91% of customer service leaders are under executive pressure to implement AI in 2026, according to Gartner's survey of 321 service and support leaders. Not just for efficiency, but for customer satisfaction.
But here's what the pressure is producing: most CS teams are deploying AI tools without training their people to use them. The result is scattered pilots, inconsistent adoption, and a growing gap between what the tools can do and what the team actually delivers.
ChurnZero's 2025 CSM Confidential research puts a number on part of the problem: 42% of CSMs cite sales and negotiation as their top skills gap, followed by business skills (37%) and analytical skills (32%). These are exactly the competencies that AI augments, and exactly the ones that untrained teams can't leverage.
The AI Pressure vs. AI Readiness Gap in Customer Success
Executive leadership wants AI in customer success. The teams doing the work aren't ready for it.
| Finding | Source |
|---|---|
| 91% of CS leaders under executive pressure to implement AI | Gartner (321 leaders, Oct 2025) |
| Only 20% have actually reduced agent headcount due to AI | Gartner (321 leaders, Oct 2025) |
| 55% report stable staffing while handling higher volumes | Gartner |
| 42% hiring specialised AI roles (strategists, automation analysts) | Gartner |
| 42% of CSMs cite sales/negotiation as top skills gap | ChurnZero CSM Confidential 2025 |
| 13% of CSMs received no training in the last 12 months | ChurnZero |
The Gartner data tells a nuanced story. AI isn't eliminating CS jobs; only 20% of leaders have reduced headcount. Instead, 55% are handling higher volumes with the same team size, and 42% are creating entirely new roles around AI. The shift is from fewer people to differently skilled people. And Gartner predicts that half of companies that cut customer service staff due to AI will rehire by 2027, because they'll realise AI augments humans rather than replacing them.
13% of CSMs received no training at all in the last 12 months, according to ChurnZero. Among those who did, the most common format was online courses (45%) and occasional manager coaching (36%). Only 23% had access to mentorship. When the dominant training model is self-directed e-learning, it's no surprise that skills gaps persist.
What Databox and Lightspeed Learned About AI in Customer Success
The CS teams getting measurable results from AI share a pattern: they treated AI deployment as a training challenge, not just a technology one.
Databox: From 30% to 55% AI Resolution, and 40% More Revenue
Databox deployed Intercom's Fin AI agent and watched its resolution rate climb from 30% in December 2023 to 55% by March 2025. Customer satisfaction scores jumped from 30% to 71% over the same period. Team productivity increased 50%, with 100% of agents now using AI copilot tools daily.
But the revenue impact is what stands out. CEO Pete Caputa said the transformation enabled Databox "to drastically increase the amount of proactive outreach to in-app users," driving a 40% increase in new revenue through booked calls for sales and account management teams.
The key to Databox's approach, according to Director of Support Emil Korpar: they treated Fin "like a new team member," onboarding it gradually, monitoring its performance, and training the human team to work alongside it rather than compete with it.
Lightspeed: 43,000 AI Resolutions Per Month Across Global Teams
Lightspeed Commerce, with hundreds of support agents across multiple regions and languages, achieved 45–65% AI resolution rates after deploying Fin. The system now handles over 43,000 resolutions per month, up from 35–40% involvement on day one. Agent productivity increased 31% through copilot-assisted workflows.
Angelo Livanos, VP of Global Support at Lightspeed, described the impact: they're "dramatically augmenting where and when our team engages with customers." AI handles triage and routine queries. Agents focus on complex issues that require judgment and relationship skills. Customer satisfaction remained stable throughout, meaning the AI didn't degrade the experience even as it handled more volume.
Both Databox and Lightspeed achieved their results by training their teams alongside the AI, not by deploying tools and hoping for the best. Databox treated AI as a team member to be onboarded. Lightspeed used copilot features to augment agent skills rather than replace them. The technology was the same; the training approach made the difference.
The Skills Modern CSMs Actually Need
The CSM role is evolving from relationship manager to value manager, someone who combines commercial confidence, data literacy, and outcome ownership. ChurnZero's research makes the skills gap concrete: the top three gaps are all areas where AI can help, but only if the CSM knows how to use it.
Data literacy for AI-powered health scores. When an AI tool ranks accounts by churn risk or flags expansion opportunities, CSMs need to understand what the model measures, how confident it is, and when to override it. This isn't data science; it's the ability to interpret algorithmic recommendations critically and act on them with judgment.
Prompt engineering for customer communications. Writing effective prompts for QBR summaries, renewal talking points, account research, and customer emails. The difference between a generic AI-generated QBR and a useful one comes down to how well the CSM frames the context, constraints, and objectives. kju covers this in the prompt engineering track with scenarios grounded in customer-facing work.
Commercial confidence with AI-sourced insights. The biggest CSM skills gap (42% cite sales and negotiation) is exactly where AI can help most. AI identifies expansion signals, calculates customer ROI, and surfaces usage patterns that predict upgrade readiness. But a CSM who can't translate those insights into a confident commercial conversation doesn't benefit from having them.
AI governance and customer data ethics. CS teams handle sensitive customer data: usage patterns, support conversations, billing behaviour. As AI processes more of this data, CSMs need to understand privacy boundaries, data handling policies, and transparency requirements. The AI governance track builds this competency progressively.
The skills CSMs need aren't technical; they're extensions of existing CS competencies applied to AI-powered workflows. Data literacy extends analytical skills. Prompt engineering extends communication skills. Commercial confidence extends account management skills. The foundation is already there. What's missing is structured training that connects AI to the work CSMs actually do.
Why Self-Directed AI Learning Fails CS Teams
ChurnZero's data reveals the training model most CS teams rely on: online courses (45%), occasional manager coaching (36%), and instructor-led training (32%). Only 23% have mentorship. Only 13% get conference access. And 13% received no training at all.
This is a problem for CS specifically because the skills gap is commercial, not technical. Self-directed AI courses teach tool mechanics. They don't teach a CSM how to interpret a health score drop from 72 to 58 and turn it into a proactive conversation that saves the account. They don't build the confidence to lead a renewal conversation using AI-generated ROI data.
BCG's AI at Work 2025 data reinforces this. Employees who receive at least five hours of structured AI training become regular AI users at a rate of 79%, compared with 67% for those who get less. And only 25% of frontline workers say their leaders provide enough guidance on AI, which maps directly to ChurnZero's finding that just 35% of CSMs feel their employer is invested in their career.
The alternative is daily, structured practice. Six minutes a day, focused on CS-specific AI skills, builds lasting competency that occasional workshops can't match. That's how kju works: short daily sessions designed for working professionals who need practical skills, not certificates. When an entire CS team trains together, they develop shared standards for interpreting AI outputs, consistent approaches to AI-assisted customer conversations, and the commercial confidence to turn data into revenue.
The CS teams pulling ahead in 2026 aren't the ones with the most AI tools. They're the ones where every CSM, from the newest hire to the team lead, has the fluency to work alongside AI rather than around it.
Frequently Asked Questions
- What AI skills do customer success managers need?
- CSMs need four practical skills: data literacy for interpreting AI health scores and churn predictions, prompt engineering for customer communications and QBR preparation, commercial confidence to lead renewal and expansion conversations using AI-sourced insights, and AI governance skills to ensure customer data is handled ethically. These extend existing CS competencies, not replace them.
- Will AI replace customer success managers?
- No. Gartner found that only 20% of customer service leaders have actually reduced staffing due to AI, while 55% report stable headcount handling higher volumes. Gartner even predicts half of companies that cut CS staff due to AI will rehire by 2027. AI handles data processing and pattern recognition; relationship building, strategic advising, and empathy remain human skills.
- How does AI predict customer churn?
- AI churn prediction processes thousands of data points including product usage drops, onboarding friction, feature adoption decline, sentiment shifts in support tickets, and billing behaviour changes. Intercom's Fin AI achieves 45-65% autonomous resolution rates across enterprise customers like Lightspeed, identifying at-risk patterns weeks before traditional methods.
- How long does it take to train a CS team on AI?
- BCG research shows employees who receive at least five hours of structured AI training become regular users at a rate of 79%, versus 67% for those with less. Most CS professionals build foundational AI fluency in 6-8 weeks of consistent daily practice. Short daily sessions outperform quarterly workshops because they build habits, not just awareness.