How Saša Uses AI Workflows to Improve Support Ops

My name is Sasa, and I'm the Support Operations Manager at Mews. I lead all the go-to-market initiatives, which means I help various internal teams work faster and more efficiently.

I mainly deal with tooling, which is what led me to AI, but my work also covers project management, documentation, and process design. Being inside a large company with a lot of internal AI tools has given me a very practical, hands-on understanding of how AI can genuinely change the way people work day to day.

I started before my current role. I'm naturally interested in tech, so as soon as AI started gaining traction, I jumped in with the standard entry point, ChatGPT. After a while, I realised AI can do much more than just edit text and create content.

That shift in understanding is what pushed me toward workflow automations, and eventually toward managing an AI chatbot at Mews. That role is where everything accelerated.

Working with a large company's internal AI stack gave me a much sharper sense of what the tools can actually do in practice.

It was more of a natural transition than a hard one, but there was still a real mental shift involved. When you come from a tech background, you think the way computers used to think: if this, then that. Logical, binary, structured.

AI requires a different mode of thinking entirely, one that is closer to how you would explain something to a person than how you would write code. Prompting was new to me, even though the underlying logic was familiar.

And for anyone coming from a non-tech background, I genuinely think that is fine. The tools are accessible. It's not as hard as it looks from the outside.

I would skip the whole content creation and text editing phase entirely and move straight to problem-solving. That is where the real value is.

AI is genuinely powerful at identifying broken workflows and fixing them, and that is a much faster path to tangible results than learning to generate content. My starting point was content, and while it taught me the basics, I could have saved a lot of time by going directly to the question: what problems do I have in my work, and how can AI fix them.

Skip the things most people start with, using AI as a text generator, and instead look hard at the problems you already have in your day-to-day.

Look at anything that old-school tooling currently handles, then ask how AI could make those same things run faster and smoother. The real opportunity is not in replacing what you know.

AI is very good at taking your existing expertise and making it better. Start there, with what you already understand, and layer AI on top of it.

I use it across a range of things. For communication and how I express myself, I still use GPT for improvements and refinements.

But the more interesting work is on the project management side. I built bots that pull data from different sources and automatically create documentation, post updates to Slack channels, generate Jira tickets, and size the work for planning. That has made quarterly planning significantly faster.

My next goal is to move further toward full automation, where the bots don't just create things for me but handle entire workflows end to end. That is the natural next step.

The learning time is real, but it compounds. Building my first bot took a few days to get the basics right. But once that first one was done, every next one became easier because of the knowledge transfer.

The initial investment pays back fast, especially when you consider how much time I used to spend manually writing documentation, chasing updates, and keeping Jira tickets current.

Once you do it for one use case, you can apply that same approach to other use cases very quickly. It takes time to set up, but it is absolutely worth it.

Two things stand out. The first is blind trust. AI can hallucinate, and that is a real problem that gets overlooked, especially when people start relying on AI-generated data without verifying it. The first step should always be to confirm that the output is actually correct before you start building any trust in it.

The second mistake is treating AI as a silver bullet, as a replacement for humans or for every other tool. AI works best alongside existing human expertise. It changes how we work and frees us to focus on building knowledge instead of doing repetitive tasks. But it does not replace judgment.

Just start. That is genuinely my main advice. Learning AI is not something that happens through reading about it. It happens through experimenting, writing prompts, trying to understand how AI thinks, and looking at your own environment to see where it can help.

It is not as hard as it seems from the outside. The barrier is mostly in getting started. Once you do, you figure out quickly that the tools are accessible, the logic is learnable, and the results come faster than expected.

The roles across the industry are changing, and mine is no exception. When I started, support operations were a more traditional function.

Now it is shifting toward AI tooling, process design, and consulting. I am personally interested in deepening my understanding of AI beyond the user level, including some certification and a stronger grasp of the backend.

Right now, I understand AI well from a practical, applied perspective. But I want to understand what is happening underneath as well. That is where I am heading.