INSIGHT

Designing Human-Led, AI-Enabled Operating Models

How to translate the principle of human-led AI into practical decisions about workflows, roles, accountability, escalation, and control.

By Juan A. Flores – Published June 4, 2026

Start With the Work, Not With AI

Designing an AI-enabled operating model should not begin by asking which activities can be automated. It should begin by understanding the work itself: what outcome is required, which decisions determine that outcome, what information is needed, where expertise matters, and who is accountable.

Only then does it make sense to decide where AI can contribute. Some activities can be substantially automated, others can be accelerated or supported, and some should remain predominantly human because context, relationships, judgment, or accountability are central to the work.

This distinction is the foundation of a human-led operating model. The objective is not to maximize AI participation. It is to design the most effective combination of human and technological capability.

Deconstruct the Workflow Before Redesigning It

AI creates an opportunity to rethink workflows that may have evolved around previous technological limitations. Simply inserting AI into each existing step can preserve unnecessary work while making it happen faster.

A useful redesign starts by separating the workflow into its underlying activities. Where is information gathered? Where is it interpreted? Which steps exist because systems do not communicate? Where are decisions made? Which approvals manage genuine risk, and which exist because the organization has always worked that way?

Once the workflow is visible, teams can determine what should disappear, what can be automated, what AI can support, and where human involvement remains essential. This prevents organizations from using new technology to reinforce old complexity.

Define the Role of AI Explicitly

AI can play very different roles within the same organization. It may retrieve knowledge, summarize information, generate alternatives, identify patterns, draft materials, recommend actions, coordinate tasks, or execute predefined activities.

Those roles should not be treated as equivalent. An AI system that helps retrieve internal knowledge creates a different operating requirement from one that recommends how a customer should be engaged or produces material intended for external use.

For each workflow, organizations should be able to explain what AI is doing and what it is not doing. Ambiguity about the role of AI quickly becomes ambiguity about accountability.

Decision Rights Must Remain Visible

Every AI-enabled workflow eventually reaches decisions. Some are routine and reversible. Others affect customers, resources, reputation, regulatory obligations, employees, or strategic direction.

The operating model should specify who owns those decisions, which can be delegated within predefined boundaries, and which require explicit human judgment. That ownership should remain clear even when AI has performed most of the analytical or preparatory work.

This is where Human-Led. AI-Enabled. becomes practical rather than philosophical. Human leadership does not require people to perform every task. It requires accountability for consequential decisions to remain identifiable.

Design Human Review Around Consequence

Putting a human approval step after every AI output is not a sustainable operating model. It can eliminate much of the efficiency AI creates while encouraging superficial review because people are asked to validate too much.

Human intervention should instead reflect the consequence and uncertainty of the activity. Low-risk, repeatable work may require limited oversight, while ambiguous or consequential decisions deserve deeper review, challenge, or specialist involvement.

The important question is not whether a human is present somewhere in the workflow. It is whether human judgment is applied where it can materially change the quality or safety of the outcome.

Build Escalation Into the Model

Not every situation will fit the rules designed in advance. AI-enabled workflows therefore need clear conditions under which work moves from normal processing to human intervention.

Those conditions might include insufficient evidence, conflicting information, unusual customer circumstances, regulatory sensitivity, outputs outside defined confidence or quality thresholds, or decisions with consequences beyond the authority of the person using the system.

Good escalation design prevents two opposite problems: people escalating everything because they do not trust the system, or failing to escalate because the workflow makes intervention difficult.

Redesign Roles Around Contribution

When AI assumes more analytical, administrative, or production work, job descriptions alone do not tell us how roles should evolve. Organizations need to reconsider where each role contributes most effectively.

Some people may spend less time assembling information and more time interpreting it. Managers may move from reviewing production to challenging recommendations and resolving exceptions. Specialists may be able to extend their expertise across more teams because AI improves access to their knowledge.

This can increase the leverage of experienced people, but only if organizations deliberately redesign responsibilities. Otherwise, AI simply becomes additional work layered onto existing roles.

Connect Governance to the Workflow

Governance should be designed alongside the operating model rather than attached after implementation. Data permissions, approved tools, validation requirements, documentation, human review, and escalation should become part of the workflow itself.

The level of control should vary according to risk. A low-consequence internal productivity use case should not necessarily require the same governance as an application influencing external communication or a material commercial decision.

When governance and workflow design happen together, responsible use becomes easier to scale. Teams know the boundaries before they begin rather than discovering them through an approval process at the end.

Measure the Operating Model, Not Just the Technology

AI initiatives are often measured through adoption, usage, productivity, or technical performance. Those measures are useful, but they do not reveal whether the operating model itself has improved.

Organizations should also examine whether decisions are better or faster, whether unnecessary work has disappeared, whether expertise is being used more effectively, whether handoffs have been reduced, whether exceptions are handled appropriately, and whether accountability remains clear.

An AI-enabled workflow that produces more output but requires more coordination, review, or correction may not represent progress.

Learn Before Scaling

Operating models should evolve through use. Initial assumptions about where AI performs well, where people need greater control, or which exceptions matter most will not always survive contact with real work.

Organizations should therefore create feedback loops that allow workflows, controls, responsibilities, and escalation rules to change as experience accumulates. Scaling should follow evidence that the model works, rather than simply the technical ability to deploy it more widely.

The objective is not to design a perfect AI operating model once. It is to build an organization capable of improving the division of work between people and technology over time.

The JUYMO Perspective

At JUYMO, we believe human-led AI operating models become useful when the principle is translated into the mechanics of everyday work. That means deciding explicitly what AI does, what people do, who makes decisions, where judgment enters, when escalation occurs, and how accountability is preserved.

The strongest model is not the one that automates the most. It is the one that uses AI where it creates meaningful leverage while keeping experienced people close to the decisions where context and consequence matter.

Human-Led. AI-Enabled. is ultimately an operating-model choice about where technology creates leverage and where humans remain responsible.