Ask an engineer why an AI project stalled, and the answer is rarely the AI model. It’s the rift between the model and network. The model has no operational context, and the systems it acts through, such as charging, order management and OSS assurance, were built for callers that behave the same way every time.
To close that rift, you need a build, not a discovery. It starts with mapping: a log line to the configuration that produced it, that configuration to the protocol behavior it causes, then a ticket to the steps an engineer follows to solve it. In our projects, we build that layer with network engineers, not data scientists alone, and keep it up to date with network changes. This way, an AI model doesn’t merely answer questions – it’s ready to operate near a live configuration.
Understanding the network is not the same as being safe to act on it, so we put a boundary in front of the integration layer: schema validation, limits on rate and scope and a policy stating which actions the system can take without asking. That boundary lets an operator allow autonomy where needed and adjust these decisions later without a redesign.
You can’t set these lines in lab conditions. We design test environments to carry real volumes and the non-standard cases: malformed messages, partial failures, timeouts after a retry. This prepares the AI model for the reality of telecom operations.
Whatever AI does alone needs a way back: a rollback path and enough observability to reconstruct what it did and why. Not every action has one. A configuration change can be reversed, a denied charging request cannot. Where nothing can be undone, the boundary tightens. In our charging and network protection work, that usually separates a platform that gets switched on from one stuck in pilot.
None of this is settled practice. What is known sits with teams that have put something autonomous in front of live traffic, and little is written down.
That is the gap Telco Days exists to close. Software Mind has run it since 2018. This year’s edition, titled “Roadmap to Agentic AI and Safer Networks”, will take place on November 6 as a free online event and will focus on solutions that have already been implemented.
One of the sessions, led by Telefónica, Tallence and Software Mind, will show how a live charging platform moved to nearshore delivery, discussing necessary changes to testing and operations that case studies often skip. In another session, Next Mobile will focus on the SMS channel, where traffic can’t be encrypted and fraud adapts as fast as the models detecting it.
The rest of the agenda is packed with more insights straight from operators, vendors and analysts. Amdocs, Alianza and IBM are official partners, and RCR Wireless News a media partner. ROCCO opens the event with a survey of more than 300 operators on where AI already pays and why pilots stall.
Everything above was learned by somebody in production. On November 6, they’ll share how, free and online. Register now.
About Software Mind
Software Mind, a global technology partner, provides consultancy and software development powered by AI accelerators. Delivering deep AI, cloud and data expertise, Software Mind turns technology into measurable business value so that companies across industries can thrive in competitive markets.
In telecom, Software Mind has spent over 25 years engineering inside operator networks: SS7, Diameter, SIP, IMS and roaming. The company develops the Amplitiv product suite for roaming, signaling security and voice, and builds AI accelerators for network operations and telecom engineering.
About the Author:
Dawid Mielnik, Telecom General Manager at Software Mind
A telecommunications expert and proficient manager with over 20 years’ experience building and integrating core network platforms. At Software Mind, Dawid leads the telecom division responsible for delivering best-of-breed and tailored software solutions for operators. Dawid enjoys building teams, cultivating their potential and collaborating to achieve shared business goals. An eager learner with a strong inner drive, Dawid avidly looks for new ideas and knowledge. He advocates open source and is a keen enthusiast of 5G and cloud native technologies in telecommunication networks.
