Joe Green (TechForge): Okay. So first of all, let’s talk about Sunil Murty. How did you end up at IBM? Are you an engineer by trade or are you a sales guy? What’s your background?

Sunil Murthy (IBM): Computer science engineer. Great. And I was a developer. I was a Java developer actually.

Joe Green (TechForge): That ages you rather, doesn’t it? Java, you say?

Sunil Murthy (IBM): Yeah. I’ve been there. A lot of battle bruises!

Sunil Murthy (IBM): I joined IBM to really work on the enterprise side of the house and been through the journey of service oriented architecture, cloud, big data, data science, and now really AI. So what’s really helped me navigate the IBM and stay at IBM is the fact that we have gone through these transitions with customers. And it’s really given me, somebody like myself, an opportunity to work through these technology transitions with our customers jointly. So never a dull day!

Joe Green (TechForge): So when you say working through these transitions, are we talking about transitions in terms of the structure of software in the enterprise? I was thinking a few years ago it was all about ERP, it was all about monolithic systems. Fast forward a few years, we’re talking about SaaS. And now all of a sudden we’re talking about hosting models back on premise. I mean, this has been an incredibly fast progression in terms of where software is deployed and AI of course is bringing something new to this.

Sunil Murthy (IBM): Absolutely. At the end of the day, customers want to drive their business outcomes with the most optimal technology choices they can make. And the technology choices they make is really the innovation at that ‘time capsule’, if you will, at that stage of the cycle of IT. And it so happens that these transitions that IBM has been working with customers with is to really realise the value of technology aligned to their business, aligned to the enterprise outcomes that the customers are driving. Let’s take AI. Everybody talks about AI. IBM talks about enterprise AI from the business alignment perspective. How do we employ AI to meet the customer’s objectives and really reimagine or redesign the business process with AI in the middle? Just like we did with when services were new, service oriented architecture was new, or the cloud came about. We went from a monolith, to your point, to a microservices world. IBM was right there to help with the transition during that journey with customers. So very similar transition.

Joe Green (TechForge): And you mentioned putting AI at the centre. Something I’ve been picking up on a bit on the show floor today, but generally talking to enterprise users of software is that the C-suite, let’s say, take them for example, they’re very excited about AI. And they want to put it at the centre of things. But the thing I’m hearing is that sometimes it’s the more peripheral elements, it’s the more boring elements that actually are what need to work. I mean, when we talk about governance, it’s dull because it’s: ‘Duncan from legal gets involved!’ Or we talk about clean data and sane data, or sanitise data and ‘Bob the DBA’ all of a sudden is involved. And actually these tasks are absolute foundation stones of AI and you can get excited about AI if you want, but it’s this other stuff that’s …

Sunil Murthy (IBM): Absolutely. Data is the important foundation and a lot of times we see that less than 1% of enterprise data, which is in the business firewall, is really used in the LLMs of today’s world. So what does that tell you? That tells you there’s a large amount of enterprise data that’s not necessarily participating in the AI transformations and the benefits of AI. So to your point, the C level can get excited – “I want AI yesterday.” But the reality is you need to first get your data ready, which is your enterprise data, ready for AI, which takes a lot of effort in terms of integration, in terms of creating that meaningful layer. We call that the contextual layer for agentic AI and enterprise AI, to use that in real time. So that’s a lot of work. To your point, it takes a lot of effort to get ready to really take advantage of the innovations in AI. And let’s also not forget the governance, the security, the policy framework, and really the overarching transparency of AI, which is a big question are right now.

Joe Green (TechForge): And this semantic layer, let’s define that. For me, semantics in this context would mean saying what data is, saying where it came from, saying who can access it, when they can access it, where they can access it from. Just creating that layer must be quite a big project?

Sunil Murthy (IBM): Semantics is all about relationships of data, and to your point, lineage, where the data came from is important. Data profiling with data quality becomes important, foundational elements, data policies, the agents can mimic human behaviour and can access sensitive data. So classification of this data based on the sensitivity and the confidentiality is important. So these building blocks create that layer which we call consumable data layer and provide that foundation for that semantics. Semantics is all about the meaning of data, the data about data.

Joe Green (TechForge): I was chatting to a fairly senior figure in the UK from Red Hat, of course, whom you acquired, oh, eight years ago now, I forget exactly. And he was talking about hosting LLMs, hosting AI on- premise. This appeared to be at least where Red Hat, UK were headed. Is this something that you see as being a viable trend not using the cloud, using the big frontier models from the big players? Are you advising customers to, “No, no, host on site, get your own hardware, lease your own hardware, lease your own data centre space, keep it in house.” Are you giving that advice to your customers or is it dependent really?

Sunil Murthy (IBM): Yeah, it’s really dependent. I would not really say either/or. It really comes down to the sensitivity of data and the industry the customer operates in terms of the… you’ve heard of sovereign data, especially in the UK and Europe and Middle East? It’s a big requirement, in terms of having the sovereignty of data and AI and operations. The independence from the cloud vendors is an important element. The fact that our foundation, our technology foundation is based off this hybrid in nature, hybrid by design, gives us that optionality and flexibility to put it in the sovereign data stack or sovereign cloud, if needed, or any cloud. We work with AWS, a very good partner of ours, Azure, or GCP, [with who] we announced a major partnership recently as well. So it gives customers optionality, but with a standardised layer, that hybrid-by-design foundation for taking that same contextual data layer for AI irrespective of wherever you want to deploy. So that standardisation gives customers that confidence, that tomorrow, if there is a requirement to move it back in- house, they can, from a cloud or take it from in-house to cloud. They can do that without a lot of moving parts, moving gear.

Joe Green (TechForge): And obviously there are several tracks on the show floor here, digital transformation, AI, cloud, cybersecurity, IoT (Internet of Things) and Edge, there’s been a lot of buzz about that on the floor on the show floor today in particular. I want to talk to you about this idea of real-time data or at least as close to real-time as you get, and working with real-time data with an AI. Is that something that the semantic layer, this contextual layer, is that something it can handle? So, it can handle older data archives and repos and pull in this fast, snappily-accessed data layer as well.

Sunil Murthy (IBM): It not only handles, but it excels. So when you think about data that’s historical as well as data addressed, and data in motion – which is IoT, edge, real-time data – you need that consistency in the data layer to provide the same semantic meaning whenever there is a request from an agentic AI [that’s] making that call. So that consistency, that standardisation, that’s what that contextual layer provides from IBM. And the best part is Confluent coming into the IBM family, integrating with watsonX data and watsonX orchestrate, you really have the technology abilities that are excelling in meeting the needs for today’s agentic AI with real-time contextual data.

Joe Green (TechForge): Yeah. It’s that Confluent plus the watsonX algorithm could be a winner! Sunil, thank you ever so much for sparing some time with me today and I wish you well in your travels [later today].