Corporate AI9 min read

How Can Credit Analysts Use AI?

On this page

Credit analysts can use AI to draft credit memo sections from figures they have already checked, explain trends in financial statements, build covenant and document checklists, and list questions for the borrower. AI should never approve, decline, score or recommend a facility. The analyst does the analysis and the approval authority decides. Customer data stays out of public AI tools.

This guide is for credit analysts, credit risk officers and credit administration staff at banks, microfinance banks and other lenders in Kenya. SACCO staff should also read our guide for SACCO loan officers. It is general information, not credit, regulatory or legal advice.

What Kenyan rules and regulators say

As of 15 November 2026, we found no final Central Bank of Kenya (CBK) rule for banks written only about AI. Four documents show where things are going.

The CBK's AI survey. The CBK surveyed the banking sector on AI in March 2025. The most common use of AI and machine learning reported by respondents was credit risk assessment, at 65 per cent. Ninety-three per cent of institutions recommended that the CBK issue guidance on AI, covering areas such as governance and compliance.

The draft Risk Management Guidelines, 2026. The CBK published this draft in September 2026 and invited public comments until 7 November 2026, together with other draft guidelines and guidance notes. It is clear about who does credit work. It says "each credit proposal should be subject to careful analysis by a qualified credit analyst". It calls for "a clear audit trail" identifying the people or committees giving input and making the credit decision. On AI systems, it expects "appropriate human oversight, model validation, bias testing, explainability measures". It is still marked as a draft, so check its status, and any final AI guidance, with your risk team.

The draft Guidance Note on Artificial Intelligence in the Banking Sector. The CBK released this draft in the same consultation, with comments due by 7 November 2026. The Standard reported that it would require banks to get CBK approval before using AI systems, and Deloitte's summary says customers should be able to ask for a person to review AI-driven decisions that affect them. It is a draft, not a final rule, so the final version may change.

The Data Protection Act, 2019. Credit files are full of personal data. Property details and family details are sensitive personal data under section 2. Section 25 requires data to be limited to what is necessary, and not transferred outside Kenya without proof of adequate safeguards or the person's consent. Section 35 gives people the right, with some exceptions, not to be subject to a decision based solely on automated processing that significantly affects them.

Your bank's own AI usage policy sits on top of all this. If it does not exist yet, ask compliance before you start.

Seven tasks where AI helps, with prompts

Each prompt uses the four Cs we teach at FuKazee: Clarity, Context, Constraints and Customisation. Replace customer names with "Company A", round or mask figures where your policy requires, and use only a tool your bank has approved.

1. Summarise financial statements you have already analysed

Do the numbers in Excel first. Then use AI to explain what they show.

Clarity: Write a one-page summary of the financial performance of Company A. Context: I am a credit analyst at a Kenyan bank. Below are three years of key figures and ratios I have calculated and checked. Constraints: use only these figures; do not calculate new ratios; do not give an opinion on whether to lend; mark any trend you cannot explain from the data as [ANALYST TO EXPLAIN]. Customisation: headings are revenue and margins, cash flow, leverage, liquidity, working capital. Plain, neutral language. FIGURES: [paste]

Check: every number in the summary matches your spreadsheet. Explanations of why a trend happened come from the borrower and the relationship manager, not from AI.

2. Build the ratio sheet faster

I have an Excel sheet with income statement and balance sheet lines in rows and three years in columns. Give me formulas for gross margin, EBITDA margin, current ratio, debt to equity, interest cover and debt service cover ratio, using our definitions below. Explain each formula in one line. OUR DEFINITIONS: [paste your credit policy definitions]

Check: test each formula on a past file where you know the answer. Use your bank's definitions, because they can differ from textbook ones. For more, see how to use Copilot in Excel.

3. Draft credit memo sections from checked inputs

Using only the masked information below, draft these sections of a credit memo in our template: borrower background, purpose of the facility, financial summary, security, key risks and mitigants. Write [MISSING] where information is not given. Do not write a recommendation or a conclusion. INFORMATION: [paste masked notes and checked figures]

Check: read the draft against the source documents line by line. The recommendation, risk grading and conclusion are yours, under your credit policy.

4. Turn a term sheet into a checklist

Below are the approved terms and conditions for a facility. Turn them into two checklists: conditions precedent before drawdown, and ongoing covenants with how often each is tested. Quote the clause number for each item. Add nothing that is not in the text. TERMS: [paste approved terms with names removed]

Check: credit administration compares the checklist with the signed offer letter before anything is released.

5. List questions for the borrower

Here is my masked summary of Company A's statements and the facility request. List the 10 most important questions a credit analyst should ask before writing the memo, grouped by cash flow, security, management and industry. Explain in one line why each matters.

Check: drop questions that do not fit the deal, and add your own. AI does not know the borrower or the market the way your relationship manager does. Our guide for bank relationship managers covers their side.

6. Write up a site visit

Turn my rough site visit notes into a short report: date, location (town only), what we saw, stock and premises condition, management discussion, concerns, follow-up actions. Neutral tone. Use only my notes and mark anything I did not record as [NOT RECORDED]. NOTES: [paste masked notes]

Check: the report reflects what you saw, and the photos and evidence stay in your bank's system.

7. Prepare sector research, then verify it

Give me a list of the main risks to check when lending to a [sector, for example a Kenyan horticulture exporter]. For each risk, say what document or data would help me test it. Do not give statistics.

Check: AI can invent figures and sources. Use the list as a starting point, then go to official publications, your bank's research and the borrower's records.

What AI must never do in credit

  • Decide. In our view, a general AI assistant should not approve, decline, score, grade or rank any application. The approval authority decides, with an audit trail.
  • See real customer files in a public tool. Names, ID and KRA PIN numbers, account numbers, bank statements, CRB reports, payslips and title documents stay in your bank's systems.
  • Check authenticity. AI cannot tell you whether a statement, title or valuation is genuine.
  • Supply numbers. Every figure in a memo comes from a document you have checked.
  • Write the recommendation. That is your judgement, and you sign for it.

For the bank-wide picture, read AI in Kenyan banking: what staff training needs to cover and our guide to the Data Protection Act and AI.

How FuKazee trains credit teams

FuKazee's AI training for banks lists first-draft credit memos, borrower and sector research summaries and structured review checklists among the tasks staff practise, "with the analyst's judgement kept firmly in the loop". Exercises use realistic, synthetic scenarios, such as an invented borrower, so no customer, account or transaction data leaves the bank for the training. Staff learn to sort what may go into a public tool, what may go only into an enterprise tool, and what must never leave core systems. We are tool-agnostic, and if your bank has licensed a platform, we build the training around it.

Questions people ask

Can AI write a credit memo?

It can draft sections from information you have checked, in your template. It should not write the recommendation or the risk grade. You read every line against the source documents before the memo goes anywhere.

Can I upload a customer's bank statement to ChatGPT?

Not into a public tool. A bank statement identifies the customer and shows their finances. If your bank has an approved enterprise tool, follow its policy on what may go in.

Does the CBK have rules on AI?

As of 15 November 2026, we found no final CBK guidance on AI for banks. The CBK's 2025 survey showed most institutions wanted guidance. In September 2026, the CBK released a draft Guidance Note on Artificial Intelligence in the Banking Sector and a draft of the Risk Management Guidelines with expectations for AI systems. Comments closed on 7 November 2026, and the final versions were not out when we checked. Check the CBK website and your compliance team for the current position.

Will AI replace credit analysts?

The draft CBK guidelines expect each credit proposal to be analysed by a qualified credit analyst, with a clear audit trail of who decided. AI can take some of the writing and formatting load. The judgement, and the accountability, stay with people.

Is it safe to use AI for financial ratios?

Use AI to write and explain formulas, then calculate in Excel yourself. Chatbots can make arithmetic errors and may use definitions that differ from your credit policy.

Sources

All sources last checked 15 November 2026.

Where FuKazee fits in

FuKazee is a Nairobi training company that has trained more than 3,000 people since 2022. We train bank and lending teams to use AI on real credit work, with synthetic borrowers and clear rules on who decides. Tell us how many people are in your credit team and which documents take most of their week. Email lead@fukazee.com

Want guided support instead?

Plan AI training around your team's real work. Answer a few quick questions about your sector, team size and goals, and we will recommend the right programme and format. Sessions are built on your workflows and include a usage policy.

Rodney MainaFounder & Director, FuKazee

Rodney has run FuKazee's training since 2022 and still teaches inside every programme line: corporate sessions, the postgraduate cohort and the KCSE examiner webinars. First Class Honours in Journalism and Mass Communication, University of Nairobi.