This loan approval workflow case study is worth telling because the fix was not what anyone expected. Everyone assumed the approval process was slow because people were slow. They were wrong. The people were fine. The documents were the problem, sitting untouched in email inboxes, waiting for someone to notice them.
A multifinance company I worked with, mid-sized, several branches around Jabodetabek, was losing deals because approvals took too long. A customer would apply, and it would take three to five days to get a decision. In consumer financing, that is an eternity. Competitors were saying yes while this company was still shuffling paper, and applicants walked.
They thought they needed more staff or a faster credit team. What they actually needed was to see where the time went. When we mapped it, the answer was almost embarrassing, and the fix required no artificial intelligence at all.
Mapping Where the Time Actually Went
Before changing anything, we tracked ten real applications end to end, timestamping every handoff. The picture was clear within a day.
The actual work, checking the documents, assessing the risk, making the decision, took a couple of hours total across the whole chain. Everything else was waiting.
| Stage | Time spent | Working or waiting? |
|---|---|---|
| Branch collects and emails documents | 20 minutes | Working |
| Sitting in credit analyst inbox | 1 to 2 days | Waiting |
| Analyst reviews | 45 minutes | Working |
| Emailed to supervisor for approval | Same day to next day | Waiting |
| Supervisor approves | 15 minutes | Working |
| Emailed back to branch, customer notified | Half a day | Waiting |
Add up the "working" rows and a decision took roughly two hours of actual effort. Add up the "waiting" rows and it took three to five days. More than ninety percent of the turnaround was documents sitting idle in someone's inbox, buried under other mail, with nobody knowing whose turn it was or how long it had been stuck.
That is the real lesson of this case. The bottleneck was not human speed. It was invisible handoffs. When a request lives in an email chain, nobody can see the queue, nobody owns the wait, and urgent and trivial cases look identical in a crowded inbox.
The Fix: One Shared Queue With SLA Timers
The solution was deliberately boring. No machine learning, no credit-scoring algorithm, no grand transformation. We built one shared queue.
Every application became a card in a single shared workflow that every person in the chain could see. The card moved through clear stages: submitted, under review, awaiting approval, decided. Three things changed the game:
- Visibility. Everyone could see every pending application and exactly which stage it sat in. The queue was no longer hidden inside individual inboxes. A supervisor could glance at the board and see the whole pipeline.
- SLA timers. Each stage had a target time. A card turned yellow as it approached its limit and red when it breached. Suddenly a stuck application announced itself instead of hiding. Nobody had to remember to chase, the board did the chasing.
- Clear ownership. At every moment, the card showed whose turn it was. No more "I thought you had it." The next actor was named on the card.
That is the whole intervention. Digitizing the handoffs, not the judgment. The credit analysts still made the credit decisions with the same expertise. We just stopped the paperwork from getting lost between them.
The Result
Turnaround dropped from three to five days to well under one, most applications decided the same day. Effectively a cut of more than half, often much more, on the metric the business cared about.
The knock-on effects mattered more than the headline number. Fewer applicants walked away, because a same-day yes keeps the deal. Supervisors stopped spending their mornings hunting through email for stuck cases. And for the first time, management could see the pipeline as a whole and spot a genuine backlog forming before it became lost revenue. That visibility is its own asset. If you want the deeper distinction between watching a live process and analyzing it after the fact, I wrote about that in dashboards vs reports.
The team's reaction stuck with me. They had braced for a painful software rollout and got, in their words, "just a board that shows us what's waiting." The best operational fixes often feel that anticlimactic.
The Lesson: Digitize the Handoffs First
If your process feels slow, resist the urge to blame the people or reach for artificial intelligence. Track ten real cases and timestamp every handoff. I would bet most of your turnaround is waiting, not working, and that the waiting happens in the gaps between people, in inboxes and chat threads and physical trays where work goes to sit.
Fix the handoffs before anything else. A shared queue with visibility, timers, and clear ownership is cheap, fast to build, and often delivers more than a far more expensive "intelligent" system would, because it targets the real bottleneck instead of an imagined one. This is the same principle behind why so many transformation projects fail: they automate judgment nobody asked them to touch and ignore the plumbing that was actually broken.
The Practical Takeaway
This multifinance company halved its approval time with a shared queue, not an algorithm. The bottleneck was documents waiting in inboxes, and the fix was making the waiting visible and owned.
Before you buy anything clever, map your own process by timestamping the handoffs. Find where work sits idle, put those handoffs into one shared queue with SLA timers and named ownership, and watch most of your delay disappear. Digitize the handoffs first. The intelligence can come later, if you even still need it.