This as-told-to essay is based on a conversation with Vineet Mehra, chief marketing officer of Chime, a financial technology company. It’s been edited for length and clarity.
Our first big use case in AI was in customer support. We get millions of calls every month from people wanting to know, “Hey, what happened to my money? What happened to my paycheck? When’s my paycheck going to come in? I sent this money to someone, I don’t see it.”
We built a generative-AI voice and chat system to handle a lot of those questions. In some cases, the AI tool is the first stop for members. In many other cases, it’s like a co-pilot that enables our agents when human interaction is required.
The shift has been transformational. AI is now handling 70% of all of our member support interactions across chat, and, importantly and uniquely, voice. It’s basically the equivalent of thousands and thousands of full-time agents.
We’re seeing wins from this all across the board. With those 70% of interactions, our cost to serve each of these sorts of calls is down 60% since launching these systems. But most importantly, our member satisfaction has doubled. We saw a 40% increase in our support satisfaction scores and a 50% drop in repeat inquiries.
We believe it’s made the experience more human, not less. AI is handling the calls that are more repetitive, so our agents then have the time to focus on those cases that require a lot more human empathy. It’s been a powerful unlock for our agent force as well. They get satisfaction when they can really help someone out when they need it the most.
AI has shrunk the time it takes Chime to make ads by 60% — and cut down agency costs
We are very far along on AI-powered creative production. We’ve cut our reliance on outside agencies. We no longer have a creative agency on retainer. We are a very big marketer — we spend a lot of dollars on brand, storytelling, and content generation. We’ve been able to in-house a lot of this work through our creative team, because of all the work we’re doing with tools like Midjourney for images, Runway, and Veo3.
Our first AI-assisted brand campaign was storyboarded and edited using these tools, all in-house. It was probably the best storyboard as a CMO I’ve ever seen. It was almost like watching a movie.
Related stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
It’s really democratized everyone’s ability to have an idea and present it. It’s unlocking my creative teams. We shrank the time it takes to produce creative campaigns by 60%.
From idea all the way to production and it being on air, we slashed the campaign turnaround by six weeks — from 10 weeks to four weeks. And we did it with no added head count.
This is going to save us millions in agency costs over the next couple of years. I don’t mean to be down on the agencies, but my job is to look at how we put the best work out in the world in the most efficient way for Chime. When I’m saving that much time and I’m saving millions in cost, we’re definitely leaning into this more and more.
How Chime shows up on AI answer engines
A lot of companies are struggling to feed AI answer engines. You need more content than ever. And the types of content you use to win in answer engine optimization is very specific.
Our first attempts with off-the-shelf AI tools weren’t hitting the quality bar for voice or insight.
We shifted to create our own content GPT, Chime GPT. We trained a content creation tool on 1,000 of our best-converting blog articles. And so the brain gets trained on our voice and our style, but also impact and conversion.
We’ve been able to increase our content velocity by orders of magnitude and reduce production costs by 30%. Most importantly, our organic search visibility — not just in SEO, but in GEO (generative engine optimization) and AEO, has increased almost 5%.
Chime created its own AI tool for consumer research
We’ve taken years and years of consumer research and implemented it into a tool that we built called SPARKI Custom GPT.
SPARKI is essentially a market research analyst. Now, everyone on my team doesn’t have to sift through years of studies or go to a research person. You type in any question into SPARKI, and it answers your question from the consumer’s perspective. It puts it right at the fingertips of every marketer or anyone in the company.
We are testing tools like Simile, Native AI, and Synthetic Users, and we’ve trained the AI to act like our consumers. We’re running focus groups with an AI. We’re running it in parallel with real groups because it’s very early. But this is going to 10x speed up our insights that are going to allow us to fuel our brand and creative strategy.
In a world of AI where everyone’s going to have the same machines and the same algorithms, the company that wins is going to be the company with a brand story that’s different. That’s going to be fueled by insights, which are, ironically, also getting powered by AI.
It’s a really exciting time in marketing, a massive platform shift. The key is we’re not using AI on the side. We’re using it to power our overall company mission. We’re measuring it on the same business results that we measured things on pre-AI.
Not every AI experiment has been a success straight off the shelf
Like any new capability, AI is only as good as the data and structure you feed it. We’ve definitely had experiments that didn’t work on the first try. One early example was with OfferFit, an AI decisioning agent, where our early attempts at personalization failed until we refined our inputs.
More broadly, we’ve learned that AI experiments often take a few iterations to hit their stride — and sometimes the impulse to move quickly can cut promising ideas off too soon. We’re building that patience into how we test and learn now, and it’s helped us make smarter long-term calls.
