BODi replaces a menu-based chatbot with agentic AI and achieves 88% containment rate 

BODi (formerly known as Beachbody) has been helping people pursue their fitness and nutrition goals since 1998. The company behind some of the world’s most recognized home fitness programs, including Shakeology, serves a passionate customer base across programs, supplements, coaching, and digital subscriptions. When customers have questions, they tend to have real ones: which program is right for a specific fitness level, whether a product contains allergens, how to troubleshoot a streaming issue, how to manage a subscription.

For years, BODi’s first-generation chatbot handled that volume on a menu-based model. Customers typed their question and discovered they had to select from a predefined list instead. Questions that didn’t match a menu option went straight to a human agent, pulling the support team away from customers who genuinely needed live assistance.

BODi’s previous chatbot was built around predefined intents. When a customer’s question matched a recognized intent, the bot returned the associated answer. When it didn’t, the conversation failed and the customer was transferred to a human agent.

This created a specific set of problems that affected both customers and the support team:

• The menu trap: Customers who typed free-form questions found themselves redirected to a list of options that rarely matched what they’d actually asked. A customer asking “Is there gluten in Shakeology?” would see a menu of product categories rather than an answer. The most natural form of communication, typing a question, didn’t work.
• Escalation of questions that didn’t need escalation: Because the bot couldn’t handle free-form input, any question outside its predefined scope went to a human agent automatically. The support team spent significant time answering questions the bot should have handled, which left less capacity for genuinely complex issues.
• No ability to identify a product from a description: BODi offers dozens of fitness programs and nutritional products. Customers often describe what they’re looking for without knowing the exact name. The old system couldn’t identify “the 21-day program that comes with containers” as 21 Day Fix without the customer using the exact product name.
• Difficult to improve: Updating the bot required manually identifying which intents were failing, writing new responses, and retraining the system. There was no automatic mechanism for surfacing which questions were falling through the gaps.

Quiq replaced the menu-based chatbot with an LLM-powered AI agent that accepts free-form questions and generates answers from BODi’s knowledge base. The system was deployed as a single, streamlined digital channel.

The deployment was built around several specific capabilities:

• Free-form question handling: Customers type exactly what they want to know. The AI agent reads the question, determines intent, and generates a response calibrated to what was actually asked. No menus, no predefined option lists.
• Product identification from description: The AI agent can match a customer’s description of a product to the actual item in BODi’s catalog. A customer who asks about “the nutrition program that comes with colored portion containers” gets an answer about 21 Day Fix without needing to know the product name.
• Disambiguation through follow-up: When a question is genuinely ambiguous, the AI agent asks a follow-up question to clarify before responding. This mimics the way a human agent would handle an unclear inquiry, rather than guessing or returning a generic answer.
• Conversation context across turns: The LLM retains context throughout a conversation. A customer who asks a follow-up question doesn’t have to re-establish the topic. The AI agent knows what was discussed and responds accordingly.
• On-brand experience: BODi has a distinctive voice, energetic, motivational, and grounded in the fitness lifestyle. Quiq configured the AI agent to reflect that tone while staying within BODi’s content and compliance guidelines.
• Knowledge gap identification: Quiq surfaces patterns from conversations the AI agent handles with lower confidence, giving the BODi team specific, actionable information about which knowledge base articles to update or create.

Ingredient and allergen questions: A customer visiting the Shakeology product page wants to know whether the formula contains gluten. They type the question directly. The AI agent identifies the product from context, looks up the relevant ingredient information, and answers with specificity. In the old system, this question would have triggered a menu; in the new one, it’s answered immediately and accurately.

Program recommendation by fitness level: A first-time buyer wants to know which program is right for someone who hasn’t exercised in over a year. The AI agent asks one clarifying question (are they looking for strength, cardio, or a combination?), then recommends two programs with a brief explanation of what distinguishes them. The human support team doesn’t get involved.

Subscription management: A customer wants to know how to pause their subscription before a planned trip. The AI agent walks through the steps specific to BODi’s subscription portal, including edge cases like mid-cycle pauses, without requiring the customer to navigate to a separate help center.

For Troy Nelson’s team, the shift was felt in the composition of the support queue rather than the volume. The questions that reach human agents now are the ones that need a human: subscription billing disputes with unusual circumstances, edge cases around coaching relationships, technical issues with streaming that require account-level investigation. The routine questions, product ingredients, program recommendations, subscription management, no longer reach the team at all.

One in three customers who contact BODi now choose the AI agent over other contact options, a signal that the experience is good enough to be a customer preference, not just a deflection mechanism.

The team has also gained insight they didn’t have before. When the AI agent handles a question with lower confidence, it flags that interaction. The BODi team reviews those flags and uses them to update knowledge base content. The system gets more accurate with each cycle.

Ask BODi AI went from a proof of concept to a primary support channel, handling 88% of conversations without escalation. The experience is on-brand, accurate, and available to customers the moment they land on the support site.

• 88% containment rate: 88 out of every 100 conversations resolved without escalating to a human agent
• 1 in 3 customers actively choose the AI agent over other contact options
• Free-form question handling: no menus, no predefined options, no mismatch failures
• Product identification from description: customers don’t need to know exact product names
• Continuously improving accuracy through a knowledge gap identification loop
• Human support team focused on genuinely complex issues rather than routine inquiries