Adding AI to a process built without it produces occasional use. Here is how operations teams pick workflows, redesign the work, and set decision rights.
TL;DR
Operations leaders are often handed an AI brief on top of a full job. The pressure arrives from three directions at once: vendors selling tools, boards asking for results, and teams quietly experimenting on their own.
This guide covers what an AI operating model means for an operations function, what the 2026 data says, how to pick the first workflows, and how to redesign the work around them.
We use a plain definition. An AI operating model is the arrangement of work, roles, decision rights, data, and tools that lets people and AI do the work together. It answers practical questions. Which steps does a system do? Which does a person do? Who decides when the two disagree? Who notices when quality slips?
Deloitte Africa describes the AI operating model as the next critical enterprise layer, institutionalised through an AI Centre of Excellence. That framing suits large organisations. For an operations team in a mid-sized business, the same questions apply at a smaller scale, and a small group with clear ownership can answer them.
We look at five dimensions: strategy, data and systems, people, process, and tooling. Operations touches all five, and it tends to stall on whichever is weakest.
Deloitte's State of AI in the Enterprise 2026 report splits organisations into three groups. About a third (34%) use AI to transform the business, by creating new products or reinventing core processes. Another 30% redesign key processes around AI. The remaining 37% use it at a surface level, with little or no change to existing processes. All three groups capture some productivity gains, and 66% of organisations report productivity and efficiency improvements. Deloitte notes that only the first group are reimagining the business rather than optimising what exists.
The same report found that insufficient worker skills are the biggest barrier to integrating AI into workflows, and that educating the workforce was the most common adjustment to talent strategy, at 53%. Redesigning career paths and mobility came in at 33%. Deloitte's own reading is that far fewer organisations are re-architecting roles and workflows than are educating employees.
McKinsey's 2026 State of AI survey points the same way. Its summary says nearly three-quarters of high performers report fundamentally redesigning workflows because of their AI use, up from 55% a year earlier. I could read only McKinsey's summary, not the full report. BCG's 2026 AI at Work survey reaches a similar conclusion: AI is changing jobs faster than companies are redesigning operations.
Start with a short list of recurring workflows and score each on five questions.
Score each from one to five and pick workflows that rate high on volume and pain with manageable risk. Our guide to why AI pilots fail explains why a small, specific workflow with an owner beats an ambitious one without.
Adding a tool to a process designed without it produces occasional use. Redesign means working through the process step by step.
First, map it as it runs today, including the workarounds. Second, mark every step as automate, AI-assist, or human. Rule-based steps go to automation, steps that interpret messy language or documents go to AI with a person checking, and steps that need judgement or accountability stay human. Third, define the handoffs. In some steps a person approves each output before it moves on. In others a person monitors the flow and steps in when something looks wrong. Decide which applies at each step, and write it down.
Fourth, build an exceptions queue. Anything the system cannot handle, or handles with low confidence, goes to a named person with a clear deadline. Deloitte describes advanced organisations as streamlining workflows that AI can execute end to end, while people focus on judgement, exception handling, and oversight. The exceptions queue is where that division becomes real.
An operating model needs named owners. Each redesigned workflow needs a process owner who can change how the team works, someone who monitors output quality, and a clear rule for who approves changes to the system. Deloitte points to new roles appearing, including AI operations managers and quality stewards, which signals that this work is becoming a structural part of how operations are organised. In a mid-sized team, one person may hold several of these roles at first.
Governance matters more as autonomy grows. Deloitte reports that only one in five companies has a mature governance model for autonomous AI agents. Set the decision rights before giving any system more independence, and see our guides to AI agents for business and AI governance.
Record a baseline before you change anything, then compare after 90 days. Useful measures for operations include:
Licences and logins are inputs. Our guide to AI adoption covers how to tell whether work has changed.
If you want a scored view of how ready your operations are across strategy, data, people, process, and tooling, the AI Readiness Diagnostic is free to take. Where the right answer is a custom workflow or internal platform, we cover that under AI consulting for operations.
It is the arrangement of work, roles, decision rights, data, and tools that lets people and AI do the work together. It defines who does what, who decides, and who monitors quality.
Choose a recurring, high-volume workflow with clear pain, available data, and manageable risk. Score candidates on volume, pain, rule clarity, data availability, and risk, and start with the best score.
Large organisations often benefit from one. A mid-sized operations team can start with a small group that has clear ownership and decision rights, then formalise it as use grows.
Strategy says where AI should create value. The operating model says how the work, roles, and decisions are arranged to deliver it day to day.
It does not have to. We have never run an engagement where headcount reduction was the goal. Deloitte describes the aim as complementary partnerships, with people focusing on judgement, exceptions, and oversight.
Ready to move from reading about AI to getting your organisation ready for it? Start with the diagnostic.
Take the AI Readiness Diagnostic