Value of analytics professionals in an AI age

By Ana Roy, Founder and Principal Consultant, Tuli Ops
#1 IAPA Top 25 Analytics Leaders 2023

I did not start my career in analytics.

My first role was as a marketing intern with Unilever in India, trying to assess the market size and viability of launching packaged soup. The distribution channel was not just the corner convenience shops, traditional grocers but also vegetable vendors in the wet markets. You might ask why the vegetable vendors? We were hoping that the freshness of the plump rosy tomatoes would wear off on the packaged Knorr tomato soup! I spent my days talking to grocers and vendors, doing dipstick surveys, understanding why a product that looked compelling on paper simply did not move in certain contexts.

From there, I moved into sales and business development - first for consumer durables, then into B2B technology, selling high-end Sun & IBM servers and storage systems, Cisco networking equipment and allied services and solutions to enterprises.

At the time, I did not think of this as ‘analytics training’. In hindsight, it was the most formative analytics education I could have received.

Sales Was My First Analytics Classroom

Sales is a humbling profession.

You learn quickly how to handle rejection, how to stay organised, how to remain resilient in the face of failure. But you also learn something else that matters deeply for analytics today - outcomes matter more than activity.

Very early on, I started doing my own analysis - not because someone asked me to, but because I needed to hit targets.

I analysed my sales funnel:

• How many weekly sales calls I needed to make?
• How many leads converted and typical lead times?
• Where deals stalled?
• Who was the decision maker, the influencers, the gatekeepers?
• Which products moved fastest?
• How inventory constraints affected fulfilment?

I tracked my own behaviour, adjusted my effort, and made trade-offs across channels depending on margin, volume, and likelihood of conversion.

This was not analytics in isolation. It was analytics in service of a decision - my own. That distinction has shaped how I think about analytics ever since and how I uncover value for the organization.

Data are people too

Fast-forward to a later chapter of my career, when I was formally part of an analytics team. A different industry, a different sales channel - this time, a call center. I had access to data - churn rates, NPS scores, resolution times. I could explain what was happening with confidence. What I struggled with was understanding why.

Because data, at its core, is a proxy for human behaviour and the processes we follow.

The analysis helped me form hypotheses - educated guesses about what customers might be thinking or feeling. But it was only when I listened to the calls behind the numbers that those hypotheses either held up or fell apart. I spent time shadowing call-center colleagues, sitting on the floor, sometimes taking customer calls myself. Slowly, the context provided meaning to the data.

One interaction has stayed with me even after all these years. A customer from a high-value segment called to cancel her policy. On paper, the call was a success. The agent acted quickly. The issue was resolved. All the right boxes were ticked, including that deceptively simple NPS question.

And yet, the customer churned. Listening to the call told a different story. What sounded like a request for cancellation was actually a request for reassurance. “What can you do?” “I’ve been with your company for X years.” She was not asking to leave; she was asking to be valued. A small renewal discount might have been enough.

What was missing was not data, nor intent, nor effort. It was judgement. And the space for humans to exercise it.

The organisation had optimised for speed, not understanding. Agents were rewarded for closing calls quickly, not for retaining customers who mattered most. In that system, the agent did exactly what they were trained to do.

That experience stayed with me because it reframed how I think about analytics. Insight doesn’t come from numbers alone. It emerges when data is interpreted through context, when metrics align with meaning, and when people are trusted to act on what the data is really saying.

Why Upstream Work Has Always Mattered

AI can write code, perform analysis, generate insights and produce narratives at scale and speed. Much of what was labelled as ‘analytics work’ sat squarely in the middle of the value chain - between processes and usage. This is where data engineers, scientists, and analysts resided and did their best work. So, when machines can do this work, it is understandable why analytics professionals would wonder about their place in the value stream.

For me, the answer is where it always should have been - in the upstream with the business processes, understanding of business levers, the dirty hands and battle scars accumulated by being at the coal face of it all - in the service of customers.

Even now, when I lead analytics or AI engagements, most of my time is spent before any data analysis and before building a product. I invest in understanding how the business functions end to end. I look closely at the metrics people are evaluated on, what ‘good to great’ really means in practice, and where the real constraints sit - across people, process, and technology.

This work is often invisible. It rarely shows up in dashboards or decks. But without it, the analytics work slate is irrelevant.

If you do not understand how a business works, you will struggle to ask the right questions. You will optimise the wrong things. You will build solutions that look impressive but fail the moment they meet reality.

Working upstream allows me to take an outside-in view of the system - to see constraints that those within the system, through no fault of their own, have stopped noticing. Only once that picture is clear does it make sense to move midstream and start building.

Follow the constraint… and I’ll show you the money

Every system has a limiting factor - a decision that caps performance, a process that throttles throughput, a capability that prevents scale. The work is to find that constraint, quantify its impact, relieve it - and accept that it will move elsewhere.

When constraint is removed, value is created. When trade-offs are stacked meaningfully, the value unlock compounds.

AI is exceptionally powerful at analysing systems and surfacing patterns. But it does not decide which constraint matters most, or when to intervene, or what trade-offs are acceptable. Those choices require judgement, context, and an understanding of consequences. And one cannot ask the right questions without a genuine desire to be curious about people, processes and technology and in that order.

How to work upstream – a practical guide

The upstream work I've described isn't abstract. It's a deliberate practice that can be learned. Here's how I approach it:

Start with the knowledge that already exists. Before I schedule a single meeting, I dig into whatever documentation exists - process maps, knowledge bases, previous analyses, training materials. This isn't just about efficiency (though it does respect my colleagues' time). It's about giving myself space to think critically about what I'm reading and formulate better questions. The goal isn't to understand what people do - it’s to develop hypotheses about why it works that way.

Conduct 1:1 interviews across the function. I deliberately seek out a cross-section of people - not just leadership, but frontline staff, subject matter experts, and those who've been around long enough to remember why certain decisions were made. These conversations reveal the gap between how processes are documented and how they actually run. They surface the workarounds, the inherited constraints, and the unofficial knowledge that makes things work.

Test hypotheses with data. Often, there's already data sitting behind these processes.

Once I understand the workflow and have collected the hypotheses that live in the organization's DNA, I seek out relevant data to test them. This is where grounding in statistics or lean methodologies becomes invaluable - it allows you to identify constraints systematically, perform gap analysis, and design to-be solutions with confidence.

Prioritize problems collaboratively. I don't decide what to solve in isolation. I facilitate workshops with the functional teams to prioritize together. Strong facilitation skills matter here - being able to create space for disagreement, synthesize competing priorities, and build alignment around what constraint to tackle first. This collaborative prioritisation isn't just good practice; it's what ensures the work gets used.

This approach has served me whether I'm working on customer retention, supply chain optimisation, or AI transformation. The mechanics change, but the principle remains - understand the system before you try to change it.

Analytics +

I think back to those days in the wet markets of India, watching vegetable vendors arrange their tomatoes just so, trying to understand why packaged soup wasn't moving off the shelves. I didn't have fancy analytics tools then. What I had was curiosity about people, processes, and the gap between what looks good on paper and what actually works in practice.

That gap hasn't disappeared in the age of AI. If anything, it's widened.

AI can compress the middle of the analytics value chain - the data wrangling, the pattern recognition, the insight generation. It does this work faster and often better than humans can. But it cannot replace the work of understanding context (at least not yet,) exercising judgment, or knowing which constraint matters most.

The analytics professionals who will thrive aren't those who compete with AI at tasks it does well. They're the ones who do the work AI cannot: building trust with colleagues, mapping the invisible networks that determine how decisions really get made, and translating between the language of data and the messy reality of business operations.

You don't need a traditional analytics background to do this work. You need genuine curiosity about how things work, the humility to learn from people on the ground, and the discipline to follow constraints wherever they lead.

The vegetable vendors taught me that. The call center agents reinforced it. And every engagement has proven it true.

The value of analytics professionals in an AI age isn't diminished. It's just returned to where it always should have been - upstream, with dirty hands and battle scars, in service of the people and outcomes that matter.

About the Author

Ana Roy is a Product and Transformation Executive who turns AI, analytics, and product innovation into engines of enterprise growth. With two decades of experience leading large-scale change across Australia and India, she combines operator instinct with systems-level thinking to deliver material commercial impact.

She has built self-funded analytics and product functions delivering up to 20× ROI, led transformations generating multi-million-dollar EBIT uplift, and embedded operating models that significantly accelerated delivery while sustaining high engagement.

Named Australia’s #1 Analytics Leader (IAPA 2023), Ana focuses on how AI and enterprise design can redefine value creation beyond efficiency. She does this via her transformation execution firm - Tuli Ops.