“We all just completed corporate AI training that told us everything we couldn’t use,” shared a CMO from a $6B conglomerate, who added sarcastically, “and that was fun.”
The huddle shuddered while I pondered the fate of big companies.
To be clear, I’m not anti-governance. Big companies have big risks, big brands, and big legal departments. A wrong email from a rogue agent is not a cute learning moment.
But corporate AI training that only teaches avoidance is not training. It is legal self-protection wearing a learning management system badge.
Congratulations, you now know 47 ways to get fired and zero ways to grow faster.
Why Big Companies Struggle to Move Fast
This is where the Innovator’s Dilemma sails into the harbor. Incumbents often protect the current business, the current process, and the current definition of “safe” until the new thing looks too messy to fund.
Then the new thing becomes the market.
Years ago, my company produced a video game called Carrier: Fortress at Sea, which meant I learned more than expected about aircraft carriers. One fact stuck with me: a carrier needs roughly five miles to come to a complete stop.
That is not a steering problem. That is physics.
Large companies have the same issue. They have scale, data, customers, brand equity, and distribution. They also have inertia, compliance layers, procurement drag, approval loops, and a thousand smart people trained to slow things down before something breaks.
So no, a $6B conglomerate will not turn like a speedboat.
But it can launch one.
Marketing Needs Its Own AI Skunk Works
That is the opportunity for CMOs at large companies right now. Corporate can and should create guardrails, but guardrails are not a go-to-market operating model.
Marketing needs its own AI skunk works.
Not a rogue team hiding from IT. Not a prompt-sharing Slack channel. A real skunk works. Sanctioned. Structured. Business-outcome obsessed.
Give it a mandate: improve pipeline, conversion, retention, customer experience, and cost per opportunity. Give it permission to rebuild workflows, not just add AI sprinkles to broken ones. Give it its own training program because “don’t paste confidential information into ChatGPT” is not a curriculum.
Snowflake offers a useful counterexample. Denise Persson’s team is building AI fluency through training, hackathons, AI goals, and governed agents. She has also cited a 30% reduction in cost per opportunity.
That is the big-company dream: carrier-sized data, speedboat-style learning.
My favorite part of Carrier: Fortress at Sea was trying to land an F-15 on the carrier. If you crashed, the game delivered a brutal little message:
“Congratulations, you just crashed a perfectly good $30 million airplane!”
AI Fluency Requires Some Trial and Error
That is what many corporate AI programs feel designed to prevent. No crashes. No mistakes. No mess.
Fair enough. Nobody wants to crash the plane.
But at some point, someone still has to learn how to land it. You may not be able to stop the ship in under five miles.
But you can launch something from the deck today.
Written by Drew Neisser