A bank approves a loan on one date. The borrower's risk keeps moving for the life of the loan.
Six months after approval, cash flow may have weakened, financing costs may have risen or a key contract may have gone. By the time arrears show it, the bank's options have already narrowed.
The task for Kenyan banks is to see that change early enough to act on it. That means moving from periodic reviews and reported arrears to a continuous view of the portfolio.
What Kenya's banking figures tell us
Kenya's banking sector ended 2025 with a lower NPL ratio. The Central Bank of Kenya's 2025 Bank Supervision Annual Report puts gross non-performing loans at 16.0% of gross loans in December 2025, down from 17.1% a year earlier.
The stock of bad loans barely moved. It stood at about KSh 696.9 billion, against KSh 697.3 billion a year earlier.
The ratio fell because the loan book grew around a stable stock of bad loans. That is an improvement. It is not evidence that every segment became less risky.
A better sector ratio tells a CRO little about their own book. Strong sector indicators and significant credit risk can sit side by side. The question is how risk is moving inside the institution's own portfolio, borrower groups and sectors.
Why arrears tell only part of the story
Missed payments are the most visible sign of borrower difficulty. They are also a late one.
A business under pressure often keeps paying by running down reserves, delaying suppliers or taking on short-term financing. A corporate borrower can meet its repayment schedule while liquidity, profitability or covenant headroom deteriorate. A retail borrower can keep up instalments for a while after income drops.
When arrears finally appear, the bank has fewer and more expensive options.
Effective monitoring therefore tracks financial, behavioural and operational indicators over time, not payment status alone. The aim is not to predict every default. It is to spot the situations that warrant attention while they are still manageable.
Different borrowers need different warning signals
No single indicator works for every borrower.
For an SME, warning signs include falling account activity, repeated requests for short-term extensions or sharp changes in working capital patterns. For a corporate borrower, they include covenant pressure, weakening liquidity or rising leverage. For a retail portfolio, they include patterns in missed payments, changing account activity and growing indebtedness.
Sector context matters too. A seasonal agricultural business should not be judged against the patterns of a salaried retail customer. Rising facility utilisation can be normal trading for one borrower and liquidity stress for another.
The right question is not whether an indicator changed. It is whether the change is unusual, material and relevant to repayment capacity. Answering it takes credit expertise as well as reliable data.
Fragmented data hides the pattern
Most banks already hold much of the information that signals early deterioration. Repayment history sits in core banking. Relationship managers hold customer knowledge. Financial statements sit elsewhere. Collateral records and limits belong to other teams.
Rising utilisation looks manageable on its own. Weaker trading looks temporary on its own. Seen together, they justify a closer review.
When the data is scattered, the pattern stays hidden. When follow-up runs on informal messages, nobody clearly owns the next step. Monitoring needs both a connected view and a defined process for action.
Early warning is a process, not an alert
An alert does not reduce credit risk. What the bank does next does.
Take a business customer whose working capital utilisation rises sharply while account inflows fall. The combination justifies a look. It does not prove distress.
The relationship manager asks the customer for an explanation. A credit officer reviews updated financials. The bank then maintains the facility, increases monitoring or reassesses the relationship, in line with its credit policy.
A sound early warning approach connects four steps: detection, investigation, decision and follow-up. Without all four, banks generate more alerts without understanding their portfolio any better.
Focus on the risks that matter most
Reviewing every account with equal intensity consumes credit teams. Relying only on fixed review cycles leaves fast-changing borrowers unattended between reviews.
A proportionate approach weighs signal severity, exposure size, borrower profile and the bank's risk appetite. Accounts with little change stay on standard monitoring. Accounts showing unusual or deteriorating patterns get enhanced review. Serious concerns go to specialist risk teams.
Prioritisation must stay transparent. Credit professionals need to see why an account was flagged and how to document the decision.
How AI can strengthen monitoring without replacing credit judgement
AI can find combinations of indicators across large portfolios that manual review would miss.
The CBK's 2025 Survey on Artificial Intelligence in the Banking Sector found credit risk assessment to be the leading use of AI among Kenyan lenders that have adopted it, with data quality and governance cited as the main risk.
Those concerns apply directly to credit monitoring. A model is only as reliable as its data, design and ongoing validation. An unusual transaction pattern can signal a business problem or a predictable seasonal event, and the model cannot always tell which.
Banks exploring AI should prioritise interpretability, human oversight and measured outcomes. They should also test whether new indicators lead to useful interventions or just add work for risk teams.
The goal is better understanding of credit risk, not the largest possible number of alerts.
Five questions for banking risk executives
- How early do we recognise deterioration? Do we usually spot important changes before repayment problems emerge, or only after customers fall into arrears?
- Can we see the complete credit relationship? Can monitoring teams see exposures, facility conditions, repayment behaviour and collateral in one place?
- Do our indicators reflect different portfolios? Or do we apply one monitoring approach to products and borrowers with very different risk profiles?
- What happens after a warning? Are responsibilities, investigation steps, escalation paths and follow-up documented?
- How do we judge effectiveness? Do we measure whether interventions got earlier and better, or just count alerts?
How ACP connects detection to action
Axe Credit Portal (ACP) treats early warning as the four-step process above, not as a dashboard.
- Detection on the bank's own terms. The bank's risk experts build and manage indicators, thresholds and scoring in ACP Early Warning System without vendor code changes. ACP draws data from core banking, credit bureaus, external media and watchlists, and CRM. It detects patterns across account turnover, payment behaviour, profit and loss trends, covenant compliance, collateral status and management changes.
- Prioritisation that shows its reasoning. ACP ranks signals by severity and priority using the bank's own indicators, so credit teams see why an account was flagged.
- Alerts that lead somewhere. Configurable alerts feed escalation workflows, recommended next-action plans and task management. Every alert has an owner and a documented outcome.
- Connected to the rest of the credit lifecycle. ACP Loan Servicing executes restructuring or deferrals from EWS triggers. ACP Collection & Provisioning picks up cases that need recovery. ACP Limit & Portfolio Management triggers limit reviews when sector-level deterioration appears.
- The bank's models, not a black box. ACP runs the bank's own or third-party predictive models inside the monitoring workflow. Credit judgement stays with the bank's credit teams.
From identifying problems to anticipating them
Portfolio quality will remain a central concern for Kenyan banks. A falling NPL ratio is welcome, but resilience depends on more than past performance.
Banks need to know how borrower conditions are changing, where vulnerabilities are concentrating and how fast their teams respond. That makes proactive monitoring an organisational priority as much as a technology one. It takes relevant data, differentiated indicators, clear ownership and consistent intervention.
The value of early warning is not knowing that a problem exists. It is recognising a meaningful change early enough to make a better decision.
Let's discuss credit risk monitoring at BFSI IT Summit Kenya 2026
· Glee Hotel, Amethyst Ballroom, Nairobi, Kenya
Meet Ferid Knani and Amine Tatouh to discuss:
- how early your current monitoring catches deterioration, and where it misses
- how to build warning indicators that fit SME, corporate and retail portfolios
- how ACP turns an alert into an owned, documented intervention
Ferid Knani
Business Development Director - Africa
Discuss your lending transformation plans and explore how ACP can support your institution's credit processes.
Book an onsite meeting with Ferid
Amine Tatouh
Business Development Consultant - Africa
Explore opportunities to improve lending journeys, configure workflows and connect your existing banking systems.
Book an onsite meeting with AmineIf the calendar does not load, open Ferid's booking page.
If the calendar does not load, open Amine's booking page.