Videos
AI Isn't Bolted On: StarHub’s Blueprint for Automation at Scale
Insights
- StarHub treats AI as the foundation of its automation strategy rather than a feature layered on afterward. Every initiative starts with AI built in.
- Running roughly 100,000 nano models across a network turns agent-to-agent communication and real-time management into a genuine scaling problem.
- Partner conversations start with understanding the problem, with the solution coming only after and always designed for efficiency and scale.
At MWC 2026, StarHub’s Chief Technology Officer, Volkan Sevindik, walks us through the hidden costs of automating at scale. He describes two early obstacles: engineers who own existing platforms often resist automating their own processes, and even once a process is built, it needs a steady supply of clean data to run on. He points to StarHub's network of 100,000 nano models as a case study in complexity, where coordinating and monitoring that many agents in real time becomes its own engineering challenge. Cost discipline matters too. Automation meant to lower expenses can quietly create a new cost center if left unmanaged, defeating the purpose of the exercise. On working with outside partners, he stresses starting from the problem rather than a pre-built pitch, and picking one process to perfect before expanding to the next. Above all, he frames automation as a cultural shift as much as a technical one, arguing that lasting results depend on building an organization that embraces it.
Volkan Sevindik:
AI is not an add on. So, we are not automating and adding AI later. We are starting with AI and AI is part of our automation journey.
Implementing automation at scale and with certain efficiency comes with challenges.
First challenge is creating the process to automate, of course, that comes with certain friction. Friction, being that engineers who are owning these platforms most of the time they don’t want to automate.
Second angle is the data. Once you create the process, how can you support the process with certain amount of data, clean data. So that you can use the data and you can use the process you have the whole set and you can apply automation to create efficiency at scale.
If you think about, we have 100,000 nano models running in a network, managing the communication among them, plus managing all these models in real-time becomes really a problem. Scaling the communication between agents and managing all these agents real-time becomes a scale issue.
You are using automation to reduce the cost, but with bringing AI to reduce the cost, you are incurring another set of cost that you have to control very carefully. Otherwise, the main objective of automation doesn't really work. We don't want to change the cost center and pay the same price while going through this significant, enormous change.
How do you do that at scale? Now we need to find good partners to work with. And we need to have good robust processes to leverage. It’s how we communicate with our partners. We always communicate with them, do not come up with the solution first, but just understand the problem and come up with the solution later, make sure that the solution is efficient and highly scalable.
Do one thing very well first, rather than boiling the ocean. Once you are convinced, go to the next level of automation, do the same thing for another process.
Automation, you cannot just solve that with technology. You need to create an organization having that automation culture.
