Gurgaon, Aug 2026 : Every Customer Operations leader I meet today seems to be running an AI pilot. In leadership meetings, discussions around using Generative AI and Agentic AI to transform day-to-day operations have become almost unavoidable. CEOs are demanding higher productivity, lower operating costs and better customer experiences, while functional heads and operations leaders are under growing pressure to demonstrate how AI can deliver all three.
Yet behind the excitement, impressive presentations and ambitious roadmaps lies an uncomfortable reality: many AI business cases simply do not hold up.
Across telecom, hospitality and other industries, I have seen CX leaders launch initiatives with compelling labels — a GenAI chatbot, a human-like voice bot, an Agentic AI coaching platform, real-time voice analytics or an agent-assist solution. Some deliver impressive early results but fail to scale. Others are quietly shelved within a year. The most concerning are those that remain operational without anyone being able to clearly explain what business impact they are actually delivering.
The problem, in my experience, is rarely the technology or the vendor.
The situation reminds me of Terminator 2: Judgment Day. Skynet does not fail because the technology is inadequate. It fails because a powerful system is given enormous authority against a narrow objective, without sufficient consideration of how that objective will play out in the real world or who remains accountable for the consequences.
The parallel with enterprise AI is striking. The technology may perform exactly as designed. The problem is that the design, operating model and business objective were not thought through properly in the first place.
With more than 25 years of experience leading customer operations as a CXO across telecom and hospitality, and having been involved in AI-led transformation initiatives, I have found that successful AI programmes have surprisingly little to do with simply selecting the right technology.
They begin with the basics: redesign the process, define the outcome, prepare the data and establish accountability.
Why AI Pilots Struggle
The difference between an AI initiative that becomes a sustainable business transformation and one that remains a pilot can often be summarised simply:
| What Organisations Often Do | What Goes Wrong | What Should Happen |
|---|---|---|
| Deploy AI on existing processes | AI accelerates inefficient workflows | Redesign the customer journey first |
| Measure bot containment and deflection | Business impact remains unclear | Measure revenue, retention and cost |
| Feed fragmented data into AI | Outputs become incomplete or inaccurate | Build a strong data foundation |
| Divide responsibility across departments | Nobody owns the final outcome | Appoint one accountable leader |
This is why I believe four principles should sit at the heart of every AI programme in customer operations.
1. Re-engineer the Customer Journey Before You AI-Enable It
One of the most common mistakes I have seen is treating AI as a technology layer that can simply be placed on top of an existing operating model.
An organisation takes its existing IVR, contact-centre workflow or customer journey and migrates it onto an AI platform. The technology changes, but the underlying process does not.
The customer journey remains the same. The objectives remain the same. Escalation logic remains unchanged. Data continues to sit in disconnected CRM, billing and backend systems.
The result is predictable: AI makes a broken process run faster rather than making the process better.
That is not transformation.
When we built the e-Care platform for a large digital-first telecom service provider, the AI and self-service layer was not added at the end of the programme. It was designed into the architecture from day one and became the backbone of inbound and outbound customer journeys across channels.
That distinction is critical.
The right sequence should be:
Business objective → Journey redesign → Operating model redesign → Data readiness → AI enablement → Business outcome → Continuous optimisation
This should become the basic operating philosophy for AI-led transformation.
2. Measure Business Outcomes, Not Just Operational Metrics
Another major problem is the way AI initiatives are evaluated.
Functional teams often celebrate metrics such as bot containment, deflection rates, number of automated conversations or percentage of calls analysed. These numbers can look impressive in an operational presentation, but they do not necessarily answer the question the CEO or business head is asking:
What did AI actually do for the business?
The distinction between intermediate operational metrics and actual business outcomes is crucial.
| Traditional AI Metrics | Meaningful Business Outcomes |
|---|---|
| Bot containment | Customer retention |
| Call deflection | Cost optimisation |
| Number of automated conversations | Revenue generated |
| Calls analysed | Agent productivity |
| Response time | Resolution quality |
| AI adoption rate | Conversion, cross-sell and upsell |
When we deployed AI-led voice analytics and Agentic AI-based agent coaching across a large contact centre, we deliberately went beyond measuring the number of calls analysed.
We linked the programme to three tangible outcomes: reducing avoidable agent call volume, correcting agent errors and, most importantly, increasing meaningful cross-sell and upsell opportunities during customer interactions.
The third outcome fundamentally changed the perception of the contact centre. It was no longer viewed purely as a cost centre; it became a contributor to the organisation's top line.
That is the level at which AI business cases need to be constructed.
Operational metrics still matter, but they should be treated as leading indicators — not the final definition of success.
Every AI business case should therefore answer a simple question: Which business metric will move because of this initiative?
If that question cannot be answered before implementation, the programme is probably not ready to proceed.
3. Make Data and Information Readiness a Prerequisite
AI cannot compensate for incomplete, inaccurate or inaccessible information.
A recommendation engine is only as effective as the customer data available to it. Similarly, an AI-powered service assistant is only useful if it can actually access and execute the underlying operational systems.
If an AI assistant cannot connect with the booking system, entitlement engine, CRM, billing platform, property management system or other relevant APIs, it may sound intelligent while remaining operationally ineffective.
This is where many impressive pilots fail to scale.
The organisation demonstrates that the AI can understand a customer question and generate an excellent response. But when the customer actually needs something done — a booking changed, a refund processed, an entitlement checked or an account modified — the system cannot complete the transaction.
The AI may be intelligent, but the operation is not connected.
The groundwork therefore needs to happen before model deployment.
In hospitality, for example, improving member holiday conversion and booking NPS required much more than introducing an AI recommendation engine. We first needed to create a personalised recommendation layer that was deeply integrated with booking, inventory and customer systems.
The same principle applied to the AI support platform we built for a telecom service provider. Resolution rates improved significantly not because the model suddenly became dramatically smarter, but because the system was continuously fed real-world edge cases and customer feedback.
AI becomes valuable when it can close the operational loop, rather than simply generate an intelligent response.
Data quality, system integration, API availability, information architecture and continuous feedback therefore need to be treated as foundational components of an AI programme.
4. Put One Person in Charge of the Outcome
Perhaps the most important lesson is also the simplest: someone must own the outcome end-to-end.
In many organisations, AI projects are divided across departments.
Data science owns the model. IT owns the technology platform. Operations owns the agents. Marketing owns the customer experience. Finance owns the business case.
When results do not materialise, accountability becomes fragmented.
Everyone has completed their part, yet nobody owns the final outcome.
The AI initiatives I have seen succeed consistently have one characteristic in common: a single accountable leader owns the entire journey and has the authority to bring together data, technology, operations and business teams.
That leader must be measured against the business outcome, not simply against whether the AI platform was successfully deployed.
Technology implementation is a milestone.
Business transformation is the objective.
The Four-Pillar AI Transformation Framework
For CXOs, CTOs and Customer Experience leaders, the framework can be reduced to four questions:
| Pillar | Key Question | Expected Impact |
|---|---|---|
| Journey | Have we redesigned the process before automating it? | Better CX and productivity |
| Data | Is information accurate, complete and accessible? | Better AI decisions |
| Outcome | What measurable business result are we targeting? | Clear ROI |
| Ownership | Who is accountable from start to finish? | Faster execution and scale |
These four pillars are interconnected. A great AI model cannot compensate for a broken customer journey. A beautifully redesigned journey cannot deliver without reliable data. Good data cannot create business value if success is measured only through activity metrics. And none of these will scale without clear ownership.
So, What Does It Really Take?
The AI conversation has become dominated by questions about models, agents, platforms and automation. But the more important questions are organisational.
What problem are we solving?
Why does the customer journey need to change?
What data and systems will AI require?
Which business outcome will determine success?
Who is accountable when the expected outcome does not materialise?
These questions may sound less exciting than talking about the latest AI model, but they are far more important to delivering ROI.
The organisations that successfully move from AI experimentation to enterprise transformation will be those that understand this distinction.
From AI Pilot to AI-Powered Transformation
| Stage | Typical Organisation | Winning Organisation |
|---|---|---|
| Strategy | “Where can we use AI?” | “What business problem must we solve?” |
| Process | Automate the existing journey | Redesign the journey |
| Data | Use whatever data is available | Build the required data foundation |
| Measurement | Track AI activity | Track business outcomes |
| Ownership | Multiple departments | One accountable leader |
| Scale | Pilot after pilot | Continuous transformation |
None of these principles is particularly fashionable or complicated. In fact, they are the same disciplines that have separated successful transformation programmes from failed ones long before AI became the centre of every boardroom conversation.
What has changed is the power of the technology.
The fundamentals have not.
AI can accelerate a well-designed process, amplify a strong operating model and unlock enormous amounts of value from data. But it can also accelerate inefficiency, scale bad processes and create impressive-looking outputs without meaningful business impact.
The organisations that win the next phase of customer experience will therefore not necessarily be those with the most sophisticated AI models or the largest technology budgets.
They will be the organisations willing to do the unglamorous work first — redesigning customer journeys, fixing data foundations, integrating systems, defining measurable business outcomes and establishing clear accountability.
Technology will continue to evolve at extraordinary speed. Models will become more capable, agents more autonomous and customer interactions increasingly intelligent.
But the discipline required to make those technologies deliver business value has not changed.
The future of AI-led customer operations will not be determined by how quickly organisations deploy AI. It will be determined by how intelligently they prepare for it.
Authored by Anupam Srivastava
(Anupam Srivastava is a former CXO with 25+ years of leadership experience across Telecom, Hospitality, IT and Managed Services, including senior roles at Reliance Jio, Tata Teleservices, Mahindra Holidays and Price Waterhouse. He now advises organizations on Customer Operations Transformation, Enterprise AI & Digital Enablement and Revenue Growth)
