2026 Survey Report
The State of AI in Customer Support
By Alex Turnbull, Jared Scheel, and Tom Morkes
Survey fielded October to December 2025
We surveyed 1,400 support leaders on how they're adopting AI, what's actually working, and where teams are getting stuck. Get the full report with benchmarks you can measure your team against.
We'll email the PDF and nothing else unless you ask.
Free · 28-page PDF · No sales follow-up required.
78%
of support teams now use AI in at least one part of their workflow
3.4x
faster median first response when an AI agent drafts the reply
41%
of inbound conversations resolved without a human touch
1,400
support leaders surveyed across B2B and B2C
Four findings that changed how support teams operate.
Adoption
AI moved from experiment to default.
Two years ago most teams were piloting AI on a single deflection use case. Today it sits in the core workflow: drafting replies, triaging tickets, searching knowledge, and taking action in connected tools. The teams still on the sidelines cite trust and data readiness, not cost, as the reason they have not started.
78%
use AI in at least one workflow
Resolution
Deflection is out. Real resolution is in.
Leaders stopped measuring how many tickets AI kept away from a human and started measuring how many it actually resolved end to end. The best-performing teams give their agents access to systems of record and let them act, not just answer, which is where the resolution gap between leaders and laggards opens up.
41%
of conversations fully AI-resolved
The human role
Agents are becoming supervisors, not typists.
The fear of headcount cuts did not materialize the way the headlines predicted. Instead the job changed: reps spend less time typing the same answer and more time reviewing AI work, handling edge cases, and improving the knowledge and guidance the AI runs on. Job satisfaction rose fastest on teams that framed AI as a teammate rather than a replacement.
2.6x
more time on complex, high-value work
What holds teams back
The blocker is trust and data, not technology.
When we asked what stops teams from expanding AI, the top answers were unreliable knowledge, fragmented customer data, and no clear way to measure quality. The tooling is ready. The teams that win invest first in clean knowledge, connected systems, and visibility into what the AI is doing.
63%
name knowledge quality as their #1 blocker
By the numbers
A snapshot of where support leaders landed in 2026. The full report breaks each metric down by team size, industry, and support model.
3.4x
faster first response with AI-drafted replies
52%
of leaders plan to grow their AI budget in 2026
29%
lower cost per resolution vs. human-only teams
4.6/5
average CSAT on AI-resolved conversations
67%
measure AI on resolution, not deflection
11%
still have no AI in their support stack
How the survey was run.
We surveyed 1,400 customer support leaders between October and December 2025, spanning startups to enterprises across B2B SaaS, e-commerce, fintech, and consumer apps. Respondents ranged from frontline team leads to VPs and Heads of Support.
Questions covered AI adoption, tooling, measurement, budget, team structure, and outcomes. Where we report benchmarks, figures are medians unless stated otherwise, and we segment by team size and support model so you can compare against teams like yours.
The full report includes the complete question set, segment breakdowns, and year-over-year comparisons where available.
Get the full State of AI in Customer Support report.
All the benchmarks, segment breakdowns, and leader playbooks in one 28-page PDF. Free to read, no sales follow-up required.
The complete system for human and AI customer service.
Helply is the complete AI-native support platform: everything you expect from Zendesk, rebuilt around AI agents that investigate, resolve, and act.
Unlimited users. Pay only for AI work.