When Michael Dell stepped onto the keynote stage at Dell Technologies World 2024 and said that the past 40 years of the organisation’s history were essentially the “pregame” for the world of artificial intelligence (AI), the people who helped him build the company from its humble beginnings into a global technology powerhouse could perhaps be forgiven for wondering what they’ve been doing with their lives.
But there is undeniable truth in the billionaire CEO’s bombastic statement. AI could yet prove to be a technology that fundamentally remakes the industry, even if some of the use cases still feel a little shoehorned in – in one session a roomful of Dell channel partners who had presumably come to hear about rebates on storage were also treated to an AI songwriting session with artist and advocate Taryn Southern.
Whatever, Dell, still very much at the wheel of his eponymous ship, is rightly or wrongly responding to AI by setting the engine order telegraph to Full Steam Ahead and ploughing straight into whatever is coming. His crew will be hoping it isn’t an iceberg.
For the passengers, this meant three days of full AI immersion as they bore witness to the formal launch of Dell’s AI Factory tie-up with Nvidia, new hardware debuts including PCs with onboard Microsoft Copilot+ as well as AI-ready storage and networking gear, and expanded AI partnerships with the likes of Hugging Face and Meta.
It’s a lot to take in piece by piece, but for those IT buyers who might be inclined to start sounding out if AI is for them, the Dell message is clear: we want to be the guys who tie it all together. Full-service AI for the enterprise, if you will.
Nick Brackney, senior consultant for generative AI at Dell, has the captain’s back. “When I look at what we’ve assembled so far as a company, it’s kind of interesting how AI really makes it all make sense,” he says.
“You used to have the client business, and then you had the infrastructure side of the house, and there was not much overlap. In my role, I’ve been messaging for all of Dell’s generative AI and it’s all coming together.”
Brackney, an avowed AI optimist, says the closest parallel in tech history to the change that AI is bringing is the advent of the smartphone in the 2000s, and it’s dwarfing even that paradigm shift.
“There’s now an entire generation of people who have never been lost, right? Because they always have a GPS on the phone. The only time you’re lost is if your battery’s dead,” he says.
“There’s going to be another generation of people who are always going to know the answer to every question because they’ll be able to ask the AI. Malcolm Gladwell talks about 10,000 hours to be an expert in everything. I think this kills that because I don’t have to be that expert. I just have to be an expert at one thing, which is working with the AI. It augments everything I do.”
As they move through this world, Brackney sees an opportunity to help Dell’s customers understand and right-size their AI needs, and wants to help them drill down to what their personalised solution will actually be – this will be crucial because in this regard, every organisation’s needs are distinct, or as he puts it “the use cases you hear about are all snowflake use cases”.
“We start off and ask what are you trying to accomplish and what is your need? How many users are we talking about? How much data are we talking about? That’s the first step. Then we go, okay, what is the role I need AI to play? Is it content generation? Is it code generation? We really narrow down on that,” says Brackney.
“And only then will we get into what kind of model to use. Is it fine-tuning? Is it RAG [Retrieval Augmented Generation]? Is it a pre-trained model? The reality is that the infrastructure required is very different depending on what you choose.”
For example, an organisation building a use case using a pre-trained model, such as Microsoft Copilot, may not need to do much more than buy a few of the newly launched AI PCs. But say the model is not delivering the right results and the customer needs to add their own data to it, the conversation then needs to encompass data preparation, more advanced services, backend storage and so on. Customers who want to get into fine-tuning or create their own models will go further still.
“The nice thing is that knowing this framework and starting with what they’re trying to accomplish, I can put somebody onto the right product for them and they’re not over-provisioning, they’re not over-buying,” says Brackney.
“To be able to do that from the laptops to workstations, all the way through to our servers and our storage, who else has the ability to do that? Not HP.
“Our entire portfolio is coming together to solve this problem, no matter what flavour of AI you need. That’s really important,” he adds.
Inside the AI tech stack
Dell introduced five new notebook AI PCs at Dell Technologies World, powered by Qualcomm Snapdragon X Elite and Snapdragon X Plus processors and featuring Microsoft Copilot+, which was launched at a separate Microsoft event, concurrently with Dell’s on 20 May 2024.
The five models, XPS 13, Inspiron 14 Plus, Inspiron 14, Latitude 7455 and Latitude 5455 are designed to “deliver exceptional speed and AI performance to elevate computing and simplify tasks.” They process AI locally on a custom-integrated Qualcomm Oryon CPU, premium GPU and neural processing unit (NPU) that delivers – so Dell claims – 45 trillion operations per second.
Brackney predicts significant interest in these product lines on the basis that as more tech companies embed AI into their products, they will want users to do the processing locally, on the basis that hosted AI will be super expensive in terms of infrastructure and the energy needed to run it.
“The benefit of this is that I want to be able to do this, for all my knowledge workers, make them more efficient, more productive. I don’t need data scientists for that, I don’t need to go and invest in a lot of infrastructure. I just need to buy them a higher-end PC,” he says.
If Dell’s AI PCs are where the average office worker in finance, HR, marketing or sales will experience its AI suite, the full service experience is the next level above that. Consider an organisation that wants to build a dedicated AI application for its end-users and needs to do more fine-tuning or model training on order to do so. This will require much more processing power, clearly, because the footprint for AI training is much larger than the footprint for inferencing.
You might think that that renders it a solution for technology companies alone. But, as Brackney points out, what constitutes a tech company, and aren’t all companies now tech companies to some degree?
“I remember a couple of years ago, I was listening to an analyst call from Walmart, and they said they were a tech company, and somebody actually laughed on the recording. [But] if you look at Walmart, they have so much data and they’re really strong with it. It’s allowed them to compete head to head with Amazon,” he says.
He predicts that many organisations will eventually realise that it makes more sense from a value standpoint to build or train their own, smaller AI model on their own data, at which point Dell can fling open the doors to the full AI product and services set.
“What’s nice is we can sit down with the customer and our services folks, and we can help them walk through everything from getting the strategy set, which use cases to start with, all the way through deploying and tuning the models, so it’s a big opportunity for us,” he says.
Building guardrails
Brackney also sees opportunities to help customers come to terms with, and appropriately address, some of the thornier ethical issues and questions that arise with the use of AI.
This need will of course vary from customer to customer, particularly if they take advantage of the full-service option and run their own AI workloads in their own datacentre. However, this should be the moment to formalise what might have been a rather more ad hoc approach.
“I talk to a lot of customers and we do these GenAI days where we’re teaching people a lot of these things. The scary thing is I ask people how many of you have formal policies in place, and most of them don’t,” says Brackney.
“The research that we’ve seen is that people are actually already embracing shadow AI – they’re just going to bring their own models in. That’s a scary thing because unlike shadow IT where it was only my developers I had to worry about, now every single person in the company is potential point of failure right from the admin to the executive. So, it’s really important to figure this out.”
In the spirit of wanting Dell to be the best partner it can be to its AI customers, Brackney says that he is already sitting down with them, sharing guidance and advice, and working on responsible use of AI holistically.
“One of the things that we offer right now is a four hour assessment where we help them start to understand the strategy, because that’s so important to get there,” he says. “This kind of thing is likely to come up as part of that services engagement – we will sit down with them and ask, what are your concerns? What are things that we need to be thinking of?”
Beta customers
Of course, none of this would be launching at all had Dell not run extensive pilots with test customers. “Probably my favourite one of those is the city of Amarillo [in Texas],” says Brackney.
“It’s one of those altruistic stories where you have you have a government entity, and I think within their city limits 25% of the people don’t speak English as a first language. When you hear that, you think Texas, they’re probably Spanish-speaking, and when you find out there are actually 65 languages, that’s a tall order.”
Amarillo has built itself an AI chatbot on Dell’s technology, which is now able to offer information and guidance on the city’s services in every one of those languages.
“Now they have like the ability to go and ask questions and talk to it and learn, so it’s making all the city’s services more addressable in a better way, for their constituents. I think that’s pretty cool,” says Brackney.