Teradata has been in the business of data warehousing so long, the term "data warehouse" didn't even exist back then, some four plus decades ago. Today Teradata is still a massive data company, but when CEO Steve McMillan took the helm in June 2020 he declard it would be a cloud-first business. That journey has paid dividends, with hundreds and hundreds of customers now part of its cloud business.

Teradata sees half a billion dollar in annual recurring revenue (ARR) through its cloud arm, and is growing over 30% year on year.

Major organisations are dependent on Teradata, and the cloud-first strategy "has given us the opportunity to demonstrate our Teradata platform can run mission critical workloads successfully in the cloud," McMillan said. "The success has built up a head of steam, and enabled us to demonstrate our maturity in this space."

Australia is a significant part of the Teradata story. "I'm really proud of the Australian team," he said. "The majority of our customers in Australia are on a modernisation journey with Teradata."

Now, you don't need to read too far in the news these days to find companies talking about cloud and modernisation. However, when it comes to Teradata these words are significant. Teradata customers are huge; the business is known for its on-premises strength since 1979. Finance, retail, telco, government, and other major industries live on Teradata with vast mountains of data. To help these organisations shift into hybrid operations with a cloud component is no mean feat, and is not without challenges of inertia. These are not overnight projects; these have complex rules around governance, providence, sovereignty, and more.

Yet, Teradata has made it happen. "Over 70% of our customers are operating in a hybrid environment today," McMillan said. "Still a lot is on-premises, but data sets are moving or have moved to the cloud."

What drives this is the desire to get the best out of that data. "Enterprise data warehousing has evolved over the last 30 to 40 years of taking data from all kinds of silos and putting it into one single store," he said. "But data has gravity."

"We have a capability called QueryGrid that enables customers to move a query to the data as opposed to moving all the data to the query. With QueryGrid we can have a Teradata ecosystem in Azure, in AWS, on-prem, and the QueryGrid enables you to send the query to the appropriate Teradata ecosystem sending back results without moving lots of data."

This is the way of modern companies. "We've found customers have a whole range of workloads, some of which require highly advanced dedicated storage mechanisms with Teradata on top to run super-complex workloads."

In fact, today's customers most likely "have 10x the data in native object stores than in structured data warehouses."

A major factor pushing companies to consider their data strategies is the real power of artificial intelligence, as seen in machine learning and generative AI, both made viable by the sheer power of today's compute engines.

"A Gartner study said over 90% of ANZ CIOs would have AI implementations in 2026," McMillan said. He's seeing huge growth in the Teradata platform being used from that AI perspective.

In fact, back in August 2022 - "just before the ChatGPT craziness in November 2022" - Teradata launched new AI and ML capabilies named ClearScape. "We've been helping customers take advantage of AI, and now GenAI, through the Teradata platform. We've seen super-interesting answers driving good use cases from Teradata customers," he said.

And over the last 12 to 18 months McMillan has definitely observed a maturing in AI projects. He's observed three important characteristics that measure if a project will be successful or not. This is valuable information from a global CEO, gleaned from the experiences of major organisations. These are:

1. The project must ensure trust. "Make sure it's training with trusted data inside your ecosystem to have more assurity over the outputs," McMillan said. Even if you don't want to have generative AI talk directly to your customers, you can still have it augment data for your staff. "We're seeing human-centric GenAI solutions that empower people inside the organisation, giving advanced capabilities."
2. The generative AI models must be ethical. "Trusted and ethical are super important," McMillan said. You must make sure no bias is introduced. "Some AIs are put in place with implicit bias in them. For example, against certain members of the population, either by economic stature or racial stature, or other. We make sure we can help customers with those ethical solutions. For example, you can have a capability to run A/B comparisons for advanced models."
3. The project must be sustainable. It has to have some green credentials; there's no point having a super advanced AI that costs so much to run it's taking your business backwards. Here Teradata can bring all its expertise to bear. "We help implement AI that can be cost effective, and cost efficient, leveraging Teradata technology to optimise language models," McMillan said.

These are the three factors that successful AI projects have in common. However, you might still be wondering what would be a good use case for AI in your company, in the first place. You know AI is the key to getting ahead, to unlocking the hidden value of your data to speed up customer service, to accelerate time to market, to create content quickly, and more. But what does this mean in a practical sense, in your specific situation? Again, McMillan has three observations from his experience. "These are the three horizons in terms of customers utilising AI," he said.

1. Improve the efficiency and effectiveness of people inside your company. This might be achieved by using commercially available AI features in tools you already use, such as the various Copilots being released.
2. Embed AI, such as generative AI, into your product or services. In this way you help your customers gain efficiency and effectiveness.
3. Identify a way to use AI to transform your industry. An example McMillan provides is that of Unilever; the chief data officer for Unilever told him how merchaniders would speak with buyers like Walmart to negotiate over the price of a bar of soap, for example. Now Unilever has developed advanced models in terms of demand and supply to help augment the merchandiser with information, while Walmart has done the same for its buyers. "It's like how high-frequency trading transformed the stock exchange. It moves to a battle of models. We find that really interesting," he said.

McMillan adds, "our approach in Teradata is not to lock a company into one kind of model. We believe in the future companies won't just use large language models with trillions of parameters, but will use small and medium language models with much smaller parameters but far more specialised in terms of results that come out."

"This will ensure better data quality, minimised hallucinations, bias removed, and better output sets."