Bad processes cost organisations an estimated 30% of annual revenue — and in most APAC banks, the team formally responsible for fixing that problem is the Process Centre of Excellence (CoE). The structural case against the traditional CoE is this: the model is wrong, not the people.
A Process CoE centralises scarce process expertise in a small team and asks that team to serve the entire organisation. That model works at low process volume. At the scale of a mid-size APAC bank — with dozens of operations functions, hundreds of sub-processes, and continuous regulatory change — the CoE is perpetually behind. Every backlogged improvement request is unresolved process debt. Every process the CoE cannot reach remains a shadow process.
This post makes a structural argument: the CoE fails because of architecture, not execution. Agentic platforms succeed because they distribute process capability to the teams that own it.
The architecture of CoE failure
Process CoEs in APAC banks typically fail for one or more structural reasons.
Scarcity by design. A CoE is, by definition, a concentration of expertise. Lean Six Sigma practitioners, process architects, and documentation specialists are expensive and scarce. The team is sized for steady-state improvement — typically 5–15 people in a mid-size bank. Demand for process work always exceeds that capacity. The backlog grows. Teams that cannot get CoE time build their own local solutions. The CoE loses visibility into operational reality.
Prioritisation becomes politics. When the CoE cannot serve everyone, someone decides who gets served first. In theory, prioritisation is based on risk and opportunity. In practice, it is influenced by seniority, persistence, and internal relationships. High-volume, low-visibility processes — trade finance exception queues, manual reconciliation, payment-ops workarounds — sit in the backlog while higher-profile transformation projects consume CoE capacity.
The handoff problem. CoE-designed processes are handed to operations teams for execution. Between the design and the execution, something is always lost: context about why a decision was made, how an exception should be handled, what the process looked like before the change. The operations team runs the designed process until it breaks, then develops workarounds that the CoE never learns about. Documentation diverges from reality within weeks.
Measurement stops at delivery. CoE mandates typically end with a delivered design or a trained team. The improvement is declared, the project is closed, and the CoE moves to the next engagement. No one measures whether the process actually improved at the 6-month mark. No one has the authority to iterate when the design meets operational reality. The process reverts — slightly different in form but equivalent in performance.
These 4 failure modes share a root cause: the CoE owns process improvement but does not own process execution. That separation means the feedback loop between design and reality is always broken.
What agentic platforms do differently
An agentic process platform like ESSAM — built on the E-S-S-A-M framework (Eliminate, Simplify & Standardize, Automate, Migrate) — does not centralise process capability. It distributes it.
The distinction matters.
In a centralised model, operations staff submit a request, wait for the CoE, receive a recommendation, and implement it. The process expert is external to the team. The feedback loop requires a new project to re-engage the CoE.
In a distributed model, operations staff describe a process directly to the platform. The platform produces the baseline, the waste analysis, and the improvement recommendation. The team reviews, confirms, and implements — without waiting for a centralised expert. The feedback loop is internal to the team and continuous.
3 capabilities make the distributed model viable in banking operations.
Conversational capture without specialist skill. ESSAM captures process information through conversation. The person who runs the process documents it, without flowchart software, without IT support, without a specialist facilitator. This removes the access barrier that makes CoE backlogs necessary in the first place.
The 7-step improvement cycle as an operational habit. ESSAM's improvement cycle — Baseline, Analyse, Optimise, Document, Deploy, Feedback, Repeat — runs inside the team that owns the process. The Feedback step is not a project re-engagement; it is a fortnightly review built into the team's operating rhythm. Process improvement becomes a habit, not a project.
Measurement as a byproduct, not a project. When the platform runs the improvement cycle, cycle time, exception rates, and straight-through processing rates are tracked continuously. The CoE's hardest problem — measuring whether the improvement held — is solved by default. Operations leadership sees performance data without commissioning a measurement study.
What CoE teams should become
The argument here is for reorientation, not disbanding.
The productive CoE model in an agentic environment is governance, not production. CoE practitioners freed from individual process-improvement engagements can focus on the work that actually requires their expertise.
Methodology governance. Setting the standard for how processes are documented, how improvement cycles are structured, and how the platform's recommendations are reviewed. An agentic platform can surface a recommendation; a governance layer decides whether it is right for a regulated environment.
Risk and compliance review. Reviewing the process changes that agentic platforms propose against regulatory requirements. MAS and BNM requirements do not disappear in an agentic environment — they require expert interpretation. CoE practitioners are best positioned to provide that interpretation systematically, across all teams.
Capability building. Training operations staff to engage with process improvement as a continuous practice rather than a one-time project. The CoE's role shifts from doing to enabling. More than 10,000 Lean Six Sigma professionals have built the foundational expertise that ESSAM is designed to amplify, not replace.
Portfolio-level visibility. When every team runs its own improvement cycles, someone needs to see the portfolio: which processes are improving, which are stalling, and where enterprise-level risks are accumulating. CoE practitioners with visibility across teams are better positioned to do this than any individual team.
The reoriented CoE is smaller but more valuable. It does not sit in a backlog. It sets the standard that all teams work to.
The Kuwait evidence: execution without CoE dependency
Abdulla Al-Awadi, former Chief Strategy Officer (CSO) of a Kuwait bank, applied ESSAM's methodology to procurement — a process that in many APAC banks sits permanently in the CoE backlog because it is complex, cross-functional, and politically sensitive.
The result: a reduction from 139 days to 57 days. A 59% cycle-time improvement. 82 days retired from the process. A 106.9% efficiency gain. This result required a full process redesign. It did not require a CoE engagement of indefinite duration. The methodology was applied directly by the team with the platform as the structure.
The Kuwait result is the only real case cited here. Comparable outcomes in APAC banking are illustrative — the exact numbers will vary by bank, by process, and by the depth of initial documentation. The structural principle, however, is the same: when the team that owns the process has direct access to improvement methodology, the feedback loop closes without waiting for a centralised expert.
Applying the structural argument to your CoE
If you lead a Process CoE in Singapore or Malaysia, 4 diagnostic questions identify where the structural failure is occurring.
How large is your current backlog? Count the process improvement requests received in the last 90 days versus the requests completed. If the ratio is below 0.5, your CoE is less than half as productive as your organisation needs it to be. The backlog is process debt accumulating faster than you can address it.
How much of your capacity is consumed by documentation maintenance? If CoE practitioners spend significant time updating SOPs rather than designing process improvements, the maintenance burden is crowding out improvement capacity. This signals that living-documentation infrastructure is missing.
How often do you measure whether an improvement held? If the answer is "rarely" or "never," your improvement cycle ends at delivery. The improvements you delivered 12 months ago may have reverted. You would not know.
What does operations staff do when they cannot reach the CoE? If the answer is "they build their own workarounds," the CoE is producing shadow processes rather than preventing them. The centralisation is creating the fragmentation it was designed to prevent.
A transition path: governance-first, not CoE-last
The shift from production CoE to governance CoE does not require a restructuring programme. It starts with a single pilot.
Choose one operations team that is willing to run the ESSAM improvement cycle without CoE involvement. Define the CoE's role for that team: governance review of the baseline, compliance sign-off on the redesigned process, methodology consultation if the team reaches a decision they cannot resolve internally.
Run the improvement cycle. Measure the result. If the process improves and the governance role held, replicate with a second team. The CoE's production backlog shrinks as teams develop the capability to improve their own processes. The CoE's governance role grows as the number of teams running improvement cycles increases.
This transition takes 3–6 months in most APAC banking operations. The CoE does not lose relevance — it gains it, because governance over a distributed improvement capability is more strategically significant than managing a backlog.
The transition also changes the CoE's relationship with operations staff. In a production CoE, staff interact with the team when they have a problem and need a project approved. In a governance CoE, the interaction is continuous: the CoE sets standards, teams run cycles, and the CoE reviews outputs. That relationship is more collaborative and better positioned to catch emerging process risks before they become audit findings. The shift is cultural as much as structural, and it requires explicit communication about what the CoE will and will not own. Teams need to know that they now own execution; the CoE needs to know that governance is not a demotion.
Where this structural argument has limits
The case for agentic platforms over centralised CoE production is strongest in high-volume, high-change environments. For complex, cross-functional transformation — reengineering an entire credit origination architecture, for instance — expert-led design remains the correct approach. The platform amplifies expertise; it does not substitute for it when the problem requires deep institutional knowledge and political navigation.
Regulated processes that are functioning correctly and producing compliant outcomes should be documented and monitored, not optimised for efficiency alone. The CoE's compliance expertise is the guardrail that prevents the platform from optimising its way into a regulatory problem.
ESSAM is certified to ISO 27001:2022 and SOC 2 Type II standards, and is GDPR-compliant — the security and data handling requirements for regulated financial services are met. But the process content that runs through the platform still requires human governance. No certification substitutes for that.
The production-vs-governance shift starts with one conversation
If your CoE backlog is evidence of the structural problem described here, the correct scope is smaller than you think. Give one operations team direct access to ESSAM for one process. The CoE provides governance. The team provides execution. Measure the result.
To start that conversation, send a description of the highest-priority process in your current backlog to https://apac.essam.ai/contact. ESSAM returns a structured baseline and waste map within 48 hours — the starting point for deciding whether the governance model is right for your CoE.
Frequently asked questions
Why do process centres of excellence fail in large organisations?
Process CoEs typically fail in large organisations because they centralise scarce expertise and cannot scale to meet the volume of process improvement demand. The backlog grows, prioritisation becomes political, and teams that cannot get CoE access develop their own undocumented workarounds. CoEs are also typically structured to deliver process designs, not to measure whether those designs held — so improvement gains are declared but not confirmed.
What is the difference between a Process CoE and an agentic process platform?
A Process CoE centralises process improvement in a specialist team. An agentic process platform distributes process improvement capability to the teams that own the processes. The key difference is the feedback loop: CoE-delivered improvements require a new engagement to measure and iterate; agentic platforms embed measurement and iteration in the team's operating rhythm. The platform does not replace expert judgment — it removes the access barrier that makes expert judgment scarce.
What should a Process CoE focus on in an agentic environment?
In an agentic environment, a Process CoE should shift from production to governance: setting methodology standards, reviewing platform-generated recommendations for regulatory compliance, building operations staff capability in continuous process improvement, and maintaining portfolio-level visibility across all teams' improvement cycles. This is higher-value work than individual process-improvement engagements. It also requires fewer practitioners, since production work is distributed.
How does the E-S-S-A-M framework apply to CoE transformation?
E-S-S-A-M — Eliminate, Simplify & Standardize, Automate, Migrate — can be applied to the CoE's own operating model. Eliminate documentation maintenance work that an agentic platform handles more effectively. Simplify & Standardize the methodology the CoE governs. Automate the measurement and reporting cycle. Migrate execution to the teams that own each process. What remains is a governance function that is faster, smaller, and more strategically valuable than a production CoE.
Is ESSAM secure enough for regulated banking environments?
ESSAM is certified to ISO 27001:2022 and SOC 2 Type II standards, and is GDPR-compliant. These certifications cover the data handling and security requirements relevant to banking operations in Singapore and Malaysia. Process content shared with the platform is handled under the controls these certifications define. For specific data residency or vendor risk management requirements, contact the team at https://apac.essam.ai/contact.
Related reading: