It’s 9:40 p.m. Someone is browsing sunscreen on your online store, adds it to their cart, and closes the tab. The next day, they message you on Instagram to ask how long shipping to Bucaramanga takes. Your team replies later that afternoon. By then, they’ve already bought it at the pharmacy around the corner.

At Asylum, we see scenarios like this almost every week. The point is simple: in ecommerce, chat and CRM often operate separately, and that’s how sales that were nearly closed slip through the cracks. When both are designed to work together, what a customer says in a chat helps the CRM determine what to send, when to send it, and through which channel.

Chat Responds, CRM Sends Emails, and Neither Talks to the Other

 

It happens in many online stores. The social media team responds to WhatsApp and Instagram messages. The CRM team builds campaigns for the subscriber database. Each team does its job well, but information doesn’t flow between them.

Someone who asked about an anti-dandruff shampoo on Tuesday receives an email promoting sunscreen on Friday. It’s not a major issue, but it’s clear that no one was really listening.

SmartChats help bridge that gap. These AI-powered agents connect to WhatsApp, Instagram, Facebook, TikTok, or a website’s live chat. They’re available around the clock, understand what people are looking for, and categorize conversations based on customer intent: someone wants to make a purchase, has a shipping question, or is reaching out with a complaint.

Those labels are particularly valuable for CRM.

What We’ve Learned from Analyzing Customer Conversations

 

We’ve been managing AI agents across the social media channels of several personal care brands for some time. Whenever we sit down to review the accumulated data, the same pattern emerges.

The most common conversation label is purchase intent. People aren’t just browsing; they’re coming in ready to buy.

Even so, the conversion rate from chat is around 3%. Interestingly, when someone does make a purchase, it happens quickly: on average, one day and five hours after their first message.

Conversations that don’t lead to a purchase remain open for almost 20 days before being considered lost. And whenever we look at the data, roughly two out of every three chats are still open.

We see those 20 days as an opportunity. That’s time during which someone has already shown interest, shared their contact information, and hasn’t heard from the brand again.

CRM can fill that gap.

There’s another, more practical finding. Thursday tends to be the busiest day for conversations, and more than half of the activity takes place between noon and 5 p.m. With that information, CRM teams no longer have to guess when to send their messages.

Four Moments When Chat and CRM Work Better Together

 

The first is turning chat conversations into audience segments.

Each label assigned by an AI agent can feed a segment in an email marketing platform. If someone asks about a product but doesn’t purchase it, they can enter an automated flow that follows up 24 or 48 hours later, reminding them about the product, providing a direct link, and addressing their original question.

The second is abandoned cart recovery, but with context.

The classic “You left something in your cart” message works. It works even better when we understand why the customer left.

If they previously messaged the brand asking about shipping costs, the follow-up can address that concern first.

The third is post-purchase engagement.

The first few days after receiving an order are often when customers have the most questions. An AI agent can answer them without keeping anyone waiting.

Meanwhile, CRM can estimate when the customer might run out of the product and schedule a replenishment reminder.

An antiperspirant doesn’t last as long as a multivitamin, and the messaging schedule should reflect those differences.

The fourth is customer reactivation.

Customers who haven’t purchased in 60 to 90 days are usually sent a discount email. Sometimes, a WhatsApp message that references their previous purchase works better.

It feels less like advertising and more like a thoughtful follow-up.

In categories such as health and wellness, there’s an important boundary that should be established from the beginning.

AI agents cannot promise health outcomes or replace medical professionals, and CRM shouldn’t encourage people to purchase more than they need.

A well-designed automation knows when to recommend a medical consultation instead of a product.

What to Measure (and What Matters Less)

 

The number of messages received may look impressive in a presentation, but it doesn’t tell us much about business performance.

In ecommerce, we focus on other metrics: the 60-day repeat purchase rate, the time between a customer’s first and second purchase, how much revenue chat generates directly, and how much revenue comes from owned channels such as email, SMS, and WhatsApp without relying on paid media.

We also monitor response times.

A customer who’s been waiting six hours for an answer probably isn’t waiting anymore.

Before Automating, Get Your House in Order

 

None of this works if your data isn’t connected.

The first step is usually the least exciting: properly tracking your online store, syncing it with your email marketing platform, defining conversation labels, and obtaining consent for personal data processing from the very first interaction, as required by Colombia’s data protection laws.

And you need to give the process time.

The first few months are about learning which segments respond, through which channels, and to what type of messaging. The following months are about refining the strategy.

One thing we’ve learned over the years is that dividing audiences into overly specific segments before having enough data leads to misleading conclusions.

Effective segments are built from months of conversations, not from the first week.

Success Story: How Somos Bienestar Introduced Intelligent Agents Across Its Channels

 

Talking about artificial intelligence in ecommerce is one thing. Seeing what happens when these agents begin managing real customer conversations is another.

At Asylum, we supported the implementation and optimization of SmartChats for Somos Bienestar, Megalabs’ ecommerce platform, connecting customer service across WhatsApp, Instagram, and Facebook through an operation powered by intelligent agents.

During September 2026, the ecommerce platform’s first full month of operation, Somos Bienestar’s three channels received 417 incoming messages, representing 79% of the 530 messages recorded across the overall SmartChats operation analyzed by Asylum.

WhatsApp became the most active channel, with 306 incoming messages. This activity was partly driven by Click-to-Chat campaigns, which connected ads directly to WhatsApp conversations.

This allowed users to move from interacting with an advertisement to accessing a customer service channel where they could get answers to their questions.

The Results: What We Found During the First Month

 

September’s performance indicators show how artificial intelligence can help manage conversation volume while also revealing the challenges that arise when a situation requires human intervention.

Metric	Result
Incoming messages for Somos Bienestar	417
Incoming WhatsApp messages	306
Connected contacts for Somos Bienestar	103
AI agents’ first response time	Between 19 and 24 seconds after assignment
Conversations escalated to Customer Service	26
Average human first response time	40 hours and 42 minutes

 

Source: Asylum Marketing SmartChats Report, September 2026. Response times refer to the overall operation involving intelligent agents and human customer service representatives, not exclusively to Somos Bienestar.

The difference in response times reveals an important insight.

While AI agents could respond within seconds of being assigned a conversation, a human customer service representative took nearly 41 hours, on average, to provide the first response.

This demonstrates that automating the initial interaction can reduce waiting times, but it doesn’t eliminate the need for efficient processes to resolve inquiries that require specialized attention.

Turn Every Ecommerce Conversation into a Business Opportunity

 

At Asylum, we connect artificial intelligence, CRM, and conversion strategies to help your ecommerce business grow.