but more of a mean to a better world (especially in IT). Further more, I think that most organizations have not yet matured enough to be automated, even without AI. For that reason I think that the first, quickest and most noticeable win of AI would be by using it as a tool to quickly get answers to questions we’re struggling on answering promptly.
As observability relies on huge amount
of signals of different types, on dependencies and behavior we don’t always see,
AI speeds things up for us.
Take for example the “latest” Dynatrace Assist. I have personally attended a few deep dive troubleshooting sessions that raised questions and simply by just asking assist the right questions provided the organization within minutes the ability of validating assumptions and theories, which otherwise would have required going through different screens and apps, trying to understand the relationships between the different signals and spend hours if not days before being able to proceed.
With that respect I believe the first goal or organizations should be maximizing the value, number and validity of “simple” AI based discussions. Only once this is achieved and people trust the answers, they’ll be able to progress further and start relying on AI based automations progress further and start relying on AI based automations.
Antonio Sousa @AntonioSousa [BTW, no AI involved in writing this, besides linguistic corrections ;)]
In IT, we have evolved in the last decades in waves. I was fortunate enough to start surfing the net in the early nineties, and I was quite aware that there was a lot of resistance at the time, especially from the traditional Telcos. This is just another big one, indeed, and one that we all have to surf! What we are seeing is just the beginning. In January, a close friend (who has a dream job of evaluating AI progress) explained to me that in mid 2025, LLMs were like 17-year-olds, studying and taking exams for what their teachers would teach them. At the time, in January, he told me that they were like second-year college students; but in essence, just transmitting back whatever they would be learning. But he also alerted me that in the second half of this year they would be PhD students, deriving new things!
I heard back from him on February 13th, when he told me that what he had prognosticated for later this year had indeed already happened, with the February 5th model launches. And indeed, every week that passes is now not about what LLMs are parroting, but what they are discovering. Take Mythos, for instance. Security is a domain where I have had quite a success with my clients, especially in the first decade of this century. At the time, I would say that Security, and Application Security, was like the Wild West of the Web (WWW). Wherever I looked, I would just see bugs & vulnerabilities! Over time, lots of years later, security is now much better, and discovering new holes became more and more difficult! So, reading what Mythos did with Firefox and OpenBSD, just to name two of the most interesting, is just taking me back to those wild days, 20 years ago!
It's much faster now! And this is clearly entering a vicious loop, where, make no mistake, we humans are not able to track this by ourselves!
We need help, and we need systems ("group of interacting or interrelated elements that act according to a set of rules or set of constraints to form a unified whole" in Wikipedia) that can observe, track, and correct what is happening already. There are great challenges in the near future, considering the quality, speed, and correctness of such systems. Dynatrace will be an important piece of this, as we urgently need to observe more and more of what is already going on...
I asked an important person (almost exactly) 10 years ago: "When will the rise of the machines happen"? I got a first answer "Never". I replied "Never say never again". And after some thought I got back: "20 years". Are we half way? ChatGPT is not even 4 years old. But the last weeks have given us major clues where this is going: AI is now authoring scientific peer-reviewed papers. It is discovering things that no one thought possible. It is not following the herd, as the disproval of the "planar unit distance problem" posed by Erdős clearly demonstrates.
So, don't try not to surf the wave. We will have problems along the way, and ultimately there might be no John Connor. We have to collectively use this new power of humanity, to make life even better! Observability will be key in moving forward, well beyond the typical IT and application quests of the past, and I'm pretty certain Dynatrace will be a huge part in this!
Daniel Stanizzo @DanielS
Now I’ll take you into the world of "fairy tales". Or perhaps fairy tales will soon become reality, or they already are?
"Aladdin and the Art of Asking for the Right Wishes: How to Navigate the Greatest Technological Wave in History"
The "Move 37": The Birth of Autonomous Cognition
For decades, it was believed that artificial intelligence was only capable of repeating and elegantly packaging what it had already learned from humans. But that theory was shattered in March 2016, during the historic Go match between world champion Lee Sedol and AlphaGo.
In the second game, in the mythical move 37, the machine placed a stone in a spot on the board that no professional human player would have ever considered. It was not a mistake. It was a move of alien beauty that resolved the scenario in a completely new way and redefined a 3,000-year-old strategy. AlphaGo was not imitating; it was discovering, deducing, and creating new knowledge on its own. It is the starting point of a new AI: a tool that no longer just processes data massively, but is capable of reasoning autonomously.
A Paradigm Shift of Unimaginable Scale
We are not facing a simple software evolution. Throughout history, humanity has surfed waves that redefined the rules of our physical existence: the steam engine multiplied our muscular strength, initiating the Industrial Revolution; the atomic bomb altered global geopolitics by making us realize that, overnight, we were capable of causing our own apocalypse; and the internet democratized information and brought the planet closer together for good.
However, all of these past tools shared a pattern: they replaced or amplified our physical strength, or connected our existing structures. This is the first time in history that we have created something capable of following patterns similar to our own thinking. A machine capable of emulating and expanding cognition on its own is a revolutionary milestone that poses a paradigm shift on a scale we cannot yet fully envision.
In the face of this unprecedented cognitive revolution, a deep digital FOMO (Fear Of Missing Out) has emerged in the market. In boardrooms, a desperate mantra is repeated: "We need to implement AI now". However, most companies want to join the trend out of inertia, but they do not really know what they want it for or how to successfully integrate it into their daily operations.
In this frantic race to achieve full automation (AIOps), we risk acting like Aladdin with the Genie of the lamp: assuming the machine understands what we actually want, and not just what we literally say. If a platform administrator asks an uncontrolled AI to "reduce infrastructure costs by half", the system, in its relentless and literal logic, might decide to shut down critical production servers over the weekend. The problem is not the Genie's lack of power; it is our inability to formulate the right wish and foresee its consequences.
From Aladdin to Socrates: A Maieutic Conversation with the DynaMights
To avoid the dangers of a hyperactive and literal Genie, the technology industry must look back, specifically to the Socratic method of maieutics: the art of bringing truth to light through iterative dialogue. Socrates did not aim to give answers, but to teach how to ask the right questions.
In the era of AI, success lies in this type of dialogue. Debating this approach with the DynaMights community makes it clear that the path to automation is not a straight line, but a multidisciplinary conversation:
The Strategy of Gradualness: Against the blind urge of FOMO, Marina Pollehn proposes a dose of elementary pragmatism: AI must not be the end, but the means. The smart path begins by breaking down the great technological challenge into small, isolated silos, identifying repetitive processes that drain the team's time to win quick, "easy wins."
The Value of the Intellectual Assistant: In line with this, Gil Givati reminds us that most organizations are not yet mature enough to fully delegate their operations. Therefore, the first great value of AI is not autonomous execution, but its use as an interlocutor capable of validating theories and resolving dilemmas in minutes—such as when querying Dynatrace Assist—saving days of analytical navigation through endless screens of data.
The Ethics of Human Control: This need to keep the expert at the center of the equation—the Human-in-the-Loop principle—transcends the technical to become a humanistic imperative. This philosophy aligns closely with the doctrine set forth by the Pope in the encyclical Magnifica humanitas, which firmly defends that technology must expand, and never replace, human responsibility, free will, and dignity. The machine proposes and assists, but the final word belongs to human conscience.
Conversational Democratization and the Cost Challenge: For his part, Antón translates this maieutics to the reality of the everyday user. AI allows us to interact with the environment using natural questions ("create a dashboard with this week's alerts"), abstracting us from the underlying technical complexity. However, he introduces a vital warning: every wish made to the Genie has a price. Translating natural language into complex queries consumes immense resources, making efficiency and computational cost control a priority.
Observability as the Lifesaver of the Revolution
As this wave accelerates, the scale of change surpasses our biological processing capabilities. As Antonio Sousa points out, the evolution of language models is so rapid that they no longer just repeat patterns, but scientifically and autonomously discover knowledge. In this "endless loop," humans cannot keep up alone. We urgently need intelligent systems that observe, contextualize, and correct what is happening in real time.
This is where observability is transformed. It is no longer enough to monitor whether servers are on; now we must apply observability to the AI itself. We need to answer urgent business questions: Which queries are draining our budget? Which inefficient prompts are driving up token consumption in our cloud environments?
Navigating this transition requires abandoning the rigidity of corporate fear and embracing Bruce Lee's famous premise, deeply rooted in the philosophy of Taoism: “Do not oppose water; be water.” Faced with an inevitable cognitive tide, rigidity will break us; fluidity and adaptability will allow us to design the technological vessel of tomorrow. The future of IT will not belong to the organization that implements AI the loudest, but to the one that, with Socratic wisdom and human temperance, learns to ask the right questions.
Antonio Sousa @AntonioSousa
I love how Daniel frames this conversation around Aladdin, Socrates, and the philosophy of Tao. There is a new trend in Silicon Valley, and they might not be hiring new programmers! They need new capabilities, and hiring Philosophers was not on my list. But it really makes sense!
BTW, the person that told me that the rise of the machines will happen in 10 years, also told me that programmers would be some of the first to lose their jobs with AI...
Phani Devulapalli @p_devulapalli
Interesting conversations and I know I am late to the party...
I follow chess quite keenly, and the first time I heard that a computer had defeated chess grandmaster Garry Kasparov, I was genuinely puzzled. Could a computer really do that?
Back then, computers mostly did what humans programmed them to do. Now, AI feels very different. It can learn, improve, and sometimes make decisions in ways that are not always obvious to us.
That is what makes instances like “Move 37” so interesting. As Daniel Stanizzo mentioned, it felt like a moment when AI started showing signs of creativity. So, yea I would not be surprised if a Terminator shows up in near future… preferably to save humanity.
There should always be checks in place
(human in the loop?) to make sure AI doesn’t turn autocratic
(Dictator vs Democracy) as it lacks the emotional intelligence that humans have when making decisions that are just
beyond data and numbers.
Maybe we just need to strike a right balance by using AI as a helping hand and at the same time letting AI run the autonomous ops where it’s safe to do. I wonder if there would ever be a time when AI is completely autonomous even with guardrails set….. what if it learns to break its guardrails (Terminators again…)?
I have followed chess developments since I programmed a little chess program on a Spectrum 48K. Deep Blue was quite an achievement, but I believe AlphaZero was really the big advance: in 2017, five years before ChatGPT, when it learned the entire history of chess moves, alone and starting only with the rules, in only 4 hours! Today, humans have no chance against chess programs...
Regarding guardrails, I believe that in the future they will not protect us against evil AI. Asimov's laws will be of no help too. There will always be hackers, be them bad agents, be them the new John Connor. As always in Human history, it will be a race, and let's hope we don’t collectively get it wrong.
I would add that things are converging. Elon Musk just reinforced what I heard 10 years ago: that Humans will no longer be in control in ten years. That’'s 20 years from 2016! Maybe we will have to start preparing for this!?
What do you think?
* This topic will be continued...
1 Comment
