Follow ZDNET: Add us as a preferred source<!–> on Google.
ZDNET’s key takeaways
- Be warned: give AI rubbish data and it’ll give you rubbish back.
- Take a strategic approach to data collection that focuses on gold dust.
- Be flexible, as language models and business demands can change.
Evidence suggests it’s tough to generate value from AI projects, but one thing we can be sure of is that successful initiatives require data – and lots of it.
Whether it’s running a generative AI rollout or exploring agentic AI, the language models that power emerging technology solutions require access to vast informational resources. As businesses scale up their AI efforts through 2026 and beyond, having access to the right data assets has never been important.
Also: You should still learn to code, says top Google AI exec – here’s why
But how can you ensure your organization is collecting the right information for its AI initiatives? Five business leaders give us their tips.
1. Take a thoughtful approach
Paul Neville, director of digital, data, and technology at The Pensions Regulator (TPR), said his public body is “very thoughtful” about collecting data for emerging technology projects.
“The result from AI depends on the data it’s looking at,” he said. “So, if you give it rubbish data, it’ll give you rubbish back. That’s very clear.”
Neville told ZDNET that the foundational elements – good data practice, governance, and ownership – help ensure his organization turns the right information into insight.
Also: Is your company spending big on new tech? Here are 5 ways to prove it’s paying off
TPR also studies external developments from technology partners, such as OpenAI and Microsoft Azure.
“Their models are developing regularly, and these new releases will impact the results that we’re getting, and so we have to be aware of that innovation cycle and be careful to ensure we’re monitoring the results and adapting our own processes to get the right results,” he said. “We do a lot of work on that. We have created some new AI governance and strategy roles precisely for that reason.”
2. Focus on the critical 20%
Ian Ruffle, head of data and insight at UK auto breakdown specialist the RAC, said professionals shouldn’t get too hung up on attempts to second-guess the types of information that will power future innovations.
“I think you should focus on what’s important to your business today,” he said. “I reckon it’s better to do fewer things well than to spread yourself too thin.”
The RAC is using Snowflake’s AI Data Cloud technology to consolidate what were once fragmented data silos into a foundation for insight-enabled projects.
Also: AI will cause ‘jobs chaos’ within the next few years, says Gartner – what that means
Ruffle told ZDNET that the storage capabilities inherent to modern data platforms mean some professionals could decide to keep everything just in case.
“It’s quite cheap nowadays to chuck data somewhere and store it,” he said. “So, maybe you could take that approach if you think data could have some value in the future.”
However, Ruffle recognized that panning data stores for gold can be a resource-intensive exercise, and he suggested it’s smarter to keep a tight grip on information.
“When you’re building forecasting models or a use case for a call center process that’s trying to aggregate information or process a complex complaint and using AI to gather data, it’s better to focus on the 20% of information that is important.”
3. Create a flexible strategy
Dominic Redmond, CIO at recruitment specialist PageGroup, said business leaders must create a forward-thinking and flexible strategy for data collection.
“Someone said to me that you can either wait for people to tell you what the impact will be, or you can try to come up with a plan for what the impact will be and manage it,” he said. “I think the same is true about data. Ultimately, you’ve only got the data in your organization. So, I think the key question is, ‘Which bits of that data are going to be most valuable?'”
Also: No ROI in AI yet? Try these six proven tactics for creating real business value
Redmond recognized that answering that question effectively is easier said than done, especially in the age of AI.
“We don’t know, when we’ve decided which data we might need in year one, whether it’s going to be the same as year two or year three,” he said. “We also don’t know whether AI can help us get that data in a different way than maybe we have been able to in the past.”
Redmond told ZDNET that the best approach is to create a plan that identifies the most important data for future projects, with room to adapt to fresh business requirements.
“Success is about being data savvy and data-led but also making sure that you’re capturing it all for whatever may come next and then being flexible enough as you need to be as the market and industry changes.”
4. Find the gold dust
Sacha Vaughan, chief supply chain officer at homeware manufacturer Joseph Joseph, said it’s crucial to keep one eye on long-term goals and the other on current priorities.
As part of this process, she said it’s best to tie data storage to clear business processes, such as her attempts to use insights to refine supply chain operations.
“In the future, I think the challenge of collecting everything might be prohibitive,” she said. “How do you collect everything? What data storage would it take to collect everything?”
Also: Anxious about AI job cuts? How white-collar workers can protect themselves – starting now
Vaughan told ZDNET she’s eager to understand how data, AI, and analytics can be used to improve business processes – and she gave an example of how this approach might work.
“It would be great if we could find a way of collecting and analyzing customer insights, down to the minute details, because, even from customer reviews and complaints, there’s some gold dust in there for design as well,” she said. “That’s about getting a mechanism to mine that data and feed it back adequately at a granular level, to a designer who can pick up a customer query about a part of the product, and that gives them an idea of how to design in the future.”
5. Consider semantics carefully
Steve Lucas, CEO of technology firm Boomi, said the question digital and business leaders should consider is “How do I store the right data that matters for AI?” – and finding a suitable solution to that challenge is a matter of semantics.
“You’ll likely have all the data you need. Data is like sand. But do you have the context for that information?”
Lucas told ZDNET that effective cataloging and tagging can help professionals find the data that matters.
“Can you query a database and get an exact response to your question? Absolutely. But if you’re looking for similarities in data, it’s those similarities that really matter. So, in that situation, catalogs and metadata will be more valuable than the raw data itself.”
Lucas said the big tech companies can help business leaders ensure their organization focuses on storing the right data for AI: “They’re spending billions. Let those billions save you money and guide your path.”
