Back to Insights
Strategy

Pega for banking ops: when a case platform meets an undocumented floor

October 22, 2026
ESSAM Team
Pega for banking ops: when a case platform meets an undocumented floor

Count your live case types, then count the processes they represent. If the first number is more than three times the second, you have a definition problem, not a platform problem. A case platform executes a definition; buy it after the definition, not instead of it.

This post is for operations and transformation leads in Singapore and Malaysian banks who already run Pega, or have a renewal in front of them. You are being asked to justify the next licence tranche. You may not be able to show what the current one encodes.

What is the short answer to "Pega or a process improvement platform"?

It is the wrong either/or. Pega and ESSAM sit in different layers.

  • Execution layer: routes work, enforces SLAs, applies decisioning rules and records every action. Pega is built for this.
  • Definition layer: establishes what the process should be, which steps are waste, who owns it and what the approved version says. ESSAM is built for this.

The useful question is which layer you are buying and whether you already own the other one. Everything below tests that.

What does Pega do well in banking operations?

Strengths come first, because they are real and they are why Pega sits in so many bank estates.

Case semantics at scale. Pega models work as a case with stages, steps, parent and child cases, and service clocks. A dispute, an investigation or an onboarding file is a long-running object with a lifecycle. Simpler tools treat work as a ticket. Pega treats it as a case, which suits banking work better.

Decisioning. Next-best-action logic, eligibility rules and routing decisions can run inside the case, not in a separate system. For a high-volume queue, that matters.

Existing enterprise footprint. Many Singapore and Malaysian banks already license Pega for dispute, investigation and onboarding work. Integrations exist, teams are trained, and security and risk functions have already reviewed it. Starting again has a price that no feature list captures.

Audit-grade case history. Every action on a case is time-stamped and attributable. Under MAS and Bank Negara Malaysia supervision, being able to replay who did what to a customer file, and when, is not optional. Pega does this well.

Pega also offers packaged investigation and dispute capabilities, including Smart Investigate, for teams that want a starting point rather than a blank canvas. The platform's flexibility for handling exceptions is a genuine strength.

Nothing in the rest of this post disputes any of that. The claim is narrower and bounded to one scope: definition versus execution.

What is the category error?

Banks often buy a case platform to run a process they have never defined. The platform then does exactly what it is built to do. It encodes what it was told.

What it was told was the org chart. Each team described its own piece of the work. Each exception raised in a workshop became a case type. Each new case type needed a release. Two years later, the "process" is a case-type inventory that only the implementation partner can read.

Pega did not fail. It built what you described, and what you described was the org chart.

A case platform makes exceptions manageable. Manageable is how they become permanent. Flexibility is the platform's strength, and it is also what lets an undefined process hide in plain sight. The platform is excellent at what happens after the process is defined. It is silent on whether the definition is right.

That is not a criticism of the product. A routing engine cannot tell you that a step should not exist. Asking it to is a category error on the buyer's side.

What does this look like on a real floor?

Illustrative example, not a client case. Consider a bank whose dispute process runs on a case platform with 34 live case types. The ops lead who owns the process has never seen 19 of them.

Walk through how that happens. A regulatory change in one quarter produced a new evidence step. A branch network issue created a workaround, and the workaround got its own case type. A joint-account variant was split out because one team wanted a separate queue. Each decision was reasonable when it was made. None was reviewed against the others.

Now ask the ops lead three questions. Which of these 34 are variants of the same process? Which are retired in practice but still configured? Which carry an SLA that anyone checks?

If the answers take a week to assemble, the platform is not the constraint. The missing definition is. This is what we mean by definition debt: the gap between the number of case types you run and the number of processes you can actually describe.

How do you run the case-type-to-process ratio test?

This is the test from the opening, in full. It takes an afternoon and needs no new software.

  1. Export the live case-type inventory. Include only types that created at least one case in the last 12 months. Retired-in-practice types go in a separate list.
  2. List the processes those types represent. A process is something a named person can describe in one conversation, from trigger to outcome.
  3. Divide. Live case types divided by defined processes gives your ratio.
  4. Read the result against the working thresholds below.
Ratio Reading What it usually means
1:1 to 3:1 Healthy Variants exist for real reasons: product, channel, regulation
3:1 to 6:1 Investigate Some variants are workarounds or org-chart artefacts
Above 6:1 Definition debt The inventory describes the org chart, not the process

These thresholds are ESSAM's working rule of thumb. They are not an industry benchmark, and your own estate may justify different lines. The method matters more than the cut-offs.

Apply it to the illustrative dispute example. Say the 34 live case types map to 4 processes that an owner can describe. The ratio is 8.5:1. Above 6:1, definition debt, and the 19 types the owner has never seen are the first place to look.

What is the definition-readiness checklist for an existing case estate?

The ratio tells you how large the problem is. This checklist tells you where it sits. Score each live case type against the four core questions, one point each.

# Question Evidence that counts as a yes
1 Is a process owner named? A person, not a team, who can approve a change
2 Is there an exception taxonomy? A written list of exception reasons, each with a defined handling path
3 Is there an SLA per case type? A target that is measured and reported, not just configured
4 Is there evidence of change? A record of who changed the case type, when and why

Then add three supporting checks at the estate level:

  • An approved process document exists for each defined process, and someone has confirmed it still matches the current configuration.
  • Frontline staff can find the current version of the work instruction without asking the implementation partner.
  • Someone has reviewed the case types for waste in the last 12 months: duplicated approvals, rework loops, handoffs that add waiting and nothing else.

Score each case type from 0 to 4. A type that scores 4 is defined. A type that scores 0 or 1 is configured but undefined. Count each group.

Scores are rarely uniform. High-volume case types tend to score 4 because someone owns them. The long tail often scores 0 or 1. That split is the useful finding, because it shows which part of the estate can be renewed on trust and which cannot.

How do you decide: renew, rebuild or wrap?

Use the checklist scores and the ratio together. This is the decision rule.

Situation Decision Action
Ratio at or below 3:1, and at least 80% of live case types score 3 or 4 Renew Renew at current scope. Spend effort on cost modelling, not redesign
High-volume flows score 3 or 4, but the long tail scores 0 or 1 Wrap Keep the platform. Define, simplify and retire the tail before the next tranche
Ratio above 6:1, or fewer than half of case types score 2 or above Rebuild Do not expand scope. Redefine the process first, then reconfigure

A note on "rebuild", because it is easy to misread. It does not mean leaving Pega. It means rebuilding the definition and then reconfiguring case types from it. Often the right outcome is fewer case types on the same platform.

"Wrap" is usually the cheapest of the three. You surround the platform with a defined, approved, living process, so every future release has something to be checked against.

What should you model about cost?

List pricing for Pega is not public. Do not trust any figure you read in a blog post, including a figure from a vendor, a consultant or this article. Model it from your own renewal.

What to put in the model:

  • The licence terms in your current agreement, and what changes at renewal.
  • The share of live case types that created cases in the last 12 months.
  • Implementation partner time per release, across the last four releases.
  • Internal time spent on regression testing for each new case type.
  • The cost of change when a regulator, MAS or Bank Negara Malaysia, asks for a different evidence step.

The fifth item is the easiest to miss. Each case type is a unit of ongoing maintenance. A ratio of 8.5:1 means you maintain roughly eight configurations for every process you can explain. That is a cost line even when no invoice names it.

The cost question is also a governance question. At renewal, licence cost and governance fit usually weigh more than features. If you cannot show what the current licence encodes, you cannot defend the next one.

When is Pega the right call?

For some flow shapes, a case platform is the right answer, and a mapping exercise would be a detour.

  • High-volume, well-bounded dispute flows. The process is stable, the exception set is known, the volume justifies configuration, and an owner exists.
  • Long-running investigations with strict evidence trails. Parent and child cases, assignment, escalation and audit history are exactly what the platform is for.
  • Regulated onboarding with many conditional steps. Stage-gated work with decisioning at each gate suits case semantics.
  • Cross-department work where the existing footprint already integrates. Reuse beats rebuild when the integrations took years.

If your estate matches these shapes and scores well on the checklist, renew it. Do not let a definition-first argument talk you out of a platform that is working.

The argument applies when the checklist says the opposite: when you are about to extend a case platform into a process nobody has defined. That is when the definition debt compounds.

Where does ESSAM fit?

ESSAM works on the definition layer. It captures a process in a single conversation, classifies every step using the E-S-S-A-M framework, and deploys living work instructions to the floor over WhatsApp. Teams in Singapore and Malaysia already use that channel.

E-S-S-A-M stands for Eliminate, Simplify and Standardise, Automate, Migrate. Automate is the fourth action on purpose. You design before you automate, and a case platform configuration is a form of automation. Eliminating and simplifying come first, so you do not configure waste into a case type.

In the 7-step AI Lean Cycle (Baseline, Map, Analyse waste, Optimise, Document, Deploy, Improve), the Document and approve step produces the artefact a case platform should consume, not replace. An approved definition gives every case type an owner, an SLA and a reason to exist. The platform then executes it. Our BPM guide for financial services covers how that layering works in practice.

The only real client outcome we can point to is a procurement cycle at a bank in Kuwait.

Real case, procurement only. The cycle fell from 139 days to 57 days, a 59% reduction. That retired 82 days of work, with the same headcount. The result came after Eliminate and Simplify, and before any automation.

Two limits apply. It is a procurement cycle, and it should not be read as a result for dispute or investigation work. It also involved no platform migration. That is the point for this comparison: the gain came from definition work alone.

For the wider tool landscape, see our comparison of low-code automation and AI process engineering and our process improvement software evaluation. If you are new to the role behind this approach, start with what an AI process engineer does. The ESSAM features page lists capture, classification and deployment in detail.

Frequently asked questions

Is ESSAM an alternative to Pega?

No. They solve different problems. Pega executes cases against a defined process. ESSAM defines, simplifies and documents the process. Many banks will use both.

Do we need to replace Pega to use ESSAM?

No. The definition work happens before, and alongside, platform configuration. The output is an approved process that your case platform can consume.

Can I run the ratio test without ESSAM?

Yes. It needs only a case-type export and a conversation with each process owner. The test works with any case platform, not only Pega.

What does a high case-type ratio cost?

List pricing is not public, so model it from your own renewal. The ratio shows how many configurations you maintain for each process you can explain, which is a cost driver on its own.

When should we skip definition work?

When the flow is high-volume, stable, owned and already scores 3 or 4 on the checklist. In that case, renew and spend effort elsewhere.

Start with the number

Run the ratio test this week. Write down the number, bring it to your next renewal conversation, and ask which layer you are paying to maintain.

A case platform will faithfully run whatever process it is given. Make sure the process is one your bank has chosen on purpose.

Map one live process in a single conversation and see which steps survive: apac.essam.ai

← All InsightsESSAM Insights