Every credit team has a spreadsheet with a name like FINAL_v7_REAL. It started as a helpful tool: a quick way to price loans, check thresholds and remember the rules in a four-hundred-page credit manual. Over the years it grew tabs, macros and personal notes. It became the thing credit officers trusted most, and the thing nobody else could see.
For SME lending, that is a real risk. Pricing, guarantor requirements and exception rules change through circulars and manual revisions. When a personal spreadsheet keeps an old rule alive, a loan can be priced or structured on terms the bank no longer offers.
Read the story behind this: The Laminated Policy Part 3
The problem
Credit manuals are long, detailed and frequently revised. Credit officers work under time pressure, often with a relationship manager and a waiting customer on the other side of the desk. The official manual is accurate but slow to search, so officers build their own shortcuts.
Those shortcuts are rarely wrong when they are made. The problem is that nobody updates them when the manual changes, and nobody outside the officer's own drive can check them.
Why it happens
- Manuals are long. A key rule can sit in a sub-clause under a heading nobody would think to search for.
- Changes arrive in pieces. A pricing circular here, a manual revision there, each correct but easy to miss.
- Personal tools have no review. A private spreadsheet is never approved, so it is never retired.
- Pressure to move fast. Good customers and urgent deadlines push teams towards the quick answer.
- Sensitive records need limits. Credit committee papers contain other customers' details, so they cannot simply be shared to help a busy team.
What good looks like
A lending team that decides on current policy usually has:
- One approved body of content: the credit manual, financing guidelines, pricing circulars and committee procedures.
- A fast way to ask a precise question and find the exact clause that answers it.
- A citation on every answer, so the officer can read the clause before relying on it.
- Access by role, so committee papers stay with committee members and credit risk.
- A clear exception path, so a good loan that does not fit the standard rules goes to committee with its reasons recorded.
- A record of how policy content is used, for supervision and audit.
How enterprise knowledge AI helps
An enterprise knowledge AI platform such as iKB, by Astiostech, turns approved organisational content into governed answers with citations. For an SME lending team, that can mean:
- Finding the right clause in a long manual. Hybrid search, structural context, reranking and knowledge relationships help surface the specific sub-clause that answers a question, even when it sits deep in a long document.
- Citations on every answer. Officers see exactly which section of the manual or which circular an answer came from, and can open it to check.
- One search across many sources. iKB connects documents, storage, databases, websites and APIs, so the manual, guidelines and circulars can be searched together.
- Role-based access control. Restricted material such as committee papers stays visible only to the roles that are permitted to see it.
- Governance and records. Audit trails, activity logging, redaction and retention rules help keep credit content and its use under control.
None of this replaces credit judgement. The officer still reads the clause, weighs the customer and decides whether a case belongs in front of committee. The platform makes sure that judgement starts from the current rules.
What to check before buying
- Precision in long documents. Test it with a rule you know is buried deep in your credit manual. Does it find the exact sub-clause?
- Citations. Every answer should link to the section it relied on.
- The old-spreadsheet test. Ask about a threshold that changed last year. The answer should cite the current rule.
- Access rules. Confirm that committee papers and customer records stay restricted by role.
- Deployment and data handling. Confirm where the platform runs and where credit documents and questions are stored.
- Content ownership. Agree who owns the credit manual in the platform and who retires superseded circulars.
Next steps
Start with the questions your credit officers check most often: pricing bands, guarantor and collateral requirements, and exception procedures. Connect the current approved versions, give one lending team access, and compare their working notes against the cited rules for a few weeks. Then ask everyone to rename their personal pricing spreadsheet to something honest. FINAL_v8_DO_NOT_USE works well.
To see how iKB works with credit manuals and lending guidelines, book a briefing on governed knowledge AI for your organisation with the Astiostech team.