top of page

AI in Customer Success and Support: What Exists Now, What's Coming and What to Weigh

  • Writer: Nadine Chucri
    Nadine Chucri
  • Jul 8
  • 10 min read

Updated: 6 days ago

Glowing AI processor chip labeled AI on a dense circuit board, with metallic black, silver, and gold components in a high-tech scene

AI is being bolted onto customer success and support faster than almost any other business function. For a discipline built on relationships and outcomes, that's both the opportunity and the trap. Here's an honest, leader-level read on what exists today, where it's heading and what's genuinely worth weighing, good and bad.


There is a peculiar tension at the heart of putting AI into customer success and support. Support has always had a volume problem that automation is well suited to solve. Customer success has always had a relationship problem that automation is uniquely bad at solving. Lump the two together under "AI for the customer," as most vendor decks now do and you get a wave of capability that is genuinely transformative in one half of the picture and quietly corrosive in the other if you're not careful about where you point it.


This piece separates the two deliberately, because the decisions a leader needs to make are different on each side. We'll look at what already exists in 2026, where the technology is heading over the next few years and then the part most vendor content skips: the benefits and the risks, weighed honestly, with the sources to back them.


What already exists in 2026


The capability has moved well past chatbots and it splits cleanly into two domains.

On the support side, the headline shift is from assistance to action. Earlier AI summarised, suggested and drafted; today's "agentic" systems can investigate an issue, access backend systems, take a step such as processing a refund or updating an account and close the interaction without a human touching it. The same systems run a quieter set of jobs behind the scenes: real-time agent copilots that draft replies and surface knowledge mid-conversation, automatic call and ticket summaries that kill manual note-taking, intelligent triage that routes by intent and sentiment, plus quality monitoring across every interaction rather than the small sample a human QA team can review. The pressure to adopt is intense. Gartner found that 91% of customer service leaders were under pressure to implement AI in 2026. On performance, a word of caution worth carrying into any vendor conversation: autonomous resolution rates for routine queries are commonly quoted in the 60% to 80% range, but those figures are largely self-reported by the platforms selling them and the honest metric to ask any vendor for is documented resolution from a deployment comparable to yours.


On the customer success side, AI has reshaped the analytics and workflow layer rather than the customer-facing one. Across 2024 to 2026, effectively every major customer success platform (Gainsight, ChurnZero, Vitally, Totango, Catalyst, Planhat) added an embedded AI copilot or agent. What that buys you in practice: health scores and churn-risk predictions synthesised continuously from product usage, support and billing data rather than hand-curated by an ops team; and AI-generated account summaries that compress a year of notes before a renewal call. The genuinely new capability is sentiment analysis that reads unstructured communication (emails, call transcripts, Slack threads) to flag when a key stakeholder is going quiet or a relationship is cooling, often before usage metrics move. ChurnZero frames the shift well: the fundamentals of health scoring haven't changed, but AI has expanded what can be measured (including how a customer feels about the partnership, not just what they click) and automated the response to it.


The cleanest way to summarise the current state is this: customer success has moved from dashboards that humans interpret toward agents that propose actions and humans approve. That is a meaningful change in the operating model and it's already here, not on the horizon.


Where it's heading


The most-cited forecast in the field comes from Gartner, which predicted in March 2025 that by 2029, agentic AI will autonomously resolve 80% of common customer service issues without human intervention, driving a roughly 30% reduction in operational costs. Treat that as a direction of travel rather than a promise. It's specifically about common issues and "begins to materialise for mature organisations" is doing a lot of work in the analyst commentary around it. Alongside it, the broader expectation is a move from channel-based, reactive support to journey-based, autonomous resolution. On the success side, that means a shift toward genuinely proactive intervention: resolving or pre-empting an issue before the customer raises it.


But the same analysts forecasting autonomy are equally clear that humans aren't leaving. In an April 2026 survey of 321 service and support leaders, Gartner found that 85% are expanding human agents' responsibilities (adding new tasks or moving agents into entirely new roles) as AI absorbs routine volume, while only 31% had implemented or were planning AI-driven frontline layoffs through early 2027. Most were reshaping the workforce rather than eliminating it: 63% said they were reducing frontline headcount gradually through attrition while redeploying that capacity toward higher-value work. The realistic future, then, isn't headcount-zero support or an AI that runs your renewals. It's a hybrid in which AI absorbs the volume and the toil and humans concentrate on the judgment, the complexity and the relationship, which, on the success side, was always the actual job.


The good: where AI genuinely helps


The strongest case for AI in this space isn't cost. It's reclaiming human attention for the work that actually drives retention.


The clearest win is removing low-value toil. Note-taking, account research, ticket triage, meeting summaries, QBR prep, first-line FAQ deflection: none of this was ever the part of the job that renewed a customer and handing it to AI frees a CSM or an agent to do the part that does. As the support cliché goes, you don't want your most experienced, most empathetic person resetting passwords.


The second win is earlier, richer risk detection. Traditional health scores fire on usage drops, which is a lagging indicator. By the time logins fall, the decision to leave is often already made. AI that reads qualitative signal (a champion's tone shifting in email, a stakeholder going quiet, a recurring complaint theme across accounts) can surface a cooling relationship earlier than telemetry alone, which is the difference between a save and a post-mortem.


The third is scale and consistency: 24/7 coverage, instant context on every account and quality review across every interaction rather than the small fraction a human QA team can realistically check. And there's a quieter benefit for smaller teams specifically: capabilities that were once gated behind enterprise platforms and dedicated ops headcount are increasingly available to lean CS functions, which lowers the barrier to running a serious, data-informed motion without a six-figure tooling bill.


The bad: what to weigh carefully


Now the part the vendor decks skip.


Customers are genuinely wary. That matters more in a relationship discipline. In a Gartner survey of 5,728 customers, 64% said they would prefer that companies didn't use AI in customer service at all and 53% said they would consider switching to a competitor if they discovered a company was going to use AI for service. Their single biggest fear wasn't a sci-fi one; it was that AI would make it harder to reach a human. A more recent Gartner survey (of 5,801 US customers in early 2025) found that the preference sharpens precisely where customer success lives: 54% said they trust a human agent more than AI for product or service recommendations, against 32% who trust AI more. That is the advisory, value-shaping conversation at the heart of the CS role. Sit with the irony for a moment: customer success exists to build a trusted relationship and a heavy-handed automation strategy can erode the exact thing the function is there to create. The lesson isn't "don't use AI." It's that the human path has to stay obvious and the AI has to earn trust by being demonstrably helpful rather than a wall.


You own what your bot says. In 2024, Canada's British Columbia Civil Resolution Tribunal held Air Canada liable after its support chatbot gave a customer incorrect information about bereavement fares. The airline argued, remarkably, that the chatbot was a separate entity responsible for its own statements; the tribunal rejected that outright. The precedent is simple and it applies to every business deploying these tools: a hallucinating bot isn't a quirky bug, it's a public, legally binding expression of your company.


The cost-cutting trap is real. When AI is deployed to remove headcount rather than to solve customer problems, the predictable result is degraded service, overloaded remaining staff and eroded trust. Gartner's own analysts put it plainly: organisations that use AI only to reduce costs risk missing the strategic opportunity, which comes from pairing AI efficiency with human judgment, empathy and experience. Their 2026 data backs the warning: only a minority of service leaders (31%) were pursuing AI-driven layoffs at all, with the majority redesigning roles instead. Klarna's widely reported retreat from an AI-first support model became the cautionary tale of the cycle, with the company moving back toward offering human agents. If the honest answer to "why are we doing this?" is only "to cut costs," the strategy tends to boomerang.


The data problem doesn't disappear. It compounds. Every AI health score is only as good as the data feeding it and most of these systems are very good at telling you what is happening (usage dropped, a ticket spiked) and much weaker at telling you why, which is where renewals are actually won or lost. Automating on top of thin or messy data just produces confident, well-formatted noise faster.


And the metrics can quietly mislead. Deflection is the obvious trap: an AI that sends a customer an FAQ link and closes the conversation looks efficient and may have resolved nothing. If you measure AI by how many humans it kept customers away from, rather than how many problems it actually solved, you'll optimise straight into the churn you were trying to prevent.


What this means for leaders


Strip it back and the through-line is a single principle: automate the toil, not the relationship.

Point AI at the work that was never the point (the notes, the triage, the research, the genuinely routine first-line questions) so your people spend their time on outcomes, judgment and the human moments that build loyalty. Keep an obvious, low-friction path to a person and be transparent that AI is in the loop. Customers tolerate, even welcome, AI when it's clearly optional and clearly useful. They resent it when it feels like a trap. Own the accountability for what your systems say. And measure AI against customer outcomes (resolution, retention, value realised) not just deflection and cost-to-serve, because the discipline is called customer success for a reason.


The companies that will come out of this period ahead aren't the ones that automated the most. They're the ones that used AI as leverage on the relationship rather than a substitute for it. The technology is finally good enough to make that choice consequential. The choice is still yours.


Frequently asked questions


Will AI replace customer success managers?

Not in the foreseeable future. AI is automating large portions of a CSM's manual work (notes, account research, health analysis, summaries) but the core of the role (judgment, relationship-building, helping customers achieve outcomes) is exactly what current AI is weakest at. The realistic shift is augmentation: agents propose actions, humans approve and own the relationship. In Gartner's 2026 data, most service leaders were expanding human roles rather than cutting them as AI absorbed routine work.


Will AI replace customer service jobs?

Routine, high-volume support is automating quickly and some roles are being reshaped or reduced. But in Gartner's April 2026 survey, 85% of service leaders were expanding human agents' responsibilities and only 31% were implementing or planning AI-driven layoffs. The dominant pattern is redesigning roles, not eliminating them. The likely outcome is fewer purely transactional roles and more focused on complex, high-value and emotional interactions.


What is the difference between AI in customer support and AI in customer success?

Support-side AI is largely customer-facing and resolution-focused: chatbots, agentic resolution, triage, copilots. Success-side AI is largely internal and prediction-focused: health scoring, churn prediction, sentiment analysis and CSM copilots. Support AI handles the volume of reactive issues; success AI surfaces risk and opportunity across the customer base so humans can act on it.


Do customers actually want AI in customer service?

The evidence is mixed and leans cautious. In Gartner's survey, 64% of customers would prefer companies didn't use AI for service, with the top concern being difficulty reaching a human. Customers are far more accepting of AI when it's fast, accurate, clearly optional and backed by an easy path to a person. They turn hostile when it feels like a barrier.


What are the main risks of AI in customer service and success?

Customer trust erosion, legal accountability for AI errors (as in the Air Canada chatbot ruling), over-automation that degrades service and erodes trust, dependence on data quality and misleading metrics like deflection that reward avoidance rather than resolution.

Not sure whether your customer success function is ready to add AI on top of it or whether the foundations need work first? The CS Maturity Scorecard assesses where your motion actually stands and The Compounding Customer makes the leadership case for building retention on relationships AI can support but not replace.


Sources


All links verified live as of June 2026. Survey figures are attributed to the organisation that ran the survey; vendor performance claims are flagged as self-reported.


Customer attitudes (primary)



Forecasts and adoption (primary)



The Air Canada chatbot ruling



Customer success AI tooling (industry and vendor reporting, descriptive, not load-bearing stats)


Comments

bottom of page