Most retailers do not have a data problem.

They have a decision problem.

Every day, retailers generate enormous volumes of information through point-of-sale systems, inventory platforms, loyalty programmes, mobile apps, websites, social commerce, delivery services and digital payments.

The data is there.

What is often missing is the ability to connect it, interpret it and turn it into decisions quickly enough to improve customer experience, operational efficiency and profitability.

That was the central issue at the heart of a discussion panel I had the privilege of moderating at the RETRAK Retail Summit 2026 on Thursday, 14 May 2026, at the Sarit Expo Centre in Nairobi.

The panel was titled:

Smarter Retail: Turning Data, AI and Insights into Competitive Advantage

It formed part of the wider two-day RETRAK Retail Summit, which brought together retailers, technology providers, financial-services organisations, consumer brands, entrepreneurs and other stakeholders from across the retail value chain under the theme “Retail Horizons: Shaping Tomorrow’s Marketplace.”

A Significant Moment for Kenyan & African Retail Markets

I opened the panel by observing that we are having this conversation at a fascinating — but also difficult — moment for Kenyan and African retail.

Modern retail chains continue to expand. Informal traders and neighbourhood dukas remain essential to the daily lives of millions of consumers. E-commerce, social commerce, mobile applications and digital payments are becoming increasingly important parts of the customer journey.

At the same time, consumers are more value-conscious, operating costs remain high and retailers are contending with currency volatility, import costs, margin pressure and an increasingly competitive physical and digital marketplace.

African digital commerce is growing rapidly, and the panel discussion was framed around the reality that smart retail must work in both the modern shopping mall and the neighbourhood kiosk.

However, simply becoming more digital does not automatically make a retailer smarter.

Many retailers already have point-of-sale, payments, inventory, customer, online-shopping and loyalty data. The problem is that much of it remains fragmented across different platforms, departments and monthly reports instead of informing decisions in real time.

The Panel

I was joined by four accomplished panelists whose experience covered payments, retail entrepreneurship, digital financial services, marketing, product development and retail technology:

• Judy Waruiru, Regional Managing Director, Network International — Judy is a fintech executive and digital-payments strategist with experience leading digital and commercial transformation across payments, aviation and media. She brought an especially valuable perspective on how payments infrastructure and transaction data can help retailers understand customer behaviour, reduce friction, manage fraud and improve commercial performance.
• Sonal Haria, Co-Founder and CEO, Canvas Cosmetics; Co-Founder, CB Consulting & Media Group — Sonal is a retail entrepreneur, marketer and brand builder.Through Canvas Cosmetics, she has helped build an inclusive Kenyan-born beauty brand that has expanded beyond Kenya into international markets. Her contribution grounded the conversation in the practical realities of developing products, pricing them, building communities and growing a consumer brand with limited resources.
• Eric Muriuki, Group Director, Digital Business and CEO, LOOP DFS, NCBA Group — Eric is a financial-services and digital-banking executive with more than 25 years of experience across Africa. He has been involved in the development and growth of major digital financial-services platforms, including M-Shwari and NCBA LOOP. He brought a platform perspective on data readiness, enterprise digitisation, artificial intelligence, business financing and productivity.
• Siddesh Narkar, Head of Product, Compulynx — Siddesh has more than two decades of experience in retail technology and B2B software across multiple markets. His contribution focused on the technology foundations that retailers need to turn point-of-sale, inventory, loyalty and operational data into practical decisions rather than simply generating more dashboards.

The Real Problem: Data-Rich but Insight-Poor

I began by asking each panelist what the phrase “data-rich but insight-poor” means in the everyday reality of African retail.

Sonal immediately introduced an important qualification.

Data and AI can produce useful outputs, but the quality of the decision still depends heavily on the person asking the question and interpreting the answer.

“The output can be very good with AI and data, but what is really important is the human input — knowing what you are looking for and how you will use the insight.”

Her point was that data should inform decisions rather than replace commercial experience, customer understanding or human judgement.

This is particularly relevant in marketing.

Conversion data is usually relatively easy to measure. Awareness, consideration, brand value and emotional connection are more complex. A retailer therefore needs to combine quantitative performance data with a deeper understanding of the customer and the brand.

Unifying The Customer Journey

For Judy, the problem begins with fragmentation.

A customer may discover a product through social media, browse it on a website, compare it through a mobile application and complete the purchase inside a physical shop.

Each interaction creates data.

However, when the retailer cannot connect those interactions, it sees several isolated transactions rather than one customer journey.

“It is about defragmenting the data, putting it together and using it to determine who your customer is, what their journey is and where you can add value.”

This is the difference between merely being present across several channels and having a genuinely omnichannel view of the customer.

The first is a distribution strategy.

The second is an insight capability.

Judy explained that unified customer and payments data can help retailers identify changes in purchasing behaviour, predict churn, detect unusual transactions and understand which channels are contributing to — or obstructing — conversion.

That means a retailer may detect declining lunchtime traffic at a particular branch before the decline becomes a serious revenue problem. It can then investigate whether customers have shifted to delivery, moved to another channel, encountered a payments problem or gone to a competitor.

The important shift is from reacting to lost customers to recognising the warning signs before they leave.

Data Readiness Comes Before AI Readiness

Eric approached the problem from another angle.

Almost every business has data. That does not mean every business has data that is ready for AI.

The information may be poorly labelled, incomplete, duplicated, stored in incompatible systems or disconnected from the processes where decisions are made.

“The first question to ask is: Is our data ready?”

That question is much more important than asking which AI platform the organization should purchase.

A business may have years of transactions and hundreds of dashboards, but if the data has no consistent structure or meaning, an AI system will struggle to generate reliable insights from it.

Eric described the opportunity to bring together enterprise operations, payments, customers, suppliers, inventory and financial data, and then place that internal picture within the context of wider industry information.

This can change the quality of conversations that business owners have with banks and other financial partners.

A bank may traditionally see only a company’s statements and repayment history. A business owner understands the wider commercial context: customers, suppliers, stock cycles, seasonal demand and expansion plans.

Better-organized operational data can reduce that information gap and lead to more intelligent discussions about working capital and credit.

AI Should Strengthen Judgement, Not Replace It

One of Eric’s most useful observations was that AI can be seen as an intelligent wrapper around the business.

It can draw on data inside the organisation, information held by trusted partners and relevant external market data.

However, it does not remove management responsibility.

“AI does not replace the business owner or the management team. It helps you obtain opinions from a much broader set of data much faster — and the judgement is still yours.”

That distinction matters.

The objective is not to allow an algorithm to run the retailer blindly. It is to help managers ask better questions, recognise patterns earlier and evaluate more options before making a decision.

AI should improve the quality and speed of management judgement, not provide an excuse to abandon it.

Clean Data Is An Asset Only When It Is Used

Siddesh reinforced the foundational importance of clean data.

“Without clean data, you will not be able to make the right decisions.”

Retailers now collect both structured data, such as sales and inventory records, and unstructured information from browsers, mobile devices, customer communication and other sources.

Modern AI tools can work with a much wider range of information than traditional reporting systems.

However, Siddesh made an equally important commercial point: data that sits unused is an investment that is not growing.

A retailer should not begin by attempting to build a giant AI programme covering every business function.

It should select one or two meaningful problems.

That might be improving replenishment, optimizing stock levels or adjusting prices without destroying sales volumes.

A focused use case is easier to implement, monitor and improve. It also provides evidence that can support the next phase of investment.

How Canvas Cosmetics Used Data To Build Products

Sonal provided one of the most practical examples of the discussion.

Canvas Cosmetics entered a market where much of the available data described either mass-market products or highly premium alternatives.

The brand wanted to occupy the space between the two: providing a premium experience while remaining accessible.

The company therefore used a combination of industry information, competitor observations, customer behaviour and its growing community to understand price sensitivity, product adoption and brand loyalty.

One of its early questions was straightforward: which product would provide the most effective entry point into the market for a young, self-funded brand?

The answer was lipstick.

As the company grew, customer behaviour also revealed the importance of complexion products in Kenya. That insight contributed to more than two years of research and development and the eventual launch of a complexion range containing more than 30 stock-keeping units.

Data did not simply improve an advertising campaign.

It influenced product development, category prioritisation, pricing and the deployment of limited capital.

That is precisely what good retail data should do.

It should move beyond describing what happened and begin influencing what the business builds next.

Payments Are Not Merely A Cost

Another major thread concerned the role of payments.

Retailers often treat payments as an operational expense: a necessary cost incurred when the customer checks out.

Judy challenged that interpretation.

Digitized payments create behavioural information. They can show where customers shop, which channels they prefer, how purchasing patterns change, where transactions fail and when unusual activity may indicate fraud.

“Retailers should stop seeing payments only as a cost and start seeing them as a revenue generator.”

Bringing payment information from multiple branches and channels into one environment can provide the retailer with a much clearer view of business performance.

Payments providers can also see broader patterns across sectors and markets — without disclosing individual businesses’ confidential data.

During the audience discussion, Wandia Gichuru of Vivo Fashion Group raised an especially important question: how can a retailer understand its performance relative to the rest of its industry when competitors do not willingly share their results?

Judy and Eric explained that payments processors can produce aggregated market insights showing whether a category or sector is growing, contracting or changing across particular periods and regions.

That allows retailers to distinguish between a company-specific performance problem and a wider market movement.

For an executive making decisions about stock, locations, financing or expansion, that distinction is invaluable.

Using AI to Eliminate Inefficient Sales Activity

Sonal also shared how an internally developed AI sales tool helped improve the efficiency of prospecting and outreach.

According to her, the system had eliminated more than 4,000 hours of inefficient outreach and significantly increased the productivity of the sales process.

Instead of growing a large team to search manually for contacts, the organisation could use a smaller team to operate the platform while employees concentrated on higher-value work.

This was a good example of the difference between adopting AI because it is fashionable and applying it to a measurable operational bottleneck.

The value was not in being able to announce that the company used AI.

The value was in producing more relevant outreach, reducing wasted time and improving the return on sales activity.

Privacy Must Be Designed into the System

The use of customer data inevitably raises questions about privacy and trust.

Eric observed that innovation often moves faster than regulation, but Kenya now has a much clearer data-governance environment.

Organizations processing personal information must understand their responsibilities as data controllers or processors. Practices that were once treated casually — such as acquiring phone numbers from unofficial sources and sending unsolicited messages — now carry significant legal and reputational risk.

The solution is not to abandon personalization.

It is to build privacy, consent and governance into the platforms through which the personalization takes place.

As I put it during the panel, this is privacy by design.

Larger, properly governed platforms may also allow smaller retailers to access sophisticated targeting and analytical capabilities without having to construct their own complex data infrastructure.

The Next Customer May Be a Machine

One of the most forward-looking moments came from Judy.

Today, generative AI largely responds to prompts. The next phase will increasingly involve AI agents acting on behalf of consumers.

Those agents may compare prices, replenish groceries, pay subscriptions and select products according to rules established by their human owners.

“Your next-generation customer may not be a Gen Z or a Gen Alpha. It may be a machine — an agent shopping on behalf of the consumer.”

This has profound implications for retail.

An AI agent will not be persuaded by a colourful shopfront in the same way as a person. It will look for structured product information, availability, price, trusted signals, convenient fulfilment and compatible payment options.

A retailer whose website cannot be understood by AI systems, whose inventory is not visible digitally or whose payment process requires too much manual intervention may simply be skipped.

In that future, payments increasingly become invisible.

The ideal transaction resembles a completed ride-hailing journey: the customer receives the service, the authorized payment happens in the background and nobody has to stop to manage the checkout.

Retailers should therefore think beyond whether a payments fee is slightly higher or lower.

The more strategic question is whether the payment journey introduces friction that prevents either a human customer — or an AI agent — from completing the purchase.

Four Recommendations for the Next 12 Months

I closed the discussion by asking each panelist to recommend one action that retailers should take during the next 12 months.

Sonal: Apply AI to Sales Efficiency

Sonal recommended using AI to make sales outreach more precise.

Retailers should focus on reaching the right customer, through the right channel, with the right information, rather than wasting human effort on large volumes of poorly targeted activity.

Judy: Just Start

Judy’s recommendation was intentionally simple.

“Look internally for the quick and easy, low-hanging fruit. As small as it is, start today.”

That could mean providing employees with a safe AI environment, reducing repetitive tasks or finding a better use for information the organization already holds.

Eric: Digitize Your Operations

Eric argued that digitization is the ticket required to board the AI train.

A retailer cannot obtain meaningful operational intelligence from activities that remain invisible in cash transactions, paper records or disconnected processes.

Digitizing payments, suppliers, inventory and customer management creates the data foundation on which future insight can be built.

Siddesh: Focus on One or Two Problems

Siddesh warned against spending millions on elaborate data lakes and fashionable tools before demonstrating value.

“AI is no longer a luxury; it is a necessity. But take one or two things, focus on the small wins and move one step at a time.”

This may be the most realistic approach for many African retailers.

The objective is not to possess the largest AI budget.

It is to solve a commercially important problem and build from the result.

My Key Takeaways

1. More data is not automatically better. The first objective is to connect the information already being created across payments, inventory, stores, e-commerce and customer engagement.
2. Data readiness comes before AI readiness. Poorly structured or poorly labelled data will limit the value of even the most sophisticated platform.
3. Begin with the business problem, not the tool. Replenishment, pricing, churn, fraud, sales productivity or customer service provide clearer starting points than a vague ambition to “implement AI.”
4. Payments are an intelligence layer. Digitised payments can provide insight into customer behaviour, channel performance, risk and wider market trends.
5. Human judgement remains essential. AI can process more information and surface options faster, but management still owns the decision.
6. Privacy cannot be added later. Consent, data protection, security and responsible targeting must be part of the design from the beginning.
7. Retailers must prepare for agentic commerce. Future customers may increasingly delegate product discovery, comparison and purchasing to AI agents.
8. Small wins beat expensive experiments. A focused project with a measurable commercial outcome is more valuable than an ambitious programme that never leaves the presentation deck.

My sincere thanks to RETRAK for inviting me to moderate this timely conversation, and to Judy Waruiru, Sonal Haria, Eric Muriuki and Siddesh Narkar for bringing four highly complementary perspectives to the stage.

The conversation made one thing clear.

Kenyan and African retailers cannot afford to ignore AI.

But the winners will not necessarily be those that spend the most money or acquire the most tools.

They will be the retailers that digitzse the right processes, unify the right data, ask the right questions and act on the answers faster than their competitors.

Watch The Video

Listen To The Podcast