Enterprise AI, Generative AI, Build vs Buy, and Data Best Practices from AI Infrastructure Field Day

I'm Alistair Cook Event Lead here at Tech Field Day, and here are my takeaways from AI Infrastructure Field Day from September of 2025. This edition of AI Infrastructure Field Day really felt like a transition from the idea that you are building something to, you're actually delivering something that AI applications aren't really about all of the learning phases at the beginning, but your AI application needs to deliver for the business to make some difference to how the business operates and to do that, in this particular edition of the AI Infrastructure Field Day, we saw a lot of platforms to help you smooth that process to get out into production. One of the big challenges in AI, and particularly generative AI applications, is to move from the experimentation phase.

And we've seen lots of demos of the experimentation phase that are characterized as by lots of manual processes and lots of Jupyter Notebooks showing you what can be done rather than the actual production deployment. And that transition from experimentation to production was definitely one of the things that I saw as a focus for the vendors who were presenting for us at AI Infrastructure Field Day, we saw much more of an emphasis on having a process that is automated, having a platform that eases the transition to production deployment, and particularly thinking about having a population of AI applications rather than just a single AI application that you typically focus on when you're in the early introductory phases. And so it was interesting to see far more maturity and, and focus and, and maturity of use for the organizations that are actually adopting AI and providers we have that are gonna help that As a significant part of that move to production.

There was a, an emphasis on whether you should build or your own solution, or whether you should actually buy a platform. A number of the vendors that we had at AI Infrastructure Field Day have open source tools underneath or at the very core of what they do, and then build those out to being more of a automated product or a platform. And so there's some interesting thoughts about whether you could just use those same open source tools and invest some technical expertise and some time in order to use those tools and just build your own platform.

But what we're overwhelmingly hearing is that companies who go down that path have a very long path to actually get applications operational and get large numbers of AI applications operational. And that often the time to value that you can get from just buying a platform. Buying an integrated solution is really valuable to getting those AI applications out in front of the users and actually having the investment in AI turn into an improvement in business and a return on that investment.

So the classic build versus buy, we definitely saw a whole collection of vendors who believe that buying a solution that rapidly gets your infrastructure out, gets your infrastructure stood up and gets you onto delivering value to business is a, a very sensible way to approach them. A lot of the companies that have been trying to build their own end up having problems and being unable to deliver their applications in the timely manner. The other major theme that we've had is one that we see right throughout itt, and we always have the good quality data is going to be really important in having good quality outcomes.

Having access and having timely access to the data that your application requires, that you are feeding into your AI application, whether you are feeding it into fine tuning of some open source model, or if you are just building out a rag solution that uses, uh, vectorized versions of your own documentation. In both of these cases, having the right data and having access to the right data in the right place is vital in order to be able to build your AI application and to make it specific to your organization rather than just having some generic chatbot in the bottom right hand corner of your website. So garbage in ai, garbage out is definitely an an element that's really important as you're building an enterprise application.

We really wanna have good quality, uh, data coming in in order to have good quality data coming out. AI infrastructure is reflecting the whole enterprise use of ai. We're transitioning from the early experimentation phases.

The times when we looked at the very low level pieces of architecture, we are seeing much more of a, a move towards the software being really vital in here. Having platforms that enable you to build applications using your own data and to build them in a timely manner to deliver value to your organization seems to be a, a really vital part of what's happening with, uh, enterprise adoption of AI and enterprise build outs of the infrastructure to support that ai. Really nice to see Refa, uh, as a presenting company and, uh, they were first time presenting with us, although, uh, Hasid who was the, um, uh, present our CEO has been a, a previous presenter at Tech Field as well.

Also first time presenting Marti. Surprisingly, we haven't had Martis presenting with us beforehand. Both Refa and Martis showed great platforms for building out your entire AI infrastructure.

Of course, we were joined by the wider Futurum group. We had Brian Martin from the Signal six five team and Guy Courier who's an analyst with, uh, Futurum as well. Great to have the Futurum team here sharing their insights and their expertise alongside the collection of awesome delegates that came to join us in, uh, Santa Clara this week.

Thanks for watching this episode of the Tech Field Day takeaway series on the Tech Field Day plus YouTube channel. If you enjoyed it, make sure to like, subscribe and share your thoughts on, uh, Colin, uh, follow Tech Field Day on X Twitter, blue Sky Mastodon, and all of your other, some favorite social medias. And check out the presentation videos on the Tech Field Day website, as well as the Tech Field Day YouTube channel.

We'll see you at the next Tech Field Day event, which will be Security Field Day 14 happening September 24th to 25th.