AI can be great. I’ve used it to add clarity on things. For example, if I felt like I’ve gone too far into the weeds, I’ll ask AI to “Rewrite for clarity” or even as we build out our Jira, I’ll ask AI to “Rewrite this to follow Jira best practices”. So it takes my ideas/thoughts and refines them for a clearer understanding.
I’ve also used AI heavily in the research segment. Research this error message, or even tell me about a certain EXE especially if the SME of the application in question is Out of Office. Keep in mind, I always call out that I am not the SME and that I’m merely just using AI to give me a basic understanding of the entity at hand, that I would always recommend syncing up with the SME.
AI has been great, is it perfect? No, not at all, but it does help. I just always double check it and remove anything that is not incorrect or misleading.
Emails is also a great usecase. Do you know situations when all you want to reply is a plain “no, go away”? :wink:
That’s too unpolite? I use AI to help draft an elborate and polite reply to reject a request with a few “good” reasons :grin:
I have changed from writing important english mails to customers by my self to writing them in my native language and asking AI to translate it to get the right tone. Privately I have used AI to recommend restaurants and tourists attractions for my holidays.
I try to use it more and more, but always keep in mind that AI can hallucinate and write gibberish.
My only issues with AI is privacy/confidentiality and price.
After enabling Microsoft Copilot in products like Teams, Outlook, and Excel, it has significantly improved my day-to-day productivity—helping with tasks such as writing emails, updating Jira stories, and creating documentation.
For automation, we’ve also enabled GitHub Copilot, which has been a great help in writing Python, Ansible, and Terraform code, as well as streamlining daily automation operations.
However, Davis Copilot is still not enabled, and I’m eagerly waiting to use it for my Dynatrace operations.
I admit I do not use AI much, at least willingly( I’m looking at you search engines with your AI overviews). In the cases that I needed to use it, it’s either for some quick skeleton code creation, or for some document parsing/translation. In the second case I discovered that it can blatantly lie in my face about the progress so I’m hesitant to trust it for work.
What I find it adequate for is for general knowledge I need to refresh, as it generally provides good breakdowns.
P.S. I hate with a passion all the “friendliness” bloat it has around it’s answers, but that’s my preference I guess
I mostly use AI to keep myself organised and remember things. I am context switching non stop all day, it is so easy for me to forget to do something I said I’d do at the end of one meeting, as I rush over to another completely different one. I have regular prompts to tell me about any prod support issues from the previous day, what I did, and any action items I have from meets and chats.
It is also way easier to ask the AI about internal documentation than search Confluence. I am dreadful at searching Confluence. And it can also present the information with different audiences in mind, so it saves me a lot of time doing knowledge transfers on more technical topics.
PLUS - it can write a nice message explaining something to someone when I’m frustrated that they haven’t understood the last five times I’ve told them 😂
I’m also using it with Davis obviously… but let’s talk about that later 😉
My first AI was a hill-climber routine used to crunch cognitive models for an academic research team studying neuroscience. Then it was Clippy. Now it’s Davis. Next it will be the MCP server/services running in Dynatrace + everything else.
I still think Clippy was actually a human. How else would Microsoft grow to have +150,000 employees worldwide back in 2002?
It was a surprise to me that compliance was concerned about the AI prompts:
- how long are they retained by the LLM + 3rd party AI?
- what fields are masked/obfuscated by the prompt interpreter?
- do I have control over blocking certain keywords in prompts?
Makes good sense now…but yeah, wasn’t thinking about that 3 years ago.
I’ve been using AI tools in my professional work with Dynatrace to solve some really complex monitoring challenges.
One example:
We had a multi-tier banking application outage (frontend → API → MQ → mainframe). Normally, troubleshooting this would take hours because:
Logs were scattered across multiple systems.
Metrics were noisy and overwhelming.
No clear root cause.
AI helped me overcome this:
Dynatrace Davis AI detected anomalies and correlated the technical root cause.
I then used an LLM (like ChatGPT) to:
Summarize the incident in plain language for stakeholders.
Suggest remediation steps based on tags (e.g., JVM tuning for Java services).
Automatically create a ServiceNow incident with all details.
This saved us hours of manual work and reduced MTTR from 3 hours to 20 minutes.
For me, AI is not just about answering questions — it’s about automating workflows and making complex data actionable.
AI has completely changed the way I work – especially in observability and automation.
I use AI daily as a silent “pair engineer” when working with Dynatrace, Terraform, Kubernetes, and automation pipelines. It helps me:
Quickly prototype DQL queries
Debug tricky pipeline and YAML issues that would normally take hours
Generate and validate Terraform, Bash, and Python automation scripts
One of the biggest improvements is speed of iteration — I can now test ideas almost instantly instead of spending hours on trial and error.
AI also helped me overcome one major struggle: documentation gaps. When something isn’t clearly documented, I can reason through it with AI and often reach a working solution faster than searching through dozens of forum posts.
Outside of pure tech, I even use AI to optimize my daily productivity and health routines
Not a developer myself, but I like to use AI to vibe code something that bothers me. It could be just as a side quest or just to engage some curiosity. The biggest use case for me is learning and having some philosophical conversations (which doesnt go anywhere beyond a certain point). As someone who is interested in most available tools and models, I like to pit Gemini vs Claude vs GPT for the same input. I also tend to run models locally and use the Perplexity Comet browser, but although I have a tough time justifying the use case for a browser like Comet.
But my top use cases are:
Vibe coding something different when I get the time - could be something simple like Computer Vision, or trying my luck out at building the next open world PC game (tried this as a browser game, results were far from satisfactory)
Personal recommendation systems for career, travel, food or anything similar
Philosophical discussions or just something to clear my head
Within a Dynatrace perspective, I use Davis as some sort of a sanity check mostly, to validate my reasoning.
I’ve used it to help with some Dyntrace items - how to create log processing rules to extract a bunch of different items from the log file path. I found I had to really feed it the proper Dynatrace Managed terminology, and my intial attempts and it helped me get 90% of the way there, but then I had to tweak it a bit because it was still brining in some of the SaaS ways of doing it.
Also used it when the lights were off and my 4 year old wanted a rhyming bed time story about a unicorn-butterfly who helps people. Then I realised I could add in additional prompts to help build/demonstrate breathing techniques to calm down/ or to include topics that she is scared of and how the unicorn is too and how she overcomes the fears..
My first experience with AI chat was with ChatGPT, searching for news about technologies around the world. Then I used it to understand some behaviors related to investments.
Currently, I use it to calculate some mathematical reports and understand new technologies or tools. To create cool images, I use Gemini.
But I think it’s important to understand that tools like ChatGPT or Gemini are just a tool to facilitate certain information, but it’s necessary to do more research in official sources.
Any experience with the Dynatrace MCP to inject data in Ollama? I’m currently starting an exploration project in an isolated network. Do you have any advises, return of experience to provide ? Did you install it on your laptop or in a server ?
I havent explicitly tried using it with Ollama, but as long as you are able to reach your environment within the same network, then the local MCP should help.
Currently I have an integration of the Dynatrace MCP server towards a a DT Playground environment. I am running it locally on my laptop using Claude Desktop. I find the OOTB MCP server quite limited and I am thinking of extending it using the Dynatrace APIs to make it more powerful. I will post once I have some progress there.
The other idea, that I initially went to was to build a KB and feed it to a a local model to be able to prompt directly from it, however I didn’t have the time for that.
I think MCP is the way to go currently, but making it custom build as of the limitations curently it has.
