Across Kenya and East Africa, boards are asking a question they could not have asked five years ago: which AI systems are we using, and what is the risk of each? An AI risk assessment in Kenya is the structured process that answers that question. It moves the conversation from anecdote and vendor pitches to a clear, prioritised picture of the organisation’s AI footprint and the risks it carries.
The exercise matters because AI is no longer confined to a data science team. It is embedded in vendor SaaS platforms, ERP modules, productivity tools, fraud monitoring systems, customer service chatbots, and ad-hoc experiments across business units. Banks, fintechs, SACCOs, insurers, manufacturers, NGOs, and public sector entities all face the same reality: AI use has expanded faster than oversight.
This article sets out how to conduct an AI risk assessment in Kenya, drawing on the Data Protection Act, 2019, the Kenya Standard KS 3007:2025, the NIST AI Risk Management Framework, and ISO/IEC 42001. It covers the model inventory, the shadow AI problem, risk tiering, and how to integrate the assessment into enterprise risk management.
Why an AI Risk Assessment in Kenya Starts with Visibility
You cannot govern what you cannot see. The first reason most organisations commission an AI risk assessment in Kenya is to establish visibility into where AI is being used, by whom, for what purpose, and against what data. That visibility, often surprising even to senior leadership, is itself the deliverable that triggers the rest of the AI governance investment.
An AI risk assessment in Kenya also creates the basis for risk-adjusted prioritisation. Not every AI use case warrants the same attention. A marketing team using a content tool with public information sits at a different risk level from a credit scoring model approving loans, a claims algorithm denying insurance payouts, or a hiring tool screening candidates. Without an assessment, organisations either over-invest in low-risk use cases or under-invest in high-risk ones.
The Shadow AI Problem in Kenyan Organisations
Shadow AI – AI systems in use without formal governance approval – is widespread across Kenyan organisations. Three patterns recur. The first is staff using public AI tools (generative AI assistants, document analysers, code helpers) for work purposes, often with confidential client or member data. The second is business units procuring SaaS tools with embedded AI features that were not flagged during procurement review. The third is vendor and partner systems that include AI components in their backend, sometimes undisclosed.
An AI risk assessment in Kenya must surface all three patterns. The discovery process typically involves staff surveys, procurement record reviews, vendor questionnaires, and direct inspection of widely-used platforms. The output is an inventory that includes not just internally built models but also vendor-embedded AI – which often represents the larger share of the actual footprint in banking, fintech, SACCOs, insurance, manufacturing, NGOs, and the public sector.
Building the AI Model Inventory
The model inventory is the central deliverable of an AI risk assessment in Kenya. For each AI system, the inventory captures the use case, the sponsoring business unit, the data inputs, the decision outputs, the affected stakeholders, the vendor or internal team, the deployment environment, and the maturity stage. The inventory becomes a living document that anchors the governance program over time.
Internally Built Models
Internally built AI systems are typically the easiest to inventory because they involve identifiable teams, named models, and dedicated infrastructure. The information needed includes the model purpose, training data sources, validation methods, deployment platform, monitoring routines, and the model owner. Even where the organisation has only a handful of internally built models, the documentation effort often reveals gaps in version control, lineage, and accountability that the assessment must capture.
Vendor Models
Vendor AI is harder to inventory because the organisation does not control the model itself. The inventory therefore focuses on the vendor relationship: which AI capability is being procured, what data is sent to the vendor, what the vendor’s controls look like, whether the vendor is willing to share model documentation, and whether the contract permits independent audit. An AI risk assessment in Kenya treats vendor AI with the same rigour as internally built AI, because the consequences of failure are the same regardless of who built the model.
AI Embedded in SaaS and ERP Platforms
The largest blind spot for most organisations is AI embedded in mainstream SaaS and ERP platforms – email systems, productivity suites, customer relationship management, accounting and finance platforms, HR software, and security tools. These embedded features are often enabled by default and rolled out through routine vendor updates. The AI risk assessment in Kenya must include a systematic review of feature flags, vendor release notes, and platform configurations to map this embedded layer.
Risk Tiering: From Minimal to Unacceptable
Once the inventory is built, the AI risk assessment in Kenya tiers each use case by risk. A widely adopted approach uses four tiers: minimal, limited, high, and unacceptable risk. Minimal-risk use cases (internal productivity tools, generic content generation) proceed with light oversight. Limited-risk cases (customer-facing chatbots, internal analytics) require basic controls and disclosure. High-risk cases (credit decisions, claims processing, hiring, fraud detection, health and safety decisions) require dedicated controls and Ethics Committee approval. Unacceptable risk cases are not deployed.
Risk Factors: Harm, Reversibility, Scale, Sensitivity
The tiering itself draws on four factors. Harm asks what happens if the AI gets it wrong – financial loss, denied service, reputational damage, physical safety impact. Reversibility asks whether the consequences can be undone. Scale asks how many people are affected. Sensitivity asks whether the system handles personal data, health information, financial data, or other regulated categories. An AI risk assessment in Kenya scores each use case across these four dimensions to derive its tier.
Aligning Tiers to KEBS KS 3007:2025 and International Frameworks
The four-tier model maps cleanly to international references. KEBS KS 3007:2025 sets out a Kenyan framework for AI applications that addresses risk, governance, and ethics. The NIST AI Risk Management Framework anchors a structured approach to risk identification and measurement. ISO/IEC 42001 provides the management system structure. The EU AI Act, which applies extraterritorially to Kenyan organisations selling into the European Union, also uses a tiered model. Aligning the AI risk assessment in Kenya to these references gives the framework credibility with regulators, customers, donors, and counterparties internationally.
From Inventory to Action: Building the AI Risk Register
The AI risk register is where the assessment becomes operational. For each tiered AI use case, the register captures the inherent risk (before controls), the controls in place, the residual risk (after controls), the risk owner, the next review date, and any open remediation actions. The register feeds management reporting, board oversight, and internal audit planning. It is also the source from which the Ethics Committee selects cases for deeper review.
A well-built AI risk register supports the FNJ Insight, Judgement, Execution risk model: insight from the inventory, judgement from the tiering, and execution from the remediation pipeline. Without all three, the assessment is a static document rather than a living risk management tool.
Integrating an AI Risk Assessment in Kenya into Enterprise Risk Management
An AI risk assessment in Kenya should not sit in a parallel structure to enterprise risk management. It should integrate with the organisation’s existing ERM framework, risk taxonomy, and risk appetite statements. In banks, that means linking AI risk to operational risk, ICT risk, model risk, and conduct risk. In SACCOs and microfinance institutions, it links to operational and member protection risk. In insurance, it links to underwriting and claims risk. In NGOs and public benefit organisations, it links to programme delivery and beneficiary protection. In manufacturing and the public sector, it links to operational and reputational risk.
Integration matters because the board sees one risk picture, not multiple parallel ones. The AI risk assessment in Kenya feeds into the enterprise risk register, the risk appetite review, and the audit plan, ensuring that AI risk is treated alongside the risks the organisation already understands.
How FNJ & Associates Conducts an AI Risk Assessment in Kenya
FNJ & Associates conducts AI risk assessments for organisations across banking, fintech, SACCOs, insurance, manufacturing, NGOs, and the public sector in Kenya and East Africa. Our methodology covers use-case discovery (including shadow AI), model inventory build, vendor AI mapping, four-tier risk classification, AI risk register design, and integration with the client’s enterprise risk management framework.
Our risk management services in Kenya cover ERM frameworks, risk assessments, risk automation on Trigarc Risk and training.
Each engagement is anchored to KEBS KS 3007:2025, the NIST AI Risk Management Framework, ISO/IEC 42001, and the Data Protection Act, 2019. We deliver the assessment alongside a remediation roadmap and, where the client chooses, operationalisation through Trigarc Risk, implemented by FNJ, allowing the AI risk register, controls, and monitoring to run on a single platform alongside the organisation’s broader risk management activity.
Frequently Asked Questions
How long does an AI risk assessment in Kenya typically take?
For mid-sized organisations, an initial AI risk assessment takes four to eight weeks, depending on the size of the AI footprint and the maturity of existing risk management. Larger organisations with extensive vendor ecosystems may take longer, particularly during the shadow AI discovery phase. The deliverables include the AI inventory, the tiered risk register, and a remediation roadmap.
Who should be involved in an AI risk assessment?
A successful assessment involves risk and compliance, IT and data teams, procurement, business unit leads, legal, and internal audit. The cross-functional involvement is essential because AI sits across all of these areas. Executive sponsorship from the CEO, COO, or CRO ensures that business units engage fully with the discovery process.
What is the difference between an AI risk assessment and a Data Protection Impact Assessment?
A Data Protection Impact Assessment under the Data Protection Act, 2019 focuses on personal data processing risks for specific high-risk processing activities. An AI risk assessment is broader, covering all AI-related risks including bias, performance, security, third-party, and ethical risks across the full AI portfolio. The two complement each other: DPIAs are typically commissioned for specific high-risk AI use cases identified through the assessment.
Should an AI risk assessment in Kenya include vendor AI?
Yes. For most organisations, vendor and embedded AI represents the larger share of the actual AI footprint. Excluding vendor AI would leave the assessment incomplete and the residual risk picture misleading. Vendor questionnaires, contract review, and platform configuration audits are standard parts of a comprehensive assessment.
Can FNJ & Associates conduct our AI risk assessment?
Yes. FNJ & Associates conducts AI risk assessments across the financial services, manufacturing, NGO, and public sectors in Kenya and East Africa. Our methodology is grounded in KEBS KS 3007:2025, the NIST AI Risk Management Framework, and ISO/IEC 42001. We can also operationalise the resulting AI risk register through Trigarc Risk. Contact us via fnjassociates.co.ke to schedule a consultation.
| Conduct an AI Risk Assessment in Kenya FNJ & Associates helps organisations across banking, fintech, SACCOs, insurance, manufacturing, NGOs, and the public sector conduct comprehensive AI risk assessments. From shadow AI discovery to risk tiering to enterprise risk integration, we provide the structured visibility boards and management teams need to govern AI confidently. Visit fnjassociates.co.ke to schedule a consultation with our AI risk advisory team. |
| About FNJ & Associates FNJ & Associates is a professional services firm offering audit and assurance, tax advisory, compliance, forensic audit, ERP implementation, and corporate training services across Kenya and East Africa. FNJ implements Trigarc Risk, a platform that helps organisations automate risk identification, measurement, and reporting. Visit us at fnjassociates.co.ke to learn more. |

