Telecoms Automation — you probably don’t need to wait for the platform.

Telecoms automation doesn’t start as a technology decision. It starts as four organisational agendas arriving in the same room: Operations looking to reduce network downtime and open ticket volumes, Product Management seeking new revenue streams and customer stickiness, Executives targeting margin improvement and competitive differentiation, and Network Engineering asking what the existing infrastructure can actually support. While these agendas aren’t necessarily opposed, they introduce so many requirements that the result is a behemoth of a programme that will take years to realise.

The automation scope is real and well understood. Networks that heal themselves when something breaks. Services that go live without manual intervention. Traffic that knows its own priority. Systems that synchronise billing and infrastructure the moment a customer clicks buy. Infrastructure that learns continuously from its own behaviour. Every operator knows what needs to be done. The sequence and the tools used to attack it are where the real decisions live.

To satisfy all of the requirements in the room and simultaneously be safe enough for the architects and engineers — a large, multifaceted unified platform is chosen. This logic is sound. The platform vendors make a compelling case: integrated capability, reference customers, and a roadmap that appears to address every requirement simultaneously. These platforms are expensive and demand several rounds of iteration, testing and reiteration. Invariably the requirements shift mid-implementation, forcing another cycle. The journey to full deployment rarely reaches a natural conclusion before the next set of requirements arrives.

The platform may still be the destination. But most operators don’t need to wait for it to start moving.

Recent developments in AI — specifically frontier models’ agentic software development tools like Claude Code, Codex and Cursor — have put powerful tools into the hands of everyday techies who have only dabbled in coding before. Every network operations team has a few good tech-minded people with the right blend of experience and domain knowledge. With a healthy bucket of tokens and some long nights experimenting, they could develop 50-60% of the automation requirements in a sandbox environment. This would jumpstart not only a potentially in-house solution but would also sharpen the requirements for an outsourced model. Both outcomes save time and money.

The operators making the most progress on automation aren’t always the ones with the biggest platform budgets. They’re the ones that found the most painful bottleneck first and put a small team on it — their top people who relish solving hard problems in unconventional ways. AI has changed what that team can do and how fast they can do it. The platform is still the destination — but the journey starts with small steps, small iterations, continuous improvement, and the clarity that comes from building something real before buying something comprehensive.