Intercom: The AI deployment gap in customer-facing products

In January 2026, Intercom published its Customer Service Transformation Report based on a survey of over 2,400 customer service professionals. The report examines a specific pattern: while 82% of organisations invested in AI for customer service in 2025, only 10% reached a level of deployment they classified as mature. The report focuses on what that gap looks like in practice and how the 10% got there.

What separates mature deployments

The data shows measurable differences across deployment stages. Teams with mature AI integration report quality and consistency improvements at nearly double the rate of early-stage teams. Among mature deployments, 87% report improved metrics compared to 62% across all respondents. The gap is most visible in measurement: mature teams have built the infrastructure to tie AI use to business outcomes, while early-stage teams often lack it and cannot demonstrate ROI even when their product works.

Mature teams also report a 28% reduction in time spent on support volume, which has freed capacity for work requiring human judgment. The report attributes this shift not to a single tool or model choice, but to the degree to which AI is embedded across the full workflow — not used for isolated tasks within an otherwise unchanged process.

The shift in priorities

One of the sharper data points is the change in what leaders say they are optimising for. Customer experience and satisfaction became the top priority for 58% of leaders in 2026, up from 28% the previous year. As AI handles more routine throughput, attention is moving toward outcomes that automation alone cannot produce.

Fifty-two percent of organisations plan to extend AI beyond customer support into departments including marketing, sales, and customer success in 2026 — a sign that what works in one function is being treated as a template for broader rollout.

Who will find this useful

The report is most directly relevant to product managers working on customer-facing products, support tooling, or internal AI rollouts. The quantified differences between deployment stages make it practical for setting stakeholder expectations about what successful AI implementation looks like beyond simple adoption metrics. For PMs building AI features, the distinction between “AI is used” and “AI is embedded in the workflow” is one of the core takeaways.