By: Kendall Stevens
Ask borrowers what they would like from a lender, and most will tell you they want a clear decision, delivered quickly.
In standardised consumer credit, technology has largely met that expectation, with decision engines now able to assess products such as credit cards, personal loans, or buy now pay later (BNPL) arrangements within seconds. They can do so because the criteria are clearly defined and individual exposures are relatively small. The questions a lender needs to answer are broadly consistent from one application to the next, so a rules-based engine, particularly when applied across a large population of similar borrowers, can be quite effective.
A lender processing thousands of consumer loan applications each day can rely on predefined rules precisely because isolated losses can be managed across a diversified book. As a result, borrowers have become used to receiving rapid credit decisions, making speed a key expectation of competent service in standardised lending.
Complex lending requires more than predefined rules
The same approach is difficult to apply to large corporate facilities, trade finance, and structured lending with strong single-name exposure. This is the biggest reason why automation adoption is relatively limited in the space, with many institutions still relying heavily on manual processes. Here, each transaction brings its own combination of considerations, and the uniformity that makes consumer scoring possible simply is not there. Rather than applying a fixed score, lenders need to assess financial performance, collateral, covenants, group structure, and sector risk together before reaching a decision. No small task.
The stakes are different, too. A single consumer default is usually small enough that it can be absorbed quite easily. Suffering a loss on a large corporate exposure, however, is a different story entirely and can easily erase the margin earned on any number of other performing loans. Where consumer lending is able to rely on the law of large numbers to contain losses, complex lending tends to be far more concentrated, which raises the cost of a poor credit decision.
Consumer scoring also draws on standardised, machine-readable inputs. A corporate assessment, by contrast, often rests on many different factors. Consider, for example, financing a multinational manufacturer. The assessment may involve multiple subsidiaries, different forms of collateral, existing banking facilities, and cross-border exposure. These factors cannot sensibly be judged one at a time. They must be considered together, and doing so requires the kind of professional judgement that a single automated decision simply does not have. This is why such judgement is and will remain central to complex lending, even as automation transforms other areas of credit analysis.
Automating preparation as opposed to judgement
While decision automation is well suited to standardised products, complex corporate lending benefits from a different type of automation – the kind aimed at reducing the time spent in preparation. Automating these top-of-funnel activities means that a credit team will have additional capacity to manage more complex facilities without thinning the analysis and professional judgement applied to each. This is the thinking behind tools such as KS-TF’s Financial Management Tool (FMT), which support the preparation stage by automating repetitive analytical tasks while keeping the final decision with the credit team.
Spreading accounts is one of the most time-consuming steps in credit analysis. The analyst must sift through financial statements to identify the relevant line items, then map those line items into a consistent structure, where they then must input historical periods into standardised spreadsheets, before they are finally able to ensure that the resulting figures can support ratios, trends, and reporting outputs. It’s essential work. But it’s also repetitive and absorbs time that would be better spent on credit assessment. It is also subject to error, and it often involves multiple spreadsheets and manual checks. Once the numbers are spread, they still need to be transferred into reports, ratio tables, and charts, creating a second layer of manual work. The same figures may appear in the spread, the ratios table, the written analysis, the covenant section, the rating input and the final credit paper. Every manual transfer means more time spent by that poor, now-beleaguered analyst, not to mention a marginally higher risk of inconsistency across it all.
Automatic spreading addresses the first part of the problem by helping to convert financial statements into structured, analysis-ready data. Instead of manually typing each line item into a spreadsheet, the system is able to help extract, classify, and map into a standard structure. Once structured, the data can be used to calculate metrics such as profitability, leverage, liquidity, coverage, cash conversion, working capital, and growth. With this data in hand, it suddenly becomes possible to analyse trends over multiple periods, which means it becomes possible to easily and regularly update rating inputs.
All of that is really just a jargon-filled way of saying that this technique makes it possible for analysts to start asking the right credit questions sooner. Is revenue growth supported by cash generation? Is leverage sustainable? Is working capital absorbing liquidity? Are covenant thresholds becoming tighter?
The next step is automatic reporting. In many credit workflows today, reporting is highly manual. Those same analysts sift through even more documents to copy figures into tables, update charts, paste ratio outputs into memos, and manually ensure that the model and the report are telling the same story. This process, too, is rather time-consuming and prone to human error. Automatic reporting, however, can help by connecting report content to the underlying financial data, which means that structured data can feed standard report sections, ratio tables, trend analysis, and charts automatically. The credit report becomes less of a manually assembled document and more of a consistently generated and structured output. Analysts rejoice!
What automation returns to the credit analyst
Having developed a taste for automation in the preparation process, many teams may then be tempted to automate the final decision as well. In complex lending, this would be a mistake. Reading how the different parts of a case fit together and interpreting what they all mean calls for explainability and human accountability, which a model cannot provide. When automation is applied thoughtfully, and only to the areas that warrant it, the analyst’s time is rightly returned to this interpretation.
Liberated from their manual efforts, analysts will have more capacity to gain a deeper understanding of the business, sector, and wider conditions and will be able to concentrate on assessing the actual drivers behind financial performance. They will also be better able to engage with the people behind the businesses. Once the routine work is automated, the analyst has the room to arrive at a clear, well-considered judgement on whether the prospective commercial relationship is viable. More importantly, they will be able to provide the reasoning behind that conclusion.
Adopting a hybrid approach to automation
In standardised lending, automation is well established. In complex lending, the case is still taking shape, given that the tools are relatively new, and their role within established credit processes is still being defined. But the direction is becoming clear, with automatic spreading functionality turning financial statements into structured, analysis-ready data, and automatic reporting drawing the ratio tables and charts from that same data. Tools such as FMT bring these together, handling the preparation while leaving the decision with the credit team. The key to automating complex lending is therefore to take the preparation off the analyst’s desk and return that time to the analysis and judgement these teams are best placed to provide.
