Composer speeds things up. Agencies know what's actually worth speeding up.
Generated by Klaviyo AI
Klaviyo Composer compresses creative production from days to minutes, pulling from your own account data and patterns across 193,000+ brands. But that data reflects the average, not the exception, which is where a specialized agency's tested playbook comes in.
- Composer handles speed and implementation: It's built into Klaviyo and draws on 14+ years of aggregate marketing intelligence to turn hours of production work into minutes.
- Benchmarks show the middle of the pack, not the top of it: Composer's data spans every account type, underperformers included, so it can't isolate what separates good results from great ones.
- Agencies test the moves an algorithm wouldn't suggest: Reframing billing reminders as gifting moments, or sending plain-text emails instead of designed ones, are wins the author found only by testing across many accounts.
- The value is in cross-account pattern recognition: Agencies can answer "how do other brands like us solve this," a question that requires seeing what's working across a curated set of top performers.
- The two are meant to work together: Composer drives speed and volume; an agency brings the strategic judgment on what to test and why, so brands get both breadth and depth.
Klaviyo Composer is a genuine breakthrough for marketing teams. What used to take days (briefing, building, copy review, quality assurance checks, etc.) can now happen in minutes, grounded in your own account data and backed by 14+ years of Klaviyo's marketing intelligence across 193,000+ brands. That's an incredibly meaningful shift in how fast brands can move.
So why, in a world where AI can produce ready-made creative on demand, would a brand still need an agency?
In my role as founder and CEO of YOCTO Agency, a boutique lifecycle marketing agency and Klaviyo Elite partner, I've been thinking about this a lot. If you manage brands on Klaviyo, or if you’re a Klaviyo brand considering hiring an agency, I’m sure you have, too.
I've been part of conversations with Klaviyo's leadership about the direction of AI for at least a year, and I've been watching our own agency go through the shift in real time. My honest take: tools like Composer don't make agencies irrelevant. They change what agencies are actually for.
What Composer does really well
Composer is Klaviyo's AI orchestration layer built natively into the platform. It sees inside your account: your flows, campaigns, segments, customer behavior, and historical performance. That means with Composer, you have two powerful data layers working for you:
- Your own performance history
- Aggregate patterns across the broader Klaviyo ecosystem
That combination is powerful. Composer can take a task that used to require hours of production time and compress that timeline dramatically. It can surface patterns you might not have noticed in your own data. It can help you implement faster and more consistently.
But here's the thing: both of those data layers are bound by what's already inside your account, and by what's average across the accounts Klaviyo can see.
The third data layer
Klaviyo benchmarks are incredibly useful. They tell you how you perform relative to businesses like yours. But "businesses like yours" includes everyone: the underperformers, the average, and the exceptional. The benchmark reflects the middle of the distribution, not the top of it.
An agency that's worked deeply in your industry has something different: a curated dataset drawn specifically from top-performing accounts. Not what's normal, but what's actually working at the highest level. And more importantly, why.
At YOCTO Agency, we work exclusively with health and wellness brands on subscription-based business models. That focus means that when a client comes to us with a specific problem (e.g., subscribers canceling, payment recovery rates dropping, psychographic segmentation they don't know how to implement), we're never starting from scratch. We've seen this exact problem before, across many accounts, and we've measured what actually moves the needle.
This is the kind of expertise that only an agency can bring to the table.
An agency that's worked deeply in your industry has something different: a curated dataset drawn specifically from top-performing accounts.
What agency expertise looks like in practice
Let me give you a couple of concrete examples.
Billing reminders vs. gifting moments
One of the most common cancellation drivers for subscription brands is the billing reminder. You know the one: "You're about to be charged $X. Your next shipment includes Y." It's accurate. It's transactional. But it triggers cancellations at a surprisingly high rate.
What we discovered, after testing this across multiple clients, is that reframing those notifications as gifting moments changes the dynamic significantly. Instead of telling a subscriber they're about to be charged, you tell them something they're about to receive. The psychology shifts from obligation to anticipation, and cancellation rates drop meaningfully.
That insight doesn’t come from a benchmark. It comes from running the test, measuring the result, and doing it again with enough accounts to be confident in the pattern. That’s the kind of practice any agency worth hiring is doing regularly.
The one-line plaint-text email
The best LLMs and vibe marketing tools probably won’t tell you to send a plain-text email without any bells or whistles.
But in my experience, plain-text email significantly outperforms every designed, fully-branded alternative. We know this because we’ve tested. The simpler the message, the more it reads like a real person reaching out, and the more people respond to it.
An LLM isn’t likely to surface this insight because it goes against conventional wisdom. You'd have to be willing to test it, and you'd have to have enough context across enough accounts to trust what the data is telling you.
Again, this is exactly the expertise an agency should bring to the table.
The questions brands actually ask
Here's the pattern I see most often. A brand comes to us and says something like:
- "We want to increase our subscription take rate. What's the best structure for our offering?"
- "We're seeing a lot of chargebacks and we don't know how to combat this."
- "We want to introduce psychographic personalization, but we don't have the capability, and we're not even sure how to measure whether it worked."
These aren't questions any AI tool can answer on its own, because it can't see what the best brands in your category are doing.
The most valuable question we get at YOCTO Agency, over and over, is some version of: "How do other clients of yours solve this?"
And we’re not alone. “Half our value is just being able to share how other brands have handled a particular problem,” says Zac Fromson, co-founder of Lilo Social, a full-funnel growth agency. “That cross-account memory is the thing brands are really paying for right now.”
It’s this type of question that requires that third data layer.
The right frame: both, not either/or
The play here isn't Composer or agency. It's Composer and agency.
Composer handles the speed and implementation side beautifully. It should. That's exactly what it's built for.
What an agency offers is the strategic layer on top: the frameworks developed from working with many brands like yours, the out-of-the-box tests you'd never think to run on your own, and the ability to look at your situation and say, with confidence, here's what's actually likely to move the needle, and here's how we know.
Think of it this way. Benchmarks give you a wide view across many accounts, and Composer gives you depth inside your own account. An agency that specializes in your category can give you both: wide coverage across comparable accounts, and real depth into what separates good performance from great.
Brands that pair AI-powered orchestration with that kind of human-curated intelligence get the best of both worlds. The efficiency gains from Composer are real. So is the strategic gap that a great agency partner fills.
“Speed without direction just means faster mistakes,” Paul Pritchard, group CEO of Overdose, a global digital commerce consultancy specializing in end-to-end retail and B2B solutions. “Agencies provide the architectural judgment from years of cross-industry experience that machines simply can’t replicate.”
Agencies provide the architectural judgment from years of cross-industry experience that machines simply can't replicate.
What this means for how agencies need to evolve
You already know: the cost of producing content is heading toward zero. If implementation, output, and volume are the main things your agency is selling, that's going to be a tough position to defend.
“The agencies that will matter moving forward will be those that use AI to accelerate delivery while partnering entirely around measurable business outcomes,” says Pritchard.
That’s because they’ve built something an AI genuinely can't replicate: proprietary frameworks tested across real accounts, in specific categories over real time. These are the agencies that will be able to walk into a client conversation and say, "Here's what we've seen work for businesses exactly like yours, and here's the data behind it."
“A model can produce the asset,” Fromson says. “But it doesn’t have the scar tissue an agency does.”
A model can produce an asset, but it doesn't have the scar tissue an agency does.
Agencies have perspectives AI can’t produce on its own. And for brands that want to operate at the top of their category rather than the middle of the pack, that's exactly what they should be looking for in a partner.
Related content
