One of the most important lessons from Kenya’s digital financial services journey is that access to credit is not only a money problem.

It is also an information problem.

A lender may have the capital and the willingness to serve a customer, but still be unable to make a confident decision because the customer’s financial information is incomplete, fragmented across different institutions or simply invisible to the formal financial system.

At the same time, the borrower may be forced through a slow and frustrating application process because the lender is operating with disconnected platforms, manual approvals, repeated document requests and outdated ways of assessing risk.

This is the tension at the centre of the future of lending in Kenya and Africa.

How can lenders make credit faster and more accessible without increasing fraud, irresponsible lending and over-indebtedness?

How can they use artificial intelligence and alternative data without compromising privacy, fairness and customer trust?

How can they digitise their operations without simply transferring the same inefficient processes from paper into software?

These were some of the questions we explored during The Future of Lending: Loan Origination, E-Sign & AI, a highly insightful event that I had the privilege of moderating on the 29th of May 2026 at Park Inn by Radisson in Westlands, Nairobi.

The event was co-hosted by Presta Technologies, Zoho and the Digital Financial Services Association of Kenya, bringing together lenders, financial technology providers, digital credit professionals and business leaders from across the lending ecosystem.

Presta Technologies brought the lending infrastructure perspective. Its platform is designed to connect activities such as customer onboarding, loan approval, disbursement, payment management, reporting, workflow automation and electronic guarantorship within a unified loan-management environment.

Zoho brought the wider enterprise technology and workflow automation perspective. The partnership was especially relevant because Presta has previously implemented Zoho Sign to replace time-consuming physical signatures with a more trackable and mobile-friendly electronic-signing process.

DFSAK brought the industry perspective, including collaboration between digital lenders, responsible market development, consumer protection and engagement around the future of digital lending in Kenya.

However, the most valuable part of the event was not the technology alone.

It was the conversation about the realities standing between the technology and the desired lending outcomes.

A Panel Reflecting The Entire Lending Journey

I was joined by a diverse and highly experienced group of panelists:

• Kris Senanu, Executive Chairman of Smith & Berkeley LLC, brought a business strategy and technology transformation perspective.
• Kevin Mutiso, CEO of OYE and Chairman of the Digital Financial Services Association of Kenya, brought the lender, industry and responsible financial inclusion perspective.
• Winnie Chira, Founder and CEO of Identify Africa, brought deep expertise in digital identity, KYC, KYB, anti-money-laundering processes and fraud prevention.
• Victor Kiplagat, CEO and Co-Founder of Spin Mobile, brought the alternative data, credit-scoring and decisioning perspective.
• Kenneth Mantu, Group CEO of The Adaptis Group, brought extensive experience in systems integration, analytics, enterprise transformation and AI-enabled business operations.

The discussion moved across the full lending journey: customer acquisition, onboarding, identity verification, credit scoring, loan approval, disbursement, collections, fraud management, reporting and the long-term sustainability of digital lenders.

Several major insights emerged.

The Biggest Lending Bottleneck May Be Fragmentation

Many financial institutions appear digital from the customer’s perspective.

They have a website.

They have a mobile application.

They may even have a chatbot or an online loan application form.

However, behind that digital interface, the actual lending process can still depend on spreadsheets, emails, paper documents, manual reviews and disconnected systems that do not communicate with one another.

Victor Kiplagat argued that this fragmentation remains one of the most significant constraints affecting African lenders.

“The majority of lenders are still operating in silos. You need a unified platform from loan origination to credit scoring, disbursement and collection.”

A customer may believe that they are completing one loan journey, but their information may actually be moving between several separate departments and platforms.

The application is captured in one system.

Identity is verified in another.

The credit score is generated elsewhere.

Approval may be sent by email.

Disbursement is processed through a separate payment platform.

Collections and reporting may sit in yet another system.

Every handover introduces friction.

It also creates the possibility of delays, inconsistencies, duplicated work, missing information and fraud.

The borrower does not care which internal department has failed to complete its part of the process.

They simply experience a slow, confusing or unreliable service.

This is why a mobile application alone does not make a lender digital.

The full operating model has to be digital.

2. Data Integrity Matters More Than Having More Data

One of Kenneth Mantu’s most important observations was that having multiple data integrations does not automatically result in better decisions.

The real issue is whether lenders can trust the data moving through those integrations.

“It is one thing to have a good set of integrations. It is another thing to be able to rely on that data to make the right business decisions.”

Poorly mapped, incomplete, outdated or duplicated data can make even the most sophisticated analytics platform unreliable.

This becomes especially dangerous in lending because organisations are continuously making decisions about:

• Who should receive credit
• How much they should receive
• How long they should have to repay
• What price reflects their level of risk
• Whether an application may be fraudulent
• How likely the customer is to default

Kris Senanu reinforced the same point from a strategic business perspective.

“Long-term sustainable businesses are not built on gut feel. They are built on data, information and trends.”

Experience and instinct remain valuable, especially among people who have worked in an industry for many years.

However, instinct should be tested against evidence.

Analytics can reveal that the customer segment management assumed was most profitable is actually underperforming.

It can reveal that a seemingly risky customer group is repaying more reliably than expected.

It can expose patterns that are invisible when individual applications are assessed manually.

The goal is not to eliminate human judgement.

It is to give human judgement better information.

Fragmented Data Creates Both Exclusion & Over-Indebtedness

Kevin Mutiso introduced an important paradox.

When lenders cannot see a complete view of the customer, two problems can happen at the same time.

The first is exclusion.

A viable borrower may be rejected because the lender cannot see enough evidence of their ability and willingness to repay.

The second is over-indebtedness.

Several lenders may independently assess the same borrower as capable of repaying a loan. Each lender then extends credit without seeing the borrower’s total exposure across the market.

The customer may be able to afford one KES 50,000 loan.

They may not be able to afford five separate KES 50,000 loans issued by five institutions that each assessed them in isolation.

This means data sharing and aggregation are not only customer-acquisition issues.

They are also responsible-lending issues.

Kevin explained that lenders often struggle to identify their next viable customers because useful financial information remains distributed across different organisations and systems.

“There are customers who need the services we offer, but we cannot see them.”

That statement captures one of the biggest opportunities in African lending.

Better information can help institutions lend more.

But it can also help them know when not to lend.

Identity Has Become A Front-Line Lending Risk

Digital lending has reduced the need for customers to visit branches, but it has also created new opportunities for identity fraud.

Winnie Chira highlighted the growing challenge posed by stolen identity documents, manipulated financial records, fake documentation and deepfake-enabled fraud.

“We are in the age of deepfakes, stolen IDs and fake documents.”

The answer is not necessarily to make every customer complete an increasingly long list of security checks.

That approach may create more friction for legitimate customers without stopping sophisticated fraudsters.

Instead, Winnie recommended a risk-based approach.

A lender should build enough information about an applicant to understand how confident it is about that person’s identity.

A low-risk application may proceed through a relatively seamless onboarding journey.

An application displaying unusual or suspicious characteristics can trigger an additional layer of verification.

“Make it easy for the genuine customer not to see all the security checks, while making it difficult for the fraudster to get through.”

This is an important distinction.

Good digital identity should be almost invisible to the genuine customer.

It should become highly visible only when the system detects that something may be wrong.

Victor highlighted practical examples including identity-document matching, phone-number verification, document analysis and checks for recent SIM changes before loan disbursement.

These controls can help identify situations where someone has acquired another person’s documents or mobile access and is attempting to borrow fraudulently.

The goal is not simply more KYC.

It is smarter KYC.

5. Alternative Data Can Make The Financially Invisible More Visible

Traditional credit decisioning normally relies on information such as income, prior borrowing, repayment history and credit bureau records.

However, millions of people across Africa may have limited formal financial histories even though they regularly earn, spend, save, trade and repay informal obligations.

These are often described as thin-file customers.

Victor explained that alternative data can help lenders understand these customers more accurately.

This could include information from:

• Mobile-money statements
• Bank transactions
• Regular payments
• Existing loan commitments
• Income and expenditure patterns
• Location and behavioural data
• The purpose for which credit is being requested

Spin Mobile’s approach combines traditional credit information with alternative data to create a broader view of the borrower.

Victor shared several memorable correlations observed in specific datasets, including links between repayment behaviour and patterns such as regular religious contributions, loan stacking, high betting expenditure and unusually high medical expenses.

These observations generated some of the liveliest moments in the conversation.

However, they also require careful interpretation.

A behavioural signal is not automatically a universal truth.

Paying tithe does not, by itself, make someone creditworthy.

Spending money in a particular category does not, by itself, prove that someone will default.

A correlation found within one dataset should not be applied indiscriminately across every borrower, community or market.

These signals must be combined with affordability, income, existing obligations, identity, the purpose of the loan and other relevant information.

They must also be governed responsibly.

Customers should understand what information is being used, why it is being used and how it affects the lending decision.

Lenders must continually test their models for bias, unfair exclusion and unintended outcomes.

AI may identify a predictive pattern.

Governance determines whether it is appropriate to use that pattern.

Alternative Data Is Useful Beyond Credit Scoring

One of the more interesting insights was that alternative data does not only help a lender decide whether to approve or reject an application.

It can also help with:

• Product development
• Customer segmentation
• Collections
• Fraud detection
• Branch and market expansion
• Identifying emerging customer needs

For instance, regular school-fee payments could indicate demand for an education-financing product.

Regular purchases from agricultural suppliers could point towards an input-financing opportunity.

Location and transaction patterns can show where clusters of customers live or conduct business.

The same data used to assess risk can therefore help a lender design more relevant products.

This changes the conversation from:

“Can we lend to this customer?”

To:

“What form of credit would be most useful and appropriate for this customer?”

That is a far more customer-centred question.

7. The Purpose Of The Loan Can Be A Powerful Signal

During the audience discussion, a representative from a solar asset-financing company asked how lenders should assess customers who use feature phones and have extremely limited transaction histories.

Victor’s response highlighted the importance of starting with the information that is available, offering a manageable initial exposure and using repayment behaviour to gradually build a customer history.

He also emphasised the importance of loan purpose.

“One of the strongest drivers of repayment was the purpose of the loan.”

This helps explain the potential of embedded finance.

A farmer who receives agricultural inputs is not simply receiving unrestricted cash.

A household acquiring a solar system is receiving a productive or quality-of-life asset.

A student receiving education financing has a clear and defined use for the facility.

The lender can connect the credit directly to the intended purpose.

This does not eliminate risk, but it can provide more context than an unrestricted cash loan.

It can also make the lender part of a wider commercial journey rather than a standalone source of money.

Convenience Can Be As Important As Price

Victor also observed that customers do not evaluate credit based on price alone.

Speed, certainty and convenience can be equally important.

A customer may prefer a facility that costs slightly more if it can be approved and delivered immediately, especially when the financing is connected to an urgent business or household need.

This does not mean lenders should disregard affordability or responsible pricing.

It means the value proposition of credit has several dimensions:

• How much it costs
• How quickly it can be obtained
• How easy the process is
• Whether the customer knows what is happening
• Whether the repayment structure matches their cash flow
• Whether the credit solves the intended problem

A theoretically cheaper loan that arrives after the opportunity has disappeared may be less useful than a slightly more expensive facility that arrives at the right time.

The Shopkeeper May Be Kenya’s and Africa’s Hidden Credit Bureau

Kevin shared what was arguably the most thought-provoking statement of the entire panel:

“The largest lender in this market is actually the shopkeeper.”

Across Kenya and many other African markets, shopkeepers extend informal credit every day.

They know which customers settle their accounts reliably.

They know who needs extra time.

They understand the income cycles of households in their communities.

They may know that when a teacher’s salary is delayed, the effects are felt across the local shop, supermarket, bar, church and wider economy.

The shopkeeper has built what Kevin described as an instinctive credit-appraisal system.

It exists in the shopkeeper’s memory.

It may also exist in a notebook behind the counter.

But it rarely exists in a structured digital form that could help the customer build a formal credit profile.

This creates an enormous innovation question:

How can Kenya and Africa responsibly digitise informal credit relationships without destroying the context and trust that make them work?

The answer cannot be to treat every informal opinion as objective data.

A shopkeeper’s judgement can contain personal bias, incomplete information or community assumptions.

However, when combined with transparent transaction records, customer consent and other reliable indicators, this type of information could help make thin-file borrowers more visible.

Kenya’s and Africa’s next major credit dataset may not come from a bank.

It may come from millions of everyday commercial relationships that have never been formally documented.

Speed Is The Clearest Test Of Whether A Lender Is Truly Digital

Kris was direct about what an outdated lending operation looks like.

“If there is paperwork, you are not digital.”

A customer should not wait several days for approval and then several more days for disbursement because the application is being passed manually between departments.

A lender should be able to see, in real time:

• Where an application is
• What information has been received
• Which checks have been completed
• What risks have been identified
• Who needs to act next
• Whether the loan has been approved
• Whether the funds have been disbursed
• How the portfolio is performing

Real-time visibility is not only an operational benefit.

It is a leadership requirement.

Management cannot respond quickly to risk, fraud, arrears or changing customer behaviour when it is working with reports that are several days or weeks old.

This is where I shared one of the AI principles I have used for several years:

Do not give a human a robot’s job.

People should not spend valuable time repeatedly transferring information between systems, calculating predictable figures or pursuing routine approvals.

Technology should handle repetitive, rules-based and high-volume processes.

People should focus on judgement, relationships, strategy, problem-solving and exceptional cases.

Automation is not primarily about removing people.

It is about using people better.

Digital Transformation Is Still A People Challenge

Kenneth offered a useful warning for organizations attempting to digitize everything at once.

“The only way to eat an elephant is piece by piece.”

A lender may need dozens of new capabilities, but its employees may only be able to absorb a few significant changes at a time.

Buying every available system simultaneously does not guarantee transformation.

In fact, it may create resistance, confusion and poor adoption.

Kenneth recommended a phased approach:

1. Identify the most urgent business problem.
2. Find internal champions.
3. Implement manageable changes.
4. Demonstrate visible value.
5. Train the people who will use the new processes.
6. Then expand the transformation progressively.

“The people are the drivers of the innovation. They are the drivers of the change and the solution.”

Many technology projects do not fail because the software is technically incapable.

They fail because workflows were not redesigned, employees were not prepared, incentives were misaligned or leadership did not communicate why the change mattered.

A successful digital lender therefore needs more than a technology roadmap.

It needs a change-management roadmap.

AI Is Not A Substitute For A Functional Lending Operation

AI was naturally a major part of the discussion.

Kevin encouraged lenders to experiment with connecting AI tools to existing lending platforms, accounting systems, customer-management applications and payment rails.

This can reduce the time required to build new capabilities and make it easier to automate processes that previously required extensive development.

However, simply adding AI to a fragmented operation will not fix the underlying fragmentation.

AI cannot compensate indefinitely for:

• Poor data
• Inconsistent workflows
• Weak identity controls
• Unclear lending policies
• Disconnected systems
• Employees who have not been trained
• A business case that has not been properly defined

The sensible question is not:

“Where can we use AI?”

It is:

“Which lending problem are we trying to solve, and is AI the most appropriate tool?”

That shift can help lenders move beyond impressive demonstrations and towards measurable business value.

AI Requires Anomaly Detection And Human Oversight

Kenneth ended the discussion with one of its most technically important insights.

“Most people look at AI as a solution in itself. You still need periodic checks to ensure it is not going off on a tangent.”

AI models do not operate in a permanent and unchanging environment.

Economic conditions change.

Customer behaviour changes.

Fraud techniques evolve.

New data enters the system.

The performance of a model can therefore weaken over time.

Anomaly detection helps lenders identify when outcomes or behaviour begin to move outside expected patterns.

This could include:

• An unusual rise in approvals
• Unexpected rejection patterns affecting particular groups
• A sudden increase in fraud
• Changes in default behaviour
• Data entering the system in an unexpected format
• A model giving excessive weight to one variable
• Scoring outcomes that no longer match portfolio performance

AI should not be treated as an unquestionable decision-maker.

It should be treated as a powerful system that is continuously monitored, tested and improved.

The future of lending is not human judgement versus machine intelligence.

It is machine intelligence strengthened by responsible human oversight.

What Should A Lender Prioritize Over The Next 12 Months?

I ended the panel by asking each participant to name the single change a lender should make to become more competitive and future-ready.

Their responses were revealing.

Victor recommended moving away from siloed platforms and towards a unified lending environment.

Winnie called for intelligent automation driven by data.

Kris emphasized complete digitization and real-time visibility for management.

Kevin encouraged lenders to begin practically experimenting with AI integrations.

Kenneth emphasized anomaly detection and the need to continuously monitor AI-driven systems.

Taken together, their recommendations form a practical transformation agenda.

A future-ready lender should:

1. Map the entire lending journey, from acquisition and onboarding through repayment and collections.
2. Identify manual and disconnected processes that create the most significant customer or operational friction.
3. Establish reliable data foundations before attempting to scale advanced analytics or AI.
4. Integrate identity, fraud, scoring, disbursement and loan-management systems into a coherent workflow.
5. Use alternative data carefully to understand thin-file customers without creating unfair or opaque decisions.
6. Apply automation to repetitive work while preserving human judgement for exceptions and high-impact decisions.
7. Give management real-time visibility into applications, portfolio performance, risk and arrears.
8. Introduce AI around clearly defined business problems, rather than adopting it simply because it is fashionable.
9. Monitor models continuously for anomalies, bias, drift and unexpected consequences.
10. Invest in people and change management as seriously as the organization invests in software.

My Final Takeaway

As I brought the discussion to a close, I summaried what I had heard into three foundational elements:

1. Data.
2. Technology.
3. People.

On reflection, I would add a fourth:

Governance.

Data makes better decisions possible.

Technology makes those decisions faster and more scalable.

People provide judgement, context, relationships and leadership.

Governance ensures that the entire system remains responsible, transparent and aligned with the customer’s interests.

None of these elements can deliver the future of lending alone.

The most successful Kenyan and/or African lender will not necessarily be the institution with the most AI tools.

It will be the one that knows which customers it should serve, understands them responsibly, makes decisions efficiently and delivers a lending experience that is fast, transparent, secure and genuinely useful.

My sincere thanks to Presta Technologies, Zoho and DFSAK for co-hosting this timely event.

I am equally grateful to Kris Senanu, Kevin Mutiso, Winnie Chira, Victor Kiplagat and Kenneth Mantu for sharing their knowledge so generously.

There was far more insight in this conversation than we could possibly exhaust in one panel.

Fortunately, the full discussion is now available to watch and listen to as below: 

Watch The Panel Discussion

Listen To The Audio Podcast