Snowflake CIO: Why Enterprise AI Needs Engineering First
Mike Blandina has seen enough technology cycles to know when something fundamental is shifting.
After five years running payments technology at JP Morgan, he could have coasted into retirement.
Instead, he’s at Snowflake Summit 25 explaining why IT departments need to stop thinking like help desks and start thinking like engineers.
“I’ve probably spent more than the last 30 years in financial services, technology and product,” he said. PayPal, Google, Blackhawk Network: if you’ve moved money electronically in the past two decades, you’ve likely used something he helped build. “Basically, moving money electronically – that’s what I’ve done.”
The retirement question came up after he left JP Morgan. “My wife and I were candidly asking ourselves, ‘Should we retire?’ And then she said, ‘I’m just not sure our marriage will survive that right now!’”
So, here he is at Snowflake, drawn by something that takes him back to the beginning.
The data philosophy that universities forgot to teach
Mike cut his teeth on data in the 1980s, when E.F. Codd and C.J. Date were writing the rules for relational databases. Back then, British consultant James Martin preached a simple gospel: get your data right first, then build your applications.
“There was this mindset of ‘get your data right, then build your software,’” Mike recalls. “That sort of got lost in the universities starting in the late ‘80s and certainly through the ‘90s and 2000s. It just really wasn’t taught anymore.”
Universities stopped teaching data modelling and engineers stopped thinking about structure. “It became ‘let’s write code and then we’ll figure out what to do with the data.’”
The result? Data chaos. “I think that led to an explosion of data lakes and data warehouses and SQL and NoSQL databases – just data everywhere.”
Now companies spend millions cleaning up the mess. “Frankly, I think this has helped create an entire industry with Snowflake and others about getting that right again.”
Why IT departments need to stop processing tickets and start solving problems
Ask most CIOs what they do and you’ll hear about service levels, vendor management and security protocols. Mike heard the same thing for years.
“Even 10 years ago, the role was to get the enterprise applications correct, service your internal customers, get your rights and privileges and privacy and trust and safety right, protect the company's assets and deliver a good product internally,” he says. “I think a lot of that over time turned into mostly ‘buy or rent software’ for a while.”
At Snowflake, he’s flipping the script. “What I’m doing is basically shifting my team from a ticket mindset to really a solution mindset – we’re going to bring engineering back into the CIO office.”
No more ticket queues. No more service requests. “It’s no longer ‘I filed a ticket, would you please service me?’ It’s ‘What are you trying to achieve? What problem are you trying to solve, and how can we help you solve that?’”
The timing matters. When AI can potentially replace entire applications, buying another SaaS tool might not be the right answer.
“The answer may not be a new app – the answer may be an AI agent that replaces that app.”
Even Snowflake, only founded in 2012, has accumulated technical debt. “I thought when I was joining I wouldn't have much legacy here, but that wasn’t true. It’s not old legacy, but in a fast-growing company, you end up buying three of this and two of that.”
How Snowflake builds its own tools on its own platform
Snowflake runs on Snowflake: called Snowhouse internally.
“Probably 60% of our stack is built on Snowflake. The stuff that’s not would be things you’d expect: horizontal global tools that no one wants to rebuild from scratch anyway.”
Even when Snowflake uses third-party tools, these are extended with Snowflake.
Take its onboarding system, Lift. The core might be Workday, but “all of those extensions were built with Snowflake,” Mike says.
“It literally is our core toolset for building new applications for the enterprise.”
Now, Snowflake is building AI agents internally and rolling them into its products.
“Some of the AI agents we’re writing, we’re putting into Snowflake Intelligence. Some of those will become public – not all of them will – but certainly our customers can learn from what we're doing, and we want to learn from what they're doing as well.”
Case in point: they built a custom employee AI tool and now they’re killing it. “We built an employee AI tool and we’re actually shutting it down. That capability is moving into Snowflake Intelligence in the next couple of weeks.”
Enterprise AI can’t afford to hallucinate about revenue numbers
Consumer AI operates in a world of creative possibilities, but enterprise AI lives in a world of absolutes.
“It’s pretty easy to take a photo off my phone and ask ChatGPT to create a colouring book for my grandson from that photo. It works. There’s no right and wrong to it – there’s no perfectly right answer.”
Try that in enterprise software, and you’re fired.
“When a salesperson or our sales leader wants to understand current sales trends and revenue up to a certain day of the week, the answer has to be right. It can’t be wrong. There’s no room for hallucination.”
Then there’s access control. “Does the person asking the question have a right to the answer? Does the person asking the question have a right to the data to get the right answer?”
Cases like this are daily realities in enterprise IT. “There’s just a depth of these topics that are actually going to get addressed in the enterprise before they get addressed in the broader SaaS world.”
Building AI systems that eventually make the CIO redundant
Five years from now, Mike wants to be unemployed – by design.
“My objective is to work my way out of a job. I think I do that by adding solutions that are AI-enabled and learn and grow. We take what we did with Snowflake on Snowflake, and we do AI on AI.”
It’s not about automation for its own sake.
“The more I do of that, I think the faster the company can be nimble internally to support our growth externally. And then, eventually, the bots will run the world, and I'm going to retire!”
Meanwhile, the industry oscillates between panic and euphoria. “The technology is evolving faster than companies are. That’s always true, but in the age of AI, it's going so fast.”
He’s seen both extremes. “A CEO says, ‘I saw this really cool demo, and you should build a bot,’ or someone at JP Morgan says, ‘We should fire 50% of our engineers because bots are going to write the code for us.’ You’ve got to bring that back to reality fast.”
His prescription? Run two tracks simultaneously. “It’s almost like you've got to split your company into two brains – one saying ‘go forward’ and you catch up behind them, cleaning all this up as fast as possible. It’s one of the reasons I’m at Snowflake, because I think that catch-up is exciting as well.”
