The strategy behind Cymbiotika's custom AI systems

Hey everyone,

Welcome back for another bite to chew on.

Everyone's talking about AI, but most brands are still just playing around. ChatGPT for copywriting. Basic chatbots for customer service. Maybe some automation tools they saw on Twitter.

Meanwhile, Cymbiotika—the premium supplement brand that's scaled to over $100 million annually—took a completely different approach. 

Instead of renting AI tools like everyone else, they built an entire in-house AI team that functions more like an R&D lab.

We recently sat down with their Chief Digital Officer, Jared Radtkey, to break down exactly how they're doing it. 

Today we’ll break down their approach and how you can apply it to turn AI from vendor dependency into a core growth driver.

🍽️ On the Menu:

You can watch the whole conversation with Jared here:

Listen on Spotify 🎧 and Apple 🍎 as well.

You're spending peak CPMs to acquire BFCM buyers, but are you maximizing their value in those critical first 30 minutes after checkout?

While you're manually testing upsell offers and guessing which products to promote, brands using Aftersell Smart Funnels are letting AI do the heavy lifting.

Their system automatically selects the perfect offer for each customer in real-time, delivering 10-15% lifts in Revenue Per Visitor without the operational overhead.

Here's what makes Smart Funnels different during BFCM:

The strategy is simple: Set your guardrails, let the AI learn from early traffic, then scale based on RPV lift.

Brands are seeing their highest-converting upsells happen automatically while they focus on other BFCM priorities.

Stop leaving money on the table during your most expensive acquisition season.

👉 Explore Aftersell Smart Funnels

After testing dozens of AI tools Jared learned that most of them are garbage dressed up with good marketing.

"A lot of them don't work right now. People get ads and people see posts and videos that you know are clickbait and they get you to overindex on what the opportunity is. When you actually demo with these products or implement them, you realize that they're not what they're advertising."

The key is having someone on your team who actually demos tools before you buy them. 

Someone who can ask the right questions during sales calls and spot the difference between a genuine solution and marketing fluff.

Here's what that means in practice:

(1) During demos, ask for edge cases 

Don't just watch them show the happy path. 

Ask: "What happens when we upload 50,000 customer records?" or "How does this handle customers who bought multiple products?"

(2) Request sandbox access before buying 

Real tools let you test with your actual data. If they won't give you sandbox access, that's a red flag.

(3) Talk to existing customers, not just sales

Ask for references from companies your size with similar use cases. Good vendors will connect you. Bad ones will deflect.

(3) Start small and measure

Even if a tool works, test it on one segment or product line first. Measure the actual impact before scaling.

This is why Cymbiotika ended up building their own systems. 

After testing tool after tool that couldn't handle their scale or use cases, they realized the fastest path forward was to build exactly what they needed.

After hitting dead end after dead end, you have 2 choices:

1) Give up 

2) Build your own

Cymbiotika chose to build.

The decision came from frustration. 

As Jared explained: "There's tons of tools out there, and a lot of them we've tested and we're not thrilled about. Having an in-house team is more like an incubator."

When you rent AI tools, you're at the mercy of vendor roadmaps and feature releases. When you build, you control the technology and can move at your own speed.

Cymbiotika hired developers with experience in language models and multi-agent systems who can create custom solutions. 

They run it like an incubator where they test different hypotheses at a small scale and then when they prove them out, they scale them out and put more budget behind them.

Not every brand can afford a full AI team. 

But every brand needs someone who can separate AI reality from AI marketing. Someone who tests tools before you buy them and knows when existing solutions won't solve your specific problems.

The lab tackles three main areas: personalization that feels natural, processing qualitative data at scale, and internal automation.

Here's how they do it:

Personalization That Feels Real

Instead of long quizzes, they use micro-surveys on product pages.

If someone lands on a vitamin C page, they ask one question: "Are you here for skin health or immunity?" The page then updates dynamically with relevant content.

This information carries across the entire site experience and powers their email and SMS segmentation. 

The goal: personalize outreach timing and messaging based on individual behavior, so each customer gets content when they're most likely to engage.

Processing What Humans Can't

When 30,000 customers visit daily, they generate massive amounts of qualitative feedback. Their in-house chatbot, trained on product data, processes these conversations in real time.

People talk to the chatbot in long sentences about their health goals. The system processes this natural language, tags customer profiles with their specific interests, and uses that data for retargeting and personalization.

The result: customers feel like the brand is talking to them personally, which builds trust and improves lifetime value over time.

Internal Automation

The lab also automates routine tasks across customer service, copywriting, and other business areas using multi-agent systems—freeing up human talent for strategic work.