AI Operating-Model Decisions for COOs
A workshop for COOs on how responsibilities, workflows, and decision rights change when AI enters operations, and which decisions must remain with people.
Direct answer
Direct answer: COO operating model
When AI enters operations, the hard questions are about responsibility, not tools. Someone must decide which steps a system may perform, who reviews its output, who owns its errors, and who can stop it. Each workflow also needs a named sponsor before a pilot starts. This session helps a COO make those operating-model choices for their own processes and produces a pilot sponsorship brief for the first candidate.
The question
“Which responsibilities and workflows must change when AI enters operations, and which decisions must remain with people?”
Who it is for
COOs, operations directors, and heads of service or delivery who own processes AI will change.
What you leave with
An operating-model decision map, ownership choices, and a pilot sponsorship brief.
- 01 An operating-model decision map for the processes in scope: assist, automate, or leave alone
- 02 Ownership choices for each AI-supported step: operator, reviewer, escalation owner, and system owner
- 03 A list of decisions that must stay with people, with the reason for each
- 04 A pilot sponsorship brief for the first candidate process: scope, measures, sponsor, and stop conditions
How it runs
Format, method, and preparation
- Duration
- A preparation call and a half-day workshop, followed by a written decision map and pilot brief
- Delivery
- In person in the UK or remote
- Participants
- The COO and operations leaders, with process owners for the workflows in scope
Map the work
Choose two or three core processes and map where judgement, data, approvals, and customer impact sit today.
Decide the role of AI per step
For each step, decide whether AI assists, acts with approval, acts and reports, or stays out, and why.
Assign operational ownership
Name who operates, reviews, escalates, and owns the system, and what changes in existing roles and service levels.
Sponsor the first pilot
Turn the strongest candidate into a pilot brief with a sponsor, measures, review effort, and conditions to stop.
Useful to have ready
- →Two or three candidate processes with their owners available for the session
- →Current service levels, volumes, and known error or rework points
- →Any existing AI pilots, purchases, or tool policies affecting operations
Evidence
What this draws on
Regulated banking AI at Aveni
Architected enterprise banking AI with conduct-risk workflows, evidence generation, human review, escalation, evaluation, versioning, and release controls. Part of the Aveni team in the first FCA Supercharged Sandbox cohort; the FCA lists Aveni as an accepted firm, which is not an FCA endorsement or certification.
Inspect the source ↗Document and CRM workflows
Python and AWS systems for document extraction, classification, validation, routing, and LLM integration with CRM and operational workflows.
AI for Everyday Automation (Packt, 2026)
Seven connected workflows for email, meetings, research, reports, task tracking, spreadsheets, and professional communication, each with people kept in review.
Read more →Scope and limits
What this is not
- —This session decides the executive operating model. It does not coach staff on daily tasks or implement workflow tools; that is enablement and delivery work at dipankar.co.
- —The decision map is a design for the organisation to adopt. It is not a regulatory opinion and it does not certify that a process is safe to automate.
- —Process measures and savings remain estimates until a pilot measures them.
If the need is different
Common questions
Answers before you commission
What should a COO decide when AI enters operations?+
Which steps AI may assist with or perform, who reviews its output, who owns its errors, how service levels and roles change, and who can stop it. These are operating-model decisions, not tool choices.
Which operational decisions should remain with people?+
Decisions that are irreversible, materially affect customers or employees, involve conflicting evidence, or require accountability that cannot be delegated to a system. The session records the reason for each.
How should an AI operations pilot be sponsored?+
With one named sponsor who owns the process, a narrow scope, measures of quality and review effort as well as speed, and conditions under which the pilot stops or is rolled back.
How is this different from the human–AI operating-model workshop?+
This session is for the COO’s own processes and ends in a pilot brief. The human–AI operating-model workshop works across functions to produce an organisation-wide decision-rights matrix.
Related