“Metrics matter. But people matter more. The day we forget that — the day we let an algorithm make the call that should have been a conversation — is the day we start losing customers we didn’t have to lose.” Laurent Pierre, Jr.
The future of customer support isn’t humans versus AI — it’s humans powered by AI.
It was 10:35 PM on a Thursday night. I remember because I was checking the timestamp on the escalation email when my phone rang. A senior executive from one of our largest global accounts. Not his assistant. Him. And he wasn’t calling to complain about downtime or a missed SLA. He was calling because he felt like no one was listening.
We had recently rolled out a new AI-powered ticketing system. Smart routing. Automated acknowledgments. Deflection logic built to reduce inbound volume by 30%. On paper, it was a win. Cost-per-contact was trending down. Handle times were improving. The dashboard was green.
And yet here was one of our most strategic customers, on the phone at late at night, telling me that his team had submitted five tickets over two weeks and every single one had been “resolved” by an automated response that didn’t actually solve anything. The chatbot had closed loops. It had not solved problems. That night changed how I think about AI in customer support for the next decade.
I’ve spent more than 30 years leading customer-facing organizations at IBM, Microsoft, Wolters Kluwer, NielsenIQ, and now Precisely, where I serve as SVP of Global Customer Support. I have seen every wave of technology promise to transform the support function. And I’ll be direct with you: AI is different. The potential is real. The risk of getting it wrong is also very real.
This article is for the VPs, Directors, and CX leaders who are either mid-implementation or staring down the barrel of an AI transformation and wondering how to do it right. Not just efficiently — right. Because in my experience, those two things are not always the same.
The State of Customer Support in 2026: A Burning Platform
Let me give you the lay of the land, because the numbers tell a story that should make every support leader sit up straight.
According to Salesforce’s 7th State of Service Report, AI is now the #2 priority for service leaders globally — second only to improving the customer experience itself. By 2027, AI is expected to handle 50% of all customer service cases, up from just 30% today. That is not a gradual shift. That is a structural transformation happening in real time.
By the Numbers • Median AI deflection rate for enterprise CX programs in 2026: 41.2% — with top-quartile organizations hitting 58.7% (Zendesk / Salesforce) • AI chatbots reduce average handle time by 30–50% on routine interactions • Junior tier-1 agent job postings dropped 21% in 2025, with a further 24% reduction planned in 2026 (Gartner) • 82% of service professionals say customer expectations are higher than ever — yet agents spend less than half their time (46%) actually with customers (Salesforce) • 54% of customers say they trust human agents more than AI for product or service recommendations (Gartner, 2025) |
The pressure is coming from every direction at once. Customers expect 24/7 availability, faster resolutions, and personalized interactions. Boards want cost-per-contact trending down. And your frontline agents — the people who actually deliver the experience — are burning out under the weight of high volume, repetitive work, and the anxiety of an uncertain future.
The good news: AI genuinely can help with all of this. The bad news: most organizations are implementing it in a way that creates new problems faster than it solves old ones. And the leaders who don’t have a clear philosophy going in are the ones most likely to find themselves on a call at midnight with an angry strategic account.
“Organizations that only use AI to reduce costs risk missing a strategic opportunity. The real advantage comes from combining AI efficiency with human judgment, empathy and experience to deliver outcomes that technology alone cannot.” — Gartner, April 2026 |
Why Most AI Support Implementations Fail
I want to be honest about something the vendor decks don’t tell you: most AI support deployments underperform because of people problems, not technology problems.
The technology, frankly, is the easy part. You can stand up a chatbot in weeks. You can configure routing logic in days. But if you haven’t thought carefully about the human dynamics on both sides of the conversation — your agents and your customers — you are going to struggle.
The Agent Fear Problem
When you tell a frontline support team that you’re rolling out AI, what they hear is: “We’re replacing you.” I’ve seen it happen again and again. The best agents — the ones with institutional knowledge, the ones customers ask for by name — start quietly updating their LinkedIn profiles. Engagement drops. Quality suffers. And you’ve undermined the very human layer that AI needs to work alongside.
Failure to address this upfront is the single most common and most costly mistake I see leaders make.
The Customer Frustration Problem
The second failure mode is what I call “deflection theater” — when a bot closes a ticket without actually resolving the underlying issue. The customer’s case disappears from the queue. Your deflection metrics look great. But the customer is still stuck, still frustrated, and now also insulted that the system pretended to help them.
Research backs this up: nuanced complaints and complex issues break through AI resolution at only about 25% success rates, even as simple password resets deflect at 70% or higher. When organizations treat all tickets the same and push everything through the same AI funnel, they create a CSAT crisis hiding behind a cost efficiency win.
The Leadership Confusion Problem
Finally — and this one stings a little because I’ve been guilty of it — leaders sometimes confuse deflection with resolution. They are not the same thing. A deflected ticket is a ticket removed from the human queue. A resolved ticket is a customer whose problem is actually solved. These metrics can look identical on a dashboard and be completely different in reality.
Watch Out For This |
The Three-Layer Support Model: A Framework That Actually Works
Over the past several years, I’ve developed and refined a framework I call the Three-Layer Support Model. It’s not a technology architecture — it’s a leadership philosophy expressed as an operating model. The core idea: not every customer interaction is the same, and your support structure shouldn’t treat them as if they are.
Here’s how the layers work:
Layer 1 AI-First Resolution — Routine, High-Volume, Time-Sensitive What it handles: Password resets, order status inquiries, FAQ responses, basic troubleshooting, billing inquiries with clear resolution paths, account lookups. When it activates: Immediately, on first contact, 24/7. No human in the loop unless the AI signals low confidence or the customer explicitly requests escalation. How to resource it: Investment goes into knowledge base quality, intent classification accuracy, and clear escalation triggers — not headcount. How to measure it: True Resolution Rate (not just deflection), Time-to-Resolution, and Customer Effort Score on AI-handled interactions. If CES is climbing, your AI isn’t actually resolving — it’s redirecting. |
Layer 2: AI-Augmented Agents — Complex, Emotional, Nuanced |
Layer 3: Expert Human Escalation — Strategic, Relationship-Critical |
The magic of this model is in the transitions. Each layer has clear escalation triggers and clear ownership. Customers never feel abandoned mid-journey. Agents know exactly what they’re responsible for. And AI is deployed where it genuinely excels — speed, consistency, scale — while humans are protected for what they genuinely excel at: judgment, empathy, and relationship.
Layer | Interaction Type | Primary Resource | Key Metric |
|---|---|---|---|
Layer 1 — AI-First | Routine, repetitive, high-volume | AI / Chatbot / Self-Service | True Resolution Rate, Customer Effort Score |
Layer 2 — AI-Augmented | Complex, emotional, multi-part | Agent + AI Co-Pilot | FCR with AI Assist, Agent Satisfaction Score |
Layer 3 — Expert Human | Strategic, relationship-critical | Senior Expert / Executive | Net Revenue Retention, Executive NPS |
How to Lead the AI Transition in Your Support Organization
Getting the model right is necessary. Getting the leadership right is what makes it work.
Here is what I’ve learned — the hard way — about leading an AI transition in a support organization:
1. Involve Your Frontline Before You Flip the Switch
Your agents know things your technology vendors don’t. They know which ticket types are genuinely routine and which ones just look routine. They know which customers need a personal touch even on a password reset. They know where the knowledge base falls short. Bring them into the design process early — not as a change management checkbox, but as genuine subject matter experts. The best AI implementation I’ve ever been part of was co-designed with the frontline team. It showed in the results.
2. Be Radically Transparent About What Changes and What Doesn’t
Don’t let the rumor mill fill the information vacuum. Tell your team exactly what AI will do, what it won’t do, and what that means for their roles. Yes, some roles will change. Be honest about that. According to Gartner’s 2026 research, 85% of service and support leaders are expanding human agent responsibilities — not eliminating them. Share that context. Help your team see where they’re going, not just what they’re leaving behind.
3. Reskill Agents as AI Orchestrators
The agent of 2026 is not a ticket processor. They are a judgment engine — someone who knows how to work with AI tools to deliver outcomes AI alone cannot. Train for this explicitly. Teach agents how to evaluate AI-generated suggestions, when to override them, how to use AI summaries to prepare for difficult conversations, and how to feed the system with better data when something goes wrong. This is a skill set. Invest in it.
4. Protect the Human Moments That Matter
Not every interaction should be optimized for efficiency. Some interactions are the relationship. When a long-tenured customer calls in distress after a service failure, that is not a ticket to be routed and deflected. That is a moment of truth. Build your system so that human moments — the ones where empathy and accountability can rebuild trust — are protected by design, not left to chance.
Leadership Principle AI should make your best agents better, not make your average agents irrelevant. If your implementation strategy doesn’t have a plan for both of those things, go back to the drawing board. |
The Metrics That Actually Matter
Let me be direct: if your AI strategy is being measured by deflection rate and handle time alone, you are flying the plane with half your instruments covered.
Here are the metrics I believe tell the complete story of a healthy, AI-augmented support organization:
Customer Effort Score (CES)
How hard did the customer have to work to get their issue resolved? This is the most honest measure of whether your AI is helping or just rerouting. A low CES means frictionless resolution. A high CES means you have a Layer 1 containment problem — things are being “resolved” without actually being fixed.
First Contact Resolution WITH AI Assist (FCR+AI)
Standard FCR tells you whether the issue was resolved on first contact. FCR+AI tells you whether AI assistance improved or degraded that resolution. Track them separately and together. If AI-assisted interactions have lower FCR than human-only interactions in Layer 2, your co-pilot tooling needs work.
Agent Satisfaction Score (AgSAT)
This one doesn’t get enough airtime. Your agents are the engine of your support operation. If they’re miserable, your customers will feel it — regardless of how good your AI is. Salesforce data shows reps using AI spend 20% less time on routine cases, freeing up an estimated four hours per week for more complex, higher-value work. That should show up in your AgSAT. If it doesn’t, something is off.
Value-Per-Contact
What is the business outcome generated by each support interaction — retention saved, upsell identified, relationship deepened? This metric reframes support from a cost center to a value driver. Salesforce projects that agentic AI will boost upsell revenue by 15% for service teams. That means your Layer 2 agents, armed with the right AI tools, can generate revenue while they resolve issues. Measure that.
Old Metric | Why It’s Incomplete | Better Metric |
|---|---|---|
Deflection Rate | Measures removal from queue, not resolution quality | True Resolution Rate + Customer Effort Score |
Average Handle Time | Incentivizes speed over quality | First Contact Resolution WITH AI Assist |
Cost-Per-Contact | Ignores value generated per interaction | Value-Per-Contact |
CSAT (point-in-time) | Snapshot; doesn’t capture cumulative effort | Agent Satisfaction Score + CES (combined) |
The Human Advantage: What AI Can Never Replace
I want to close the loop on something, because I’ve seen this debate go sideways in boardrooms: AI will not replace the human element in customer support. Not the parts that matter most.
Here’s what AI genuinely cannot do — at least not yet, and arguably not ever in the way humans do it:
Emotional Intelligence Under Pressure
When a customer is not just frustrated but scared — when a system failure is affecting their business, their patients, their operations — they need to feel heard by a human being who actually understands the stakes. AI can detect sentiment. It cannot hold space for someone’s fear. That is a human skill, and it is irreplaceable.
The Judgment Call in Gray Areas
Real support involves ambiguity. A customer who technically violated a service agreement but had extenuating circumstances. A request that is outside policy but would be the right thing to do for the relationship. These calls require contextual wisdom, moral reasoning, and an understanding of long-term relationship value that no algorithm can fully replicate. Gartner found that 54% of customers trust human agents more than AI for product or service recommendations — and that number holds even as AI capabilities advance.
De-escalation Mastery
I’ve watched great support agents do things in three minutes that could prevent a six-figure churn. Not because they said the magic words — because they were genuinely present, accountable, and willing to own the problem. That is a craft developed over years of practice. AI can surface the right talking points. It takes a human to land them with credibility.
Relationship Building at Scale
The best support leaders I know have built genuine relationships with their key accounts — relationships where customers call them when something goes wrong, because they trust they’ll get a straight answer and real action. AI can personalize interactions. It cannot build trust the way a human can. Not at the level that protects revenue and drives retention.
| “Metrics matter. But people matter more. The day we forget that — the day we let an algorithm make the call that should have been a conversation — is the day we start losing customers we didn’t have to lose.“
|
The organizations that will win in the AI era of customer support are not the ones that eliminate human touchpoints the fastest. They are the ones that are most thoughtful about which human touchpoints to protect — and most disciplined about investing in the human skills that AI genuinely cannot replicate.
Let’s Keep This Conversation Going
• Read The Unfinished Work — my book on the leadership journey, purpose, and why the most important work we do is often the work we do on ourselves The future of customer support belongs to leaders who refuse to choose between efficiency and humanity. I believe you can have both. This is how you build it. |
Laurent Pierre Jr.
SVP, Global Customer Support • Precisely | Former IBM • Microsoft • NielsenIQ
Laurent Pierre Jr. is a global technology executive with more than 30 years of experience leading customer-facing organizations at scale. As SVP of Global Customer Support at Precisely, he oversees worldwide support operations serving enterprise customers across data integrity, integration, and analytics solutions. His leadership philosophy — metrics matter, but people matter more — drives everything he writes, speaks, and builds. The Future Ready Leader is his platform for sharing what he has learned leading teams through transformation, disruption, and growth.
Topics: AI Customer Support Customer Support Leadership Scaling Customer Support with AI AI in Customer Service Customer Experience Strategy
Leave a Reply