Tech CCO Says CX Leaders are Over-focused on AI Results While CX Atrophies
Oremeyi Akah explains why CX leaders must protect trust, empathy, and human connection as AI makes measurable customer outcomes easier to optimize.
Oremeyi Akah
CCO at Interswitch
Full Visibility: AI-powered quality assurance reviews every customer contact, helping teams detect risks and recurring problems without adding headcount.
Human Judgment: AI handles scale, while people manage sensitive escalations, service recovery, and firsthand understanding of customer needs.
Measurable Gains: Interswitch reduced manual requests 54 percent, increased CSAT 60 percent, and improved employee satisfaction over 70 percent.
Better Handoffs: AI integration succeeds when escalations preserve context, retain trust, and connect customers smoothly with human specialists.
Prove Value: CX leaders should establish baselines and financial measures before pilots, separating genuine returns from impressive demonstrations.
Oremeyi Akah is CCO at Interswitch, a pan-African payments infrastructure company serving both B2B and B2C segments.
We sat down with Oremeyi to hear about the impressive results she's seeing from AI integration across her CX organization. Along the way, she told us where CX organizations are going wrong in their AI integrations.
From Service Function to Strategic Function
I'm Oremeyi Akah, Chief Customer Officer at Interswitch Group — a pan-African payments infrastructure company operating across 23 African markets, connecting individuals, businesses, and governments to digital financial services.
Continue Reading for Free
Create a free account to finish this article, plus get ongoing access to timely insights and practical resources.
Have an account? Log In
The customer organization I lead spans the full spectrum of Interswitch's commercial relationships — from large financial institutions and fintechs through our Key Account Management function, to merchants, billers, and end consumers interacting across digital, contact center, and in-person channels. We serve both B2B and B2C segments.
Our CX mandate ranges from high-touch strategic account relationships with banks like Stanbic IBTC and Providus to the experience of millions of everyday Nigerians transacting on Verve cards, Quickteller, and our switching infrastructure.
My team manages relationships at the intersection of revenue, trust, and national financial inclusion — making customer experience not just a service function, but a strategic one.
How AI Radically Expands What CX Teams Can Do
A year ago, with a team of 15, we manually reviewed only 10% of customer contacts across our B2B and B2C channels. This meant we could not see 90% of our customers' experiences — every complaint, every friction point, every regulatory risk signal. We made customer experience decisions based on a very incomplete picture.
Then, we implemented an AI-powered Quality Assurance system which completely changed this. We now analyze 100% of customer contacts. We review every single interaction, pre-categorize it by severity, and flag regulatory risks, repeated failure patterns, or areas requiring critical attention.
Three things changed as a result.
- Our coverage went from 10% to 100% — without adding headcount. This represents a 10x leap in visibility with the same resources.
- We shifted from reactive to proactive. Previously, we found problems only when customers escalated. Now, we find them before they escalate — and in payments infrastructure, this distinction matters enormously, both for customer trust and for regulatory standing.
- It changed what our 15-person team does. The team moved from sampling contacts to acting on intelligence. They now focus human judgment where it matters most: on the highest severity and highest risk interactions the system surfaces.
This is what AI-enabled CX looks like in practice — not replacing people, but radically expanding what the same people can see and do.
A Real-world AI Workflow for Handling Customer Interactions
AI continuously aggregates insights across every interaction, identifying recurring pain points, emerging operational failures, and customer trends that individual cases would never reveal. Those insights feed directly into our operational excellence initiatives, product improvements, process redesign, and leadership dashboards, allowing us to address root causes rather than simply resolving individual complaints.
Here's the workflow.
Our AI-powered quality assurance platform, Gladly, automatically ingests every customer interaction — whether through voice, email, chat, or other digital channels.
First, the AI understands the conversation. It classifies the issue, assesses sentiment, determines severity, identifies potential regulatory or reputational risk, detects repeat failure patterns, and predicts the most appropriate routing and likely resolution path based on similar historical cases.
The workflow then diverges based on complexity.
For straightforward issues, AI directs the case to the right team with relevant context already assembled for faster resolution.
For higher-risk or high-value cases — particularly those involving our banking partners or issues with regulatory implications — we immediately escalate the case with a recommended priority. Human expertise then takes over. Our specialists review the AI's recommendations, apply judgment, consider the broader customer relationship, and determine the appropriate response and service recovery.
The workflow doesn't end when the case is closed.
AI continuously aggregates insights across every interaction, identifying recurring pain points, emerging operational failures, and customer trends that individual cases would never reveal. Those insights feed directly into our operational excellence initiatives, product improvements, process redesign, and leadership dashboards, allowing us to address root causes rather than simply resolving individual complaints.
Why AI and Human Touch are Both Necessary
AI expands what we can see. Humans decide its meaning and what to do about it. Both are necessary. Neither is sufficient alone.
Overall, AI handles scale work — work requiring breadth, speed, and pattern recognition across millions of data points. Specifically, this includes complete contact analysis, severity pre-categorization, issue prioritization, regulatory risk flagging, and identifying repeat failure patterns before they become crises. It also provides an early warning layer for escalation — the system tells us what needs human attention and how urgently.
Three tasks remain human, and deliberately so.
- Escalation judgment: AI tells us what to look at; humans decide what to do, especially when relationship stakes or regulatory sensitivity are high.
- Service recovery: No model handles this — especially on the B2B side, where a conversation with a tier-1 bank is a relationship moment, not just a ticket.
- Understanding the real customer experience: This is the one I care most about. Real customer experience is not reducible to data alone. We invest in ethnographic and immersive customer engagement. We sit with customers, watch how they interact with our products, and listen to what they say when nobody runs a survey. This kind of insight — the unfiltered, unprompted, behavioral truth — informs our most important journey design decisions. AI can identify what went wrong at scale. A human in the room understands why.
AI expands what we can see. Humans decide its meaning and what to do about it. Both are necessary. Neither is sufficient alone.
How AI Improves Request Handling and Satisfaction
The numbers tell a compelling story.
As I mentioned, our AI-powered QA system moved us from reviewing 10% to 100% of all customer contacts with the same 15-person team — a 10x visibility gain at zero additional headcount cost.
Our AI-powered self-service portal also reduced manual request handling by over 54%, and critically, it doesn't just deflect — it prioritizes requests intelligently based on customer volume and trend data, so humans address the right things first.
And the downstream effect on experience has been measurable: CSAT grew by 60%. Turnaround times improved significantly. And perhaps the figure I'm most proud of — employee satisfaction improved by over 70%. When your team stops drowning in volume and starts doing meaningful work, that shows up in how they show up for customers.
Qualitatively, we've shifted the culture of the team from firefighting to intelligence-led management. That is harder to quantify but easier to feel — in how we run reviews, how we engage with accounts, how we talk about the customer in leadership conversations.
Why CX Leaders Must Focus on Handoff and Escalation
Most redesign energy focuses on the front end—the chatbot, the personalization engine, and the automation of repetitive contacts. This work is necessary and largely already underway. But the seam that determines whether AI helps or damages the relationship is often treated as a feature to be configured later, despite being the system that most determines trust.
Research converges on this from several directions. A 2025 CX study found 79% of respondents prefer a human over an AI agent for complex issues — so the escalation path isn't optional; what's optional is whether it's graceful. The cost of getting it wrong is now quantified: 48% of customers would abandon a brand if they had to re-explain their issue after being transferred to a human, and 40% would leave if they had to re-verify their identity. The failure mode isn't slowness — it's that a bad handoff erases everything the customer already told you, and the trust the AI built evaporates just as the stakes rise.
Two findings make this board-level worthy. First, the industry quietly admits it over-automated: Gartner predicts that by 2027, half of companies that cut customer service staff because of AI will rehire for similar work under different titles. Second, the strategic verdict is explicit — the brands most likely to earn trust with AI agents are not the ones racing to eliminate human interaction, but the ones designing for clear escalation, context retention, and honest disclosure.
Why AI Does Not Reduce the Need for Human Contact
Two areas didn't pan out how we expected, but both taught us something important.
The first: We thought AI could significantly reduce the need for human contact. It didn't — not as we hoped. Volume reduced, yes. But the need for genuine human interaction didn't. Reflecting on it, that shouldn't have surprised us. We are in the business of money — people's money, businesses' money — and when something goes wrong, or stakes feel high, people want a person. Not a bot, not a portal. A human voice that understands, takes responsibility, and acts. AI can contain volume. It cannot replicate presence. We are human, after all.
The second: product design intelligence. We expected AI-derived data to tell us how customers wanted products shaped. It couldn't — at least not fully. Behavioral data tells you what people do. It does not tell you what they wish existed, or what frustrates them at a level they haven't yet articulated. For that, we went back to basics — sitting with customers, having real conversations, doing the ethnographic work. The insight that moved our product thinking came from those rooms, not from dashboards.
AI is exceptionally good at scale and pattern. It is not good at nuance, emotion, or imagination. Those remain stubbornly, irreducibly human — and perhaps that is exactly as it should be.
Why CX Leaders Must Learn How to Measure Results
Our first pilot worked. The model performed, the dashboards lit up, the stakeholder demos landed. I didn't understand that none of that answered the only question the C-suite would ask six months later: "What did this return, and how do you know?"
I'd optimized for proof of capability. I should have optimized for proof of value by establishing a baseline for the metric we intended to move, isolating AI’s contribution from other changes, and agreeing with Finance in advance on how to measure success. We hadn't, so when the investment case came due, I reconstructed the baseline after the fact and defended attribution I couldn't cleanly prove.
The harder lesson sits with the C-suite, not the vendor or the tool. A bright demo is easy to buy and hard to justify. Once you own the budget line, enthusiasm isn't the deliverable — a number you can stand behind in front of the board is. Most AI pilots impress. Far fewer are instrumented to prove they earned their cost, and you must design that instrumentation in from day one, because you cannot retrofit it.
So the thing I'd tell anyone launching their first pilot is this: Decide how you'll prove the return before you build anything, and treat the bright result with suspicion until the number holds up.
How the Unmeasurable Parts of CX are Slowly Failing
Here's what I'm concerned about. We're automating the cheap-to-measure half of customer experience and quietly letting the expensive-to-measure half atrophy — and the numbers will tell us we're winning until we've lost something we can't rebuild.
AI excels at the parts of CX that show up in dashboards: deflection, handle time, cost per contact, resolution speed. So we optimize these, because a board finds them legible. But the things that build a durable relationship — why a customer feels safe, what earns forgiveness after a failure, the sense that someone owns their outcome — don't sit in the data, and we can't automate them.
So, my worry isn't that AI will do measurable work badly. It's that by making measurable work so cheap and visible, it slowly redefines "good CX" as only the measurable work. The unmeasured half doesn't fail loudly; it just thins out while every metric on the wall keeps improving.
The part I haven't resolved — and this is the honest version — is that I can't cleanly separate my own bias from the real risk. It's convenient for someone whose career relies on human judgment and empathy to believe those things are irreplaceable. I'm guessing how much of my conviction is genuine foresight versus professional self-interest.
Some of the "AI can't do empathy" position is probably a comfort story CX leaders tell themselves, and the capability line moves faster than that story assumes. So the thing genuinely on my mind is two-sided: I believe we're under-protecting the unmeasured half of CX, and I'm suspicious of how much I want that to be true.
The practical residue is that the leaders who will succeed will make the unmeasured stuff measurable — instrument trust, ownership, and post-failure recovery directly — rather than automating it away or defending it on faith. I currently rely on faith for part of it, and I don't think faith is good enough.
Three Pieces of Advice for CX Leaders Implementing AI
Overall, here's my advice to CX leaders:
- Stay real. Stay compassionate. The noise around AI in customer experience is enormous — and a lot of it is hype chasing hype. Cut through it by anchoring to one question: "Does this make life better for my customer?" Not better on a dashboard. Better in their day. If the answer is yes, pursue it. If it's just impressive, leave it. Customers don't remember your technology stack. They remember how you made them feel. That hasn't changed, and AI doesn't change it.
- Keep learning — relentlessly. This landscape is moving faster than any of us were trained for. The leaders best positioned to serve their customers five years from now are those curious enough to admit today that they do not have everything figured out. Read. Ask questions. Talk to your peers across industries and geographies. Sit with your customers. The moment you stop learning in this environment, you start falling behind — quietly, and then suddenly.
- Don't be afraid to experiment. Not every initiative will land. Some of our biggest insights came from initiatives that didn't work as planned. That is not failure — that is intelligence, expensively acquired. Build a culture where your team feels safe to try, to learn, and to adjust. Organizations that will lead in CX do not have the most sophisticated AI — they have the most courageous and curious teams behind it.
Strive to win for people, not for presentation.
Follow along
You can follow Oremeyi Akah's work on LinkedIn.
More expert interviews to come on The CX Lead!
