APPLICATIONS
“AI IS A NEW PIECE OF TECHNOLOGY – IT’ S AN ADDITIONAL ATTACK SURFACE”
Markus Nispel Head of AI Engineering and CTO EMEA Extreme Networks
“ You need to know which AI technologies and agentic systems are deployed,” Markus states.“ That will give you an idea of the upcoming capacity requirements, because you can project whether your existing infrastructure will cope.”
Markus also advises scrutinising existing AI projects to check they have a viable business case before investing further. Meanwhile, for firms that have already deployed agents without the right data or infrastructure in place, it isn’ t too late to recover.
“ You need to go back and reset expectations early enough with your customers,” Markus advises,“ because when data isn’ t ready, that usually leads to lower accuracy. If you don’ t reset expectations, once your reputation is down it’ s difficult to recover.”
Alongside data readiness, Markus stresses that governance and security must be built in from the start, as opposed to bolted on later. He compares it to the established practice of“ shifting left” in secure software development – addressing security earlier in the process rather than at the end.
“ AI is a new piece of technology – it’ s an additional attack surface,” highlights Markus.“ We’ ve already seen incidents bubbling up specifically around agentic systems, so policy enforcement and governance are really important.”
Regular AI model reviews Even the choice of underlying AI model tends not to stay fixed for long, Markus notes. Extreme reviews its options every quarter rather than committing longerterm to a single provider.
He adds:“ Who knows how this is going to look six months from now? It’ s also going to be interesting what the frontier labs’ business models will look like moving forward, because ultimately tokens will become, to a degree, a commodity.”
Markus suggests this will eventually force AI providers to compete on something other than price alone.
For enterprises, the same principle applies: those treating AI as a genuine source of differentiation – rather than a quick fix for costs – will likely come out on top once the current wave of hype dies down. aimagazine. com 105