Clearly Filtered's AI handles 72% of technically challenging conversations end to end.

• 72% chat containment rate
• 4.36 ★
  CSAT on subscription management
• 2,120
  automations per week

Challenge

Clearly Filtered’s CS team fielded technically complex questions at scale. The chat tool in place handled basic FAQ lookups acceptably, but it ran into limits on anything requiring live subscription data from Recharge, structured contaminant lookups from a 250-plus item dataset, or actions in their order stack. There was also no way to test chat changes before pushing them live.

• Subscription actions (cancellation, frequency changes, skips) required agent involvement because the tool couldn’t read or write to Recharge.
• Contaminant questions needed per-filter accuracy the tool couldn’t deliver from a structured technical dataset.
• No ability to distinguish a customer asking about cancellation policy from one who actually wants to cancel.
• No systematic testing capability before pushing chat changes to production.

"The team at Kodif is top-notch. They really understand the eCommerce space and are there to support us as we continue refining our AI tooling. They genuinely want to see us succeed and are true partners."

Matt Dale
Head of CX

Solution

Clearly Filtered launched ‘Crystal’, their chat AI agent built on Kodif, in February 2026. The deployment runs on native integrations into Shopify, Recharge, Loop Returns, and Gorgias, with a custom lookup tool built specifically for their contaminant dataset.

Phase 1

Subscription management with retention logic built in

Six subscription policies run through Recharge, each writing back to the customer’s account directly, with no agent required.

• Subscription Cancellation: the AI identifies the reason and offers a retention option before processing. Cancellation only executes after the customer confirms past the offer.
• Change Next Charge Date, Change Frequency, Skip Subscription, Pause a Subscription, Change Subscription Shipping Address.
• Each policy reads the customer’s live Recharge data first, so the conversation reflects their actual account state.

Phase 2

Product, Contaminant, and order support

• Order Status: WISMO queries pulling live data from Shopify.
• Contaminant Questions: a custom lookup tool built on Clearly Filtered’s 250-plus item dataset, returning specific removal percentages by filter.
• Returns: initiated through Loop Returns with a Gorgias ticket created automatically for agent follow-through.
• Fridge Filter Selection and Exchange: model identification and replacement workflows.

Results

• 72% conversations contained
• 4.36 ★ Subscription Cancellation CSAT
• 54.9% CSAT ratings five stars
• 4.56 ★ Change Next Charge Date CSAT

The low reviews cluster around a single scenario: customers who reached the agent handoff queue and were told to expect a response in one to two business days. Those reviews are a ceiling marker, pointing directly to the next set of flows to automate.

What's next

Expanding order modification flows:

• Mid-shipment address changes

• Multi-item modifications

• Tighter Shopify order cancellation logic

• Increasing automated return throughput through deeper Loop integration

• Extending the contaminant lookup to surface proactive filter recommendations based on what the customer’s current filter addresses

• Reducing the handoff queue by adding policies for the edge cases that currently require an agent, improve the 72% containment.