The unstoppable march of AI continues to assemble tempo. Analyst Gartner not too long ago forecast that half of all enterprise selections will likely be absolutely automated or a minimum of partially augmented by AI brokers throughout the subsequent two years.
Some organizations have experimented greater than others. 4 enterprise leaders who’ve explored AI shared classes discovered at a latest media roundtable occasion at Snowflake Summit 2025 in San Francisco. Here is what they needed to say.
1. What’s my cloud technique?
Wayne Filin-Matthews, chief enterprise architect at AstraZeneca, defined how his group is pioneering AI implementations in a number of areas.
The pharma large has developed an AI-enabled analysis assistant that enhances the productiveness of scientific researchers by specializing in the reproducibility of scientific strategies and the event of latest medicines.
AstraZeneca companions with main educational establishments, comparable to Stanford College, to run agentic AI experiments.
“We’re occupied with how one can have a group of brokers that may help the standard scientists who do their analysis,” mentioned Filin-Matthews.
The corporate can be exploring the way to apply AI in industrial areas. AstraZeneca operates in 126 markets, and serving these assorted places with content material is a posh problem. That is the place AI is available in.
“We have leveraged the know-how from an AI perspective to automate the creation of selling materials and details about drug improvement,” he mentioned.
Whereas these experiments have highlighted the advantages of AI, they’ve additionally proven the significance of stable knowledge foundations.
Filin-Matthews mentioned corporations can solely clear up issues with AI in the event that they’ve constructed a robust underlying cloud infrastructure.
“There are such a lot of use circumstances the place the profit is turning into clear as we have gone on this journey,” he mentioned.
“We’re positively within the period of AI-enabled decision-making. However the important thing for me is you possibly can’t neglect these different underlying parts. You can’t be AI-first with out being cloud-first.”
2. Have I addressed knowledge governance considerations?
Amit Patel, chief knowledge officer for wholesale banking at Truist, mentioned he discovered two key classes from rolling out AI use circumstances.
Primary was the significance of the underlying knowledge basis.
“As a financial institution, we’ve to show, ‘The place did the info come from? Is it appropriate? Is it ruled? Do I’ve lineage? Do I’ve metadata? Do I’ve knowledge high quality checks?’ I’ve to show these factors to an exterior regulator,” he mentioned.
“I am unable to simply launch a big language mannequin (LLM) into the wild, proper? And I am unable to level it at simply any sources that I’ve internally. It is obtained to be a ruled supply. It is obtained to be a certified provisioning level.”
Patel mentioned this deal with regulated sources helped elucidate a standard drawback level for CDOs: getting your knowledge so as.
“By means of that course of, I’ve found that I haven’t got as many dependable sources as I wish to level to,” he mentioned. “I’ve obtained to allow that basis first, after which I can construct on high.”
Patel mentioned the second factor he discovered is that individuals who use AI at residence assume it is going to be simple to deploy LLMs in an enterprise surroundings.
“It is not that straightforward,” he mentioned. “You need to outline guardrails round what the fashions can take a look at. It’s best to outline the metadata to information the fashions’ interpretations. And that course of takes time.”
Patel mentioned his group has addressed employees misconceptions in regards to the time to take advantage of AI by way of expectation-setting workout routines.
“As we have began to allow use circumstances, folks have began to know that it is not as simple as a point-and-click course of,” he mentioned.
“Whereas implementing know-how is quicker than it was, it is nonetheless difficult, and it requires time and thought round how you place governance and construction round AI earlier than you allow it for work.”
3. What is the high quality of my outputs?
Anahita Tafvizi, chief knowledge and analytics officer at Snowflake, mentioned her group helps the tech firm develop the AI-enabled merchandise its prospects use.
Nonetheless, Tafvizi mentioned her firm does not simply promote these merchandise — the group additionally will get to experiment with these applied sciences.
“The attention-grabbing factor about being the CDO at an information firm is that I get the privilege of being the very first buyer of a variety of our merchandise,” she mentioned.
Tafvizi drew consideration to Snowflake Intelligence, a know-how launched at Summit that enables enterprise customers to create knowledge brokers.
Her group partnered intently with the product group to develop an AI-enabled assistant for the interior gross sales group.
She acknowledged that implementing new AI instruments brings challenges, notably in relation to balancing the speed of innovation with governance necessities.
One essential consideration is high quality. As her group pushed the software to the gross sales group, they contemplated vital questions, comparable to, “Is 95% high quality adequate?”
Tafvizi suggested different enterprise leaders to think twice about these challenges, as employees should belief the outputs of AI experimentation.
“The deal with high quality has been vital for us,” she mentioned. “The precise governance constructions, entry controls, lineage, metadata, and semantic fashions are additionally important. We consistently take into consideration these issues as a part of the stress between innovation and velocity.”
4. Have I thought-about unanticipated advantages?
Thomas Bodenski, chief knowledge and analytics officer at finance know-how specialist TS Think about, mentioned his firm has been utilizing AI to cut back worker workloads since October 2023.
Nonetheless, whereas the main target of AI is usually on automating guide processes, his experiences recommend enterprise leaders ought to acknowledge the know-how additionally produces different advantages.
“Utilizing AI isn’t just about lowering effort,” he mentioned. “You get to do issues quicker, higher, and have an unbelievable protection enchancment as nicely.”
He defined how TS Think about buys knowledge from specialist distributors that ship emails about upcoming product adjustments.
The corporate receives 100,000 of those emails a yr. Every electronic mail must be learn and its implications understood. Historically, that work-intensive course of has consumed, on common, two and a half full-time equivalents per yr.
“It is nerve-racking as a result of you possibly can’t make errors,” he mentioned. “If we miss data in an electronic mail, our programs will go down. Hundreds of merchants can’t commerce and hundreds of danger managers cannot assess their publicity, so it is doubtlessly catastrophic.”
To keep away from this state of affairs, Bodenski mentioned the corporate makes use of Snowflake’s AI fashions to finish this time-intensive work.
“Now, we by no means miss the consequence,” he mentioned. “These two and a half full-time equivalents can do data work somewhat than guide knowledge curation or entry.”
Bodenski mentioned AI may handle what was beforehand a weak spot: guaranteeing buyer requests are handled on Saturdays.
“No one labored on these days. Now, there’s AI, and she is going to reply to buyer inquiries and assign the ticket to the suitable individual,” he mentioned.
Need extra tales about AI? Join Innovation, our weekly e-newsletter.
