Digital transformation doesn’t mean pain, say Apstra and Deloitte

Immediate savings

Immediate savings will be achieved, says Karam, through “powerful full lifecycle automation of network services on day 0” where networking staff get “a simple operational model that hides the complexity of the infrastructure while helping them to make moves and changes significantly faster and more reliably through root cause identification”.

The newest version of the Apstra operating system (AOS 3.3) now includes Juniper and SONiC contributions among its multi-vendor support, so you get more switching options in the software.

Support has also been added for ESI, the new standard protocol and reference design. This creates new models for creating backup capacity and for setting performance goals.

The third tranche of improvements is in operational power and efficiency, as network operators can now exert control over their networks in much greater incremental detail. They can see and effect change by fine-tuning IP address allocation, for example.

There is an improvement in the way role-based access control (RBAC) is delivered and the configurations locked in. Improvements have been made to an array of intent-based analytics functions, most notably the headroom probe.

“We have liberated the network operators from the tedium of having to type arcane commands, device by device, to configure or troubleshoot networks,” says Karam.

He promises that Apstra frees network operators from nightmares about making configuration-error changes, which can result in massive, business-impacting outages.

It will be fairly easy too, for those “fluent in the field”, says Karam. The network operators won’t need to learn Python or any other new programming language.

Companies need to demystify “digital transformation” because it sounds ominous to most people. Even some networks staff are afraid of losing their jobs to a machine. Meanwhile, many company executives are secretly paralysed by fear.

“Clients have a lot of questions about digital risk and everyone wants to understand the implications,” says Gribben.

When there is so much you don’t know, you have to fire out tracer questions to establish your range before you can home in on your target of understanding. But the irony of asking questions about AI – which, by its nature, is a technology that’s not afraid to ask dumb questions – is that you risk looking stupid, and nobody can afford to look vulnerable at the moment.

Silly questions

Robots don’t lose status by asking silly questions in meetings – but there is a massive risk for humans.

The stupid questions are the ones you don’t ask, says Gribben. “Digital ethics and the impact on society are a big part of the Digital Risk Academy curriculum,” she adds. “There are many questions that affect the company and its compliance to standards. The line of questioning starts with what do you use it for and ends with the bottom line: will it affect the brand?

“As humans, we need to feel comfortable and confident in working with AI. In order to understand how to manage risk tolerance, we need to set up guard rails – to limits on where AI will go. You are setting a machine off, but you want to manage the parameters it stays in.”

There are some clear parameters – the law is not negotiable, for one. But everything else needs boundaries. It’s your baby – you need to establish boundaries as they grow up so fast.

We do that by re-examining where we stand on risk. “Once it was acceptable to run a yearly risk analysis, but technology evolves much faster now,” says Gribben. “These days it makes sense to fine-tune the directions of AI at least once a month.”

There is a skills gap that affects the way AI is implemented, controlled and managed. So once it’s installed, they can’t afford not to employ you. Sounds like a classic IT industry sales tactic.