Robots seemed to be the dominant force at NRF until I was reminded how we always have to take care of the poor old infrastructure, however sophisticated the innovation.

Here were the robots, the future of retail, manoeuvring around the show, capable of understanding your requests, taking payments, offering advice and product information. But suddenly some had come to a grinding halt. Why? Because there were not enough charging points to keep them powered up or in other cases, because their Wi-Fi connections had gone down. It doesn't matter how clever you are in retail technology, you still need the infrastructure.

Essentially this eruption of robots at NRF is led by Pepper (sadly not named in honour of my Vista colleague, James Pepper), the device born of the partnership between IBM's Watson supercomputer and SBRH.

Pepper now has a gang of imitators and whether it is IBM or one of the others, a winner may emerge that has a role to play using biometric recognition to facilitate payment, recognise customers and administer loyalty rewards on the spot. This is a set of attributes that should have application in the theme park sector, but in mainstream retail settings robots still feel like something of a novelty item, still performing tasks that can ultimately be done better by humans.

The truth is that Pepper is an excellent promotional tool for IBM Watson and its artificial intelligence (AI) capabilities. Throughout the show we can see how major retailers and hospitality operators are increasingly looking to AI as the cognitive technology that will give them the edge, constantly learning about individual customers and bringing them the benefit of a vast amount of background information that it can tailor to specific requirements and tastes with minimal interaction.

These AI capabilities are being meshed into the growth of voice-first commerce through Amazon Alexa, Google Home or Siri. For the moment, no single technology stands out, but Starbucks is demonstrating voice-led ordering that makes getting a coffee almost frictionless.

On a grander scale, IBM is collaborating with General Motors (GM) to take the connected car a lot further into retail. It is clear at this year's show that these cognitive capabilities will recognise the driver's preferences and loyalty schemes, making recommendations and enabling in-car ordering that ensures quick service and precise, highly personalised service at petrol stations, stores or restaurants. This is much further on from simply displaying a list of the options for filling up when the tank gets low and is the fruit of what is clearly a lot of research into driver habits by GM.

This year, NRF is vaster than ever with a huge emphasis on technological innovation. There is, for example, a bewildering number of hardware vendors, yet many have technologies that are hard to differentiate.

The virtual reality (VR) sector, however, is proving irresistible to anyone at the show with an eye for innovation. How much of it proves to be effective is open to question, as with the robots, but the ability to stroll down the "endless aisle" picking up items, turning them over to inspect them and place them in the trolley you are pushing is dazzling.  When the store doesn't have what you want on the shelf, a trip into VR may well persuade you to purchase. We could well see this as part of the gamification of retail, readily accepted by millennials, even if there is a barrier to overcome about using a headset just worn by someone else. It remains, however, a very expensive technology that has yet to prove it can earn its money back.

Which brings us back to infrastructure. Verizon, for instance is showing technology using the very latest in mobile data. It relies on 5G connectivity and in the UK that is still a long way off.  The big players at NRF are making a convincing case for the use of big data and cognitive learning to usher in a new age of hyper-personalised, almost frictionless service, but it still needs the right infrastructure.