Conventional wisdom says that you need clean data before you can use AI. But Rebecca Sykes isn’t so sure the data needs to be pristine.
Sykes, a partner at The Brandtech Group who leads its emerging technology practice, says generative AI can work with the kind of fuzzy, siloed and sometimes badly labeled data that used to stop projects cold.
“We’re less worried about messy data or bad data than no data,” she says on this week’s episode of AdExchanger Talks. Brand-specific data, even when it’s disorganized, is critical, because it’s what keeps AI outputs from turning into a generic sea of sameness.
“For brands, the biggest risk here is that AI takes you toward convergence rather than divergence,” Sykes says. “And divergence is the lifeblood of brands.”
But even with the right data, plenty of AI pilots stall at scale. To explain why, Sykes has a fable she likes to cite from a new business book about AI transformation, and it involves a frog and a sock.
Think of AI as a frog, which was born to leap. Now picture the frog stuck inside a sock, which represents a company’s existing operating model. The frog can try to leap all it wants, but if it’s trapped in a sock, it’s not going anywhere.
“In general, no one’s sock is wrong or broken,” Sykes says. “It was just built for a different time.”
For example, a client might use AI to reduce the amount of time it takes to do one part of a task by, say, 70%, but then be disappointed, because the entire task still takes four days to complete. Where’s the efficiency?
But the reason why it still takes so long is because the overall work still moves through multiple different systems and rounds of cutting and pasting, and the data in one system doesn’t always match the data in another.
In other words, AI sped up one step, but the process around it never changed. “You have to look at the entire end-to-end efficiency of a work process,” Sykes says, “not just the piece you can automate today inside a single system.”
But shedding the sock isn’t something companies can do thread by thread, she says. They have to start over and build new ways of working around what technology can do. The majority of effort in any AI transformation should go toward people and processes, she says, and that means creating new roles, including jobs related to AI governance, systems ownership and workflow maintenance on both the client and agency sides.
As the cost of making content falls, Sykes says, the value shifts to owning the system. But that also requires treating transformation as an ongoing capability rather than a one-time initiative.
“Transformation is not a project; this is a capability,” she says. “It’s a muscle you have to exercise.”
Also in this episode: The thinking behind Brandtech’s new global data and AI practice, how Jellyfish’s “share of model” tool tracks what AI models say about a brand, the role video plays in AI brand recommendations and why Sykes has never seen “The Goonies.”
For more articles featuring Rebecca Sykes, click here.
