We used to model commercial lending as a 47-step BPMN diagram. Average time to decision was 11 business days. We rebuilt the same workflow as a case with structured stages and adaptive tasks. Decisions now take 2.9 business days on average. The bank shipped the rebuild in a quarter.
Commercial lending at Accrual covers mid-market companies across APAC and EMEA, with ticket sizes from S$500K to S$250M. It is a business where getting to yes or no quickly, with the right structure on the deal, matters enormously. For two years we tried to run it on a BPMN diagram. This is the story of why that stopped working and what we did instead.
When we first modeled commercial lending in 2023, we mapped every approved path through the process. Intake to KYC to financial analysis to covenant modeling to credit committee to term sheet to legal to signature to funding. Every known fork, every approval tier, every special case was a branch on the diagram. The result was a 47-step BPMN process that covered our complete credit policy.
It was beautifully accurate and operationally useless. Relationship bankers did not look at the diagram to find out where a deal was. They checked email, Slack, and three different spreadsheets. The diagram described the happy path in granular detail and described real life badly, because real deals are almost never on the happy path.
Every interesting commercial deal has exceptions. Most have six or seven. A client's financials are out of date. A KYC refresh is stale. Sanctions rescreening flags a director with a shared name. Covenants need a modification. The client changes the request mid-flow. Each of these, on a diagram, has to be a branch. We kept adding branches. We crossed 47 steps and were heading for 80 when we stopped and asked whether the shape of the tool was the problem.
A diagram that models every known branch is a diagram nobody maintains. We were spending more time updating the BPMN than we were shipping credit decisions.
The rebuild reframed commercial lending as a case with four stages and a bank of adaptive tasks. The stages are fixed. The tasks are flexible.
Stage 1: Intake. Client submits, relationship banker enriches, documents ingested and classified. KYC refreshed where stale. Initial sizing completed.
Stage 2: Credit analysis. Financials parsed, covenants modeled, peer benchmarks pulled, internal exposure computed. Risk appetite checked against policy.
Stage 3: Approval. Tiered approvals per ticket size. Credit committee review where required. Conditional approvals captured as part of the case record.
Stage 4: Closing. Legal review, documentation, signature, funding. Funding confirmation closes the case.
Within each stage, tasks are created, reassigned, reopened, and closed as the deal demands. KYC refresh is a task. Covenant modification is a task. Sanctions rescreening is a task. Conditional approval sign-off is a task. A deal can have five tasks in flight simultaneously inside a single stage. The case does not mind.
The relationship banker's day changed first. Before the rebuild, a banker spent most of their time chasing the status of deals across multiple tools. After the rebuild, the case is the tool. The banker opens a deal, sees every open task, every pending decision, every document in one place, and acts from there.
The credit analyst's day changed next. Credit analysis used to be a hand-off where the banker sent a package and the analyst reviewed in isolation. Now, analysis happens inside the case. The banker can see what is being analyzed and can respond to analyst queries without a separate email thread.
Exception handling changed most dramatically. Exceptions used to be the thing that broke the diagram and forced the deal into manual handling. Now exceptions are just additional tasks on the case. The tool accommodates whatever the deal needs.
Time-to-decision is the headline. The number we care about more is zero exception tickets. Before the rebuild, we averaged four exception tickets per week, where a deal had hit a case the diagram did not cover and the work had to route around the tool. Since go-live three quarters ago, we have filed zero. The tool holds every deal without needing to be reshaped.
Internal audit was the most sceptical stakeholder in the rebuild. Auditors are used to tracing a path through a diagram. A case does not have a path. It has stages and tasks. Our answer, developed over eight conversations across two months, was to treat every task closure, approval, and stage transition as an immutable event on the case.
An auditor asking "how was this decision made" replays the event log in order. Every approval has a named approver, a timestamp, and the document set that was in the case at the moment of approval. The chain is continuous and cannot be edited retroactively. Audit accepted the model, and noted on review that the new system captured more detail than the old diagram plus email threads did.
A common question during the rebuild was whether we were moving away from BPMN. We were not. Accrual Capital still uses BPMN for the parts of commercial lending that are genuinely linear, such as the funding disbursement sub-flow. We now use case management for the parts that are not linear, which is the credit analysis and approval stages. The two run on the same orchestration runtime. A case can call a BPMN sub-process. A BPMN can open a case and wait for its closure.
This layered architecture is what lets us pick the right tool for each shape of work. Linear where linear. Case where case. Same platform.
We would have started with one market rather than all of APAC and EMEA simultaneously. We bit off too much in the first phase and spent the first six weeks fighting regional differences in credit policy that we could have deferred.
We would have drafted the audit story before we built the case model. We built the model first, then spent two months retrofitting the audit story onto it. Building the two together would have saved six weeks.
We would have involved risk committee chairs from day one, not month two. They turned out to be among the most supportive stakeholders once they understood the model. Starting with them earlier would have made the rollout smoother.
Our VP Commercial Lending wrote a more opinionated post on this: Credit is not a straight line. For the orchestration runtime context, see Same product, new front door.
Talk to an Accrual Capital relationship banker. Response within one business day.