How Cleo scales lifecycle marketing with an AI Agent 

Cleo, an ai-powered financial assistant and budgeting app, operates in a financial services context, where compliance shapes nearly every decision the CRM team makes. The team is expanding into new markets across the UK, the US, and beyond, so getting lifecycle marketing right at that intersection—personal, on-brand, and compliant—is not a small ask.

A year ago, product teams would arrive with "we need these three emails next week," and Ed Preston, Sr. Growth Marketing Manager at Cleo, would build them and hit send. Strategy took a back seat to execution because there was no room for anything else.

The Cleo lifecycle team turned a one-person email operation into a strategic function that plans campaigns months ahead—using Customer.io's Agent to clear away the manual work that used to swallow entire days.

Cleo runs enterprise AI access across the business, and Ed connected Customer.io directly into his AI assistant alongside his meeting-notes tool and the team's Notion strategy docs. That setup lets him generate weekly updates that pull from his meetings and planning docs automatically. Even when a tool only saves 10% of someone's day, he notes, that time goes straight back into the strategic work that actually moves the business.

The momentum is part of a longer arc. On stage at a Customer.io event in Lisbon last January, Ed described the tool he wished he had: an agent that could just build his campaigns for him. At the time it sounded like a someday feature. Three months later, when Customer.io shipped its AI Agent, someday arrived a lot faster than he’d expected.

Team growth helped too. Two CRM managers joined in January, two more hires will start in June, and a one-person function is on its way to a team of five. But the headcount only goes so far on its own—what lets a five-person team operate like a much larger one is automating the work that used to eat days at a time.

With the busywork handled by AI, the team can strategize. Cleo's CRM team now works in four-month terms. Instead of rushing, the team set baselines and concrete targets, used holdout groups it had configured in Customer.io to measure against, and looked back at past test results to make informed calls. By Ed's estimate, the planning behind the second term was at least 50% more thorough than the first.

The clearest sign of change was a kickoff for an August launch that happened months in advance—something that, by his account, had basically never happened before. Product launches that used to land with a week's notice now get audience definition, segmentation, and strategy worked out long before a single email gets built.

Ed is the first to say his team isn't using the Agent to its full potential yet, but the early, unglamorous wins are exactly the ones that add up on a small team.

The example he keeps coming back to is snippets. Cleo attaches a snippet to nearly every disclaimer in its emails—roughly 100 of them in the production instance—and none of them existed in the staging environment the team uses to build and test. Recreating them by hand would have taken someone the better part of a day.

Ed gave his AI Agent in Customer.io a few prompts of instruction, and it copied every original production snippet into staging so the two matched. In a financial services business where disclaimers are a compliance concern, keeping those in sync matters.

The Agent synced Cleo's Design Studio assets, fonts, colors, the full design scheme, between production and staging so the two environments stayed in step. When an email threw a rendering error buried in its HTML, the team asked the Agent to find it. The culprit was a stray space inside an image href, the kind of thing that's painful to catch by eye, and the fix cleared the email to go out the door.

Then there's the profile limit. Cleo had run over its 7 million profile limit twelve months in a row, and checking usage meant manually adding up profiles across instances. Now the Agent sends Ed a daily email showing how many new profiles were created and how close the team is to its ceiling—an automatic nudge to clean things up before an overage hits.

Every net-new email Cleo sends is now built in Design Studio (https://customer.io/platform/design-studio), one of the team's goals for the term. The team moved off a third-party HTML email builder it used to import code from, which means no more broken imports and no more design drift between what's designed and what ships.

The bigger payoff is shared modules. Change the footer once and it updates everywhere, which matters more and more as Cleo expands into new markets.

Cleo's longer-term goal is consolidation. Push notifications are split roughly half-and-half between Customer.io and an internal tool today, SMS verification runs through a separate system, and WhatsApp is under investigation—held up less by technology than by how the company collects phone numbers in the UK and by message-category rules in the US. Ed's view is that CRM channels belong in one place, under one team, because that's what makes for a coherent customer experience.

The capability he most wants next is testing velocity. Today, producing and testing a batch of email variants can take close to three months. Ed wants to spin up 20, 30, even 100 variants of a single email and run them through automated multi-armed bandit testing—letting the system retire the underperformers, push traffic toward the winners, and ideally surface different winners for different audiences, like engaged versus unengaged customers. He's already trialing an outside service that generates dozens of design and copy variants from a single email, and he'd like to pair that kind of output with Customer.io's testing—turning a three-month cycle into something closer to three days.

It's an ambitious task, but it fits the pattern of the last few months at Cleo. And as Cleo consolidates more of its CRM onto a single platform and builds out faster testing, Customer.io is the foundation it's choosing to grow on.