Operations Center of Excellence
An Operations Center of Excellence is a centralized function that codifies and scales operational best practices — process design, automation, performance management, change governance — across the enterprise. The disciplined version is an enablement function: it owns standards, methodologies, training, and tooling, while business units retain ownership of execution. The undisciplined version is a delivery function: it tries to own execution itself, becomes a bottleneck, and rapidly degrades into the layer everyone routes around. IBM's published CoE patterns (across automation, AI, process, and analytics) and Bain's research on CoE effectiveness both consistently show the same finding: enablement-model CoEs deliver 2-3× the durable value of delivery-model CoEs, while delivery-model CoEs are dramatically more common.
The Trap
The trap is exactly what the title implies: the CoE is supposed to be a Center of Excellence and instead becomes a Center of Mediocrity. KnowMBA take: Centers of Excellence often become Centers of Mediocrity if they own delivery instead of enablement. Once the CoE owns delivery, three things happen reliably: (1) intake queues stretch to 9-12 months, business units build shadow capability, and the CoE drowns in maintenance; (2) the function becomes a career graveyard for senior people who once delivered and are now custodians; (3) leadership disbands the CoE in a reorg 18-24 months later, having confirmed that 'CoEs don't work' — when in fact the operating model was wrong, not the concept.
What to Do
Stand the CoE up around five enablement workstreams: (1) Standards & Methodology (frameworks, playbooks, reference architectures); (2) Training & Certification (role-based curricula, internal certification ladders); (3) Tooling & Reusable Assets (libraries, templates, accelerators); (4) Governance & Quality (peer review, gate criteria, controls); (5) Community & Advocacy (practitioner community of practice, internal events). Crucially: the CoE does NOT own execution. Business units own execution; the CoE makes them better at it. Headcount: typically 6-15 FTE for an enterprise of 5,000+. If headcount climbs above ~20, you are likely sliding into delivery mode and should restructure.
Formula
In Practice
IBM has published patterns and customer stories across its various CoE practice areas (Automation CoE, AI CoE, Process CoE) consistently arguing for the federated/enablement model: a small central team owns standards and reusable assets while distributed practitioners in business units own delivery. Bain & Company's published research on CoEs across procurement, analytics, and shared services reaches a similar conclusion — that the CoEs delivering durable value are the ones structured as enablement functions, not delivery functions. Both bodies of work are widely cited in enterprise transformation playbooks.
Pro Tips
- 01
The most reliable test of whether your CoE has slid into delivery mode is the intake backlog. If it exceeds 60 days for the average request, the CoE is now a bottleneck. The fix is not more headcount — it is restructuring scope so business units can self-serve against CoE standards.
- 02
The career path for senior CoE talent must be back into business units, not deeper into the CoE. The best CoE leaders rotate through every 18-24 months. Permanent CoE careerists become defensive of the function's centralization and resist the federated model that would make the CoE more effective.
- 03
Measure the CoE on adoption, not output. 'We trained 400 people' is an output metric. 'We trained 400 people and 280 of them shipped a project against CoE standards in the last 90 days' is an outcome. The second number is what the CFO should fund against.
Myth vs Reality
Myth
“A CoE will scale best practices across the enterprise”
Reality
A CoE alone scales nothing. What scales is a working operating model: clear standards, distributed delivery capability, and a community that maintains the standards. The 'CoE' is just the org-chart label for that operating model. Calling something a CoE doesn't make it work — and most CoEs in practice scale very little.
Myth
“Bigger CoEs deliver more value”
Reality
Past 15-20 FTE, CoE productivity per head typically declines because coordination overhead exceeds added enablement capacity. The highest-performing CoEs are lean (6-12 FTE) with broad distributed practitioner networks (50-300 trained builders), not large centralized teams. Bain's published research on CoE effectiveness has documented this pattern across multiple practice areas.
Try it
Run the numbers.
Pressure-test the concept against your own knowledge — answer the challenge or try the live scenario.
Scenario Challenge
Your CFO walks in: 'The Process Improvement CoE has 32 FTE, a 14-month intake backlog, and three business units have started building shadow process teams. I'm tempted to disband it. What do you do?'
Industry benchmarks
Is your number good?
Calibrate against real-world tiers. Use these ranges as targets — not absolutes.
CoE Headcount as % of Practitioner Network
Enterprise CoEs across automation, process, analyticsLean enablement
5-15%
Healthy
15-30%
Heavy
30-60%
Centralized delivery (anti-pattern)
> 60%
Source: Hypothetical: synthesized from IBM and Bain CoE pattern research
CoE Intake Backlog (Days)
Average request from business unit to CoESelf-serve mature
< 14 days
Healthy
14-45 days
Strained
45-90 days
Bottleneck
> 90 days
Source: Hypothetical: synthesized from enterprise transformation surveys
Real-world cases
Companies that lived this.
Verified narratives with the numbers that prove (or break) the concept.
IBM (CoE pattern library)
2018-present
IBM has published patterns across its various CoE practice areas (Automation CoE, AI CoE, Process CoE) consistently arguing for the federated/enablement model: a small central team owns standards and reusable assets while distributed practitioners in business units own delivery. The patterns are referenced in IBM Institute for Business Value publications and in customer-facing transformation engagements.
Recommended model
Federated enablement
Central team size
Lean (typically 6-15 FTE)
Distributed network
50-300+ trained practitioners
The published vendor and consulting consensus is unambiguous: enablement-model CoEs work, delivery-model CoEs don't. The widespread persistence of delivery-model CoEs in enterprises is an organizational, not analytical, problem.
Bain & Company (CoE effectiveness research)
2019-present
Bain's published research on CoEs across procurement, analytics, and shared services reaches the same conclusion as IBM's pattern library: enablement-model CoEs deliver durable value; delivery-model CoEs do not. Bain's surveys consistently find that ~70% of enterprise CoEs are structured in delivery mode and ~80% of those are rated as underperforming by their executive sponsors within 24 months.
% of CoEs in delivery mode
~70%
% rated underperforming within 24 months
~80%
Recommended structure
Enablement
The data has been clear for a decade. The persistence of delivery-mode CoEs is a leadership failure, not an analytical one.
Related concepts
Keep connecting.
The concepts that orbit this one — each one sharpens the others.
Beyond the concept
Turn Operations Center of Excellence into a live operating decision.
Use this concept as the framing layer, then move into a diagnostic if it maps directly to a current bottleneck.
Typical response time: 24h · No retainer required
Turn Operations Center of Excellence into a live operating decision.
Use Operations Center of Excellence as the framing layer, then move into diagnostics or advisory if this maps directly to a current business bottleneck.