The unstoppable march of AI continues to gather pace. Analyst Gartner recently forecast that half of all business decisions will be fully automated or at least partially augmented by AI agents within the next two years.
Also: 4 ways to turn AI into your business advantage
Some organizations have experimented more than others. Four business leaders who have explored AI shared lessons learned at a recent media roundtable event at Snowflake Summit 2025 in San Francisco. Here’s what they had to say.
1. Get your cloud strategy right
Wayne Filin-Matthews, chief enterprise architect at AstraZeneca, explained how his organization is pioneering AI implementations in several areas.
The pharma giant has developed an AI-enabled research assistant that boosts the productivity of scientific researchers by focusing on the reproducibility of scientific methods and the development of new medicines.
AstraZeneca partners with leading academic institutions, such as Stanford University, to run agentic AI experiments.
“We’re thinking about how you can have a team of agents that can support the traditional scientists who do their research,” said Filin-Matthews.
Also: 4 ways your organization can adapt and thrive in the age of AI
The company is also exploring how to apply AI in commercial areas. AstraZeneca operates in 126 markets, and serving those varied locations with content is a complex challenge. That’s where AI comes in.
“We’ve leveraged the technology from an AI perspective to automate the creation of marketing material and information about drug development,” he said.
While these experiments have highlighted the benefits of AI, they’ve also shown the importance of solid data foundations.
Also: Integrating AI starts with robust data foundations. Here are 3 strategies executives employ
Filin-Matthews said companies can only solve problems with AI if they’ve built a strong underlying cloud infrastructure.
“There are so many use cases where the benefit is becoming clear as we’ve gone on this journey,” he said.
“We’re definitely in the era of AI-enabled decision-making. But the key for me is you can’t forget those other underlying elements. You cannot be AI-first without being cloud-first.”
2. Focus on data governance concerns
Amit Patel, chief data officer for wholesale banking at Truist, said he learned two key lessons from rolling out AI use cases.
Number one was the importance of the underlying data foundation.
“As a bank, we have to prove, ‘Where did the data come from? Is it correct? Is it governed? Do I have lineage? Do I have metadata? Do I have data quality checks?’ I have to prove those points to an external regulator,” he said.
“I can’t just release a large language model (LLM) into the wild, right? And I can’t point it at just any sources that I have internally. It’s got to be a governed source. It’s got to be an authorized provisioning point.”
Also: This free Google tool turns AI into your research assistant
Patel said this focus on regulated sources helped elucidate a common problem point for CDOs: getting your data in order.
“Through that process, I’ve discovered that I don’t have as many reliable sources as I would like to point to,” he said. “I’ve got to enable that foundation first, and then I can build on top.”
Patel said the second thing he learned is that people who use AI at home assume it will be easy to deploy LLMs in an enterprise environment.
“It’s not that simple,” he said. “You have to define guardrails around what the models can look at. You should define the metadata to guide the models’ interpretations. And that process takes time.”
Also: Is your business AI-ready? 5 ways to avoid falling behind
Patel said his team has addressed staff misconceptions about the time to exploit AI through expectation-setting exercises.
“As we’ve started to enable use cases, people have started to understand that it’s not as easy as a point-and-click process,” he said.
“While implementing technology is faster than it used to be, it’s still challenging, and it requires time and thought around how you put governance and structure around AI before you enable it for work.”
3. Consider the quality of your outputs
Anahita Tafvizi, chief data and analytics officer at Snowflake, said her team helps the tech company develop the AI-enabled products its customers use.
However, Tafvizi said her company doesn’t just sell these products — the organization also gets to experiment with these technologies.
“The interesting thing about being the CDO at a data company is that I get the privilege of being the very first customer of a lot of our products,” she said.
Tafvizi drew attention to Snowflake Intelligence, a technology launched at Summit that allows business users to create data agents.
Also: The top 20 AI tools of 2025 – and the No. 1 thing to remember when you use them
Her team partnered closely with the product team to develop an AI-enabled assistant for the internal sales organization.
She recognized that implementing new AI tools brings challenges, particularly when it comes to balancing the velocity of innovation with governance requirements.
One crucial consideration is quality. As her team pushed the tool to the sales team, they pondered important questions, such as, “Is 95% quality good enough?”
Tafvizi advised other business leaders to think carefully about these challenges, as staff must trust the outputs of AI experimentation.
“The focus on quality has been important for us,” she said. “The right governance structures, access controls, lineage, metadata, and semantic models are also critical. We constantly think about those things as part of the tension between innovation and velocity.”
4. Look for unanticipated benefits
Thomas Bodenski, chief data and analytics officer at finance technology specialist TS Imagine, said his company has been using AI to reduce employee workloads since October 2023.
However, while the focus of AI is often on automating manual processes, his experiences suggest business leaders should recognize the technology also produces other benefits.
“Using AI is not just about reducing effort,” he said. “You get to do things faster, better, and have an unbelievable coverage improvement as well.”
He explained how TS Imagine buys data from specialist vendors that send emails about upcoming product changes.
Also: 10 strategies OpenAI uses to create powerful AI agents – that you should use too
The company receives 100,000 of these emails a year. Each email has to be read and its implications understood. Traditionally, that work-intensive process has consumed, on average, two and a half full-time equivalents per year.
“It’s stressful because you can’t make mistakes,” he said. “If we miss information in an email, our systems will go down. Thousands of traders cannot trade and thousands of risk managers can’t assess their exposure, so it’s potentially catastrophic.”
To avoid this scenario, Bodenski said the company uses Snowflake’s AI models to complete this time-intensive work.
“Now, we never miss the result,” he said. “Those two and a half full-time equivalents can do knowledge work rather than manual data curation or entry.”
Bodenski said AI can also manage what was previously a weak spot: ensuring customer requests are dealt with on Saturdays.
“Nobody worked on those days. Now, there’s AI, and she will respond to customer inquiries and assign the ticket to the right person,” he said.
Want more stories about AI? Sign up for Innovation, our weekly newsletter.
Leave a Reply