The AI technology is impressive, accessible, and often genuinely useful. But as enthusiasm accelerates, many organisations are discovering an uncomfortable gap between expectation and reality. Despite their heavy investment, real value remains elusive.
This article is shaped by a recent conversation on The CX Algorithm Podcast with Professor Kiron Ravindran, where we explored why this gap exists — and why it has less to do with the technology itself, and more to do with how organisations frame value, measure success, and allocate attention.
Across industries, leaders feel growing pressure to demonstrate progress on AI. Boards ask about it. Investors expect it. Competitors announce initiatives. Internally, teams feel compelled to embed AI into every function, every initiative, and every layer of the organisation.
AI is no longer seen as optional. CFOs increasingly rank it alongside cost-cutting as a top strategic priority, and leadership teams are expected to show tangible progress quarter after quarter.
The result is that AI gets deployed broadly across functions — marketing, CX, operations, HR — often without clarity on where value should emerge first. The mandate shifts from “solve a problem” to “use AI.” Attention fragments, expectations inflate, and AI is positioned as a magic bullet rather than a capability that requires time, skill, and context to mature.
This is where many AI programs begin to struggle — not because the technology is flawed, but because focus is diluted and success is judged through short-term lenses and gains.
This gap is part of a broader concept known as the productivity paradox. AI delivers impressive surface-level outputs, particularly in the early stages. The interface is intuitive. The responses are convincing. It feels like productivity has increased.
But deeper research tells a more nuanced story.
Studies across programming, call centres, and knowledge work show that AI delivers productivity gains only under specific conditions — when tasks are well-defined, complexity is manageable, and domain expertise is either absent or applied effectively. Novices often benefit more than experts. In many cases, AI doesn’t replace jobs; it reshapes how work is distributed.
The problem arises when organisations extrapolate these early gains into unrealistic expectations of continuous growth. When those expectations aren’t met, AI initiatives are labelled failures — not because they failed, but because they were misunderstood.
This misunderstanding becomes most visible in how AI is measured.
Traditional ROI assumes a direct, near-term relationship between investment and return. But as Kiron explained, this assumption breaks down quickly for AI. Technology introduces lag effects, indirect benefits, and capability shifts that don’t map cleanly to financial outcomes.
A marketing team may generate better ideas. A service agent may resolve issues faster. A strategist may spot patterns earlier. These are real improvements, yet they don’t always appear on a quarterly ROI scorecard. As a result, organisations conclude they are “not getting value,” even as internal effectiveness quietly improves.
A more useful way to think about AI value is through return on equity — not as a financial ratio, but as a way of tracing value through process.
Consider a marketing team using generative AI to support campaign development. A traditional ROI question would be simple: Did AI increase revenue? If the answer isn’t clear, the initiative is deemed unsuccessful.
A return-on-equity lens breaks this into stages. First, did AI improve the quantity or quality of campaign ideas? Second, did those ideas translate into campaigns that actually launched? Third, did those campaigns lead to customer engagement or conversion?
Tracking performance across these steps reveals where value is created — or where it breaks down. Strong ideas may fail in execution. Effective campaigns may struggle because of product-market fit. The conclusion isn’t that AI failed, but that the system around it needs adjustment.
This approach turns measurement into diagnosis, not judgment.
Automation is often treated as the primary goal of AI. It’s visible, measurable, and easy to justify in a business case.
But the deeper value comes from augmentation.
Augmentation means using AI to support human judgment rather than replace it. It helps people focus attention, recognise patterns earlier, and navigate complexity with greater confidence. In functions that depend on context, empathy, or creativity — such as customer experience, leadership, and product design — augmentation consistently outperforms full automation.
This also explains why experience still matters. AI amplifies domain knowledge rather than eliminating the need for it.
When these ideas come together, a different leadership question emerges.
The issue is no longer “How fast can we automate?” It’s “Where should AI help us think better, decide faster, and learn sooner?”
Organisations that ask this question resist the urge to short-circuit people and process. They invest in skills, job design, and governance alongside technology. And they accept that meaningful value takes time to surface.
That’s not a failure of AI. It’s the reality of transformation.
🎧 Listen to the Full Conversation
If this perspective resonates, the full episode explores these ideas in much more depth, alongside broader themes around organisational readiness, skills, and the future of work. The link to the episode is available below for those who want to continue the conversation.
Spotify https://open.spotify.com/episode/52mS4ZrJxUjEtcXDh4i6wq Apple Podcasts https://podcasts.apple.com/podcast/id1868364883?i=1000744809801 Amazon Music https://tinyurl.com/hhdvs3cj
Dr. Eslam Afifi is a customer experience and data leader whose work sits at the intersection of human behaviour, technology, and decision-making. His career spans multiple industries and regions, where he has helped organisations navigate complexity, design better experiences, and make sense of the growing role of data and AI in everyday decisions.
Curious by nature and sceptical of one-size-fits-all thinking, Eslam created The CX Algorithm to explore how customer-centricity actually works in practice. Through conversations with leaders from different contexts, the podcast unpacks the choices, trade-offs, and design decisions that quietly shape customer experience over time.
Follow The CX Algorithm for thoughtful conversations on experience, intelligence, and leadership — and a reminder that great CX is designed, not declared.
There is a phrase I hear in boardrooms constantly. Every time I hear it, I tell myself I should reward myself with a treat. That phrase is "channel strategy." Not…
Reflections from my conversation with Mohamed Soliman, Regional VP at Energy Exemplar, on The CX Algorithm Podcast.
What does it take to survive — and thrive — through the internet boom, the booking.com wave, and now the rise of agentic AI? I recently sat down with Kate Couling…
