Articles
From hierarchy to intelligence: Why operating models must evolve for AI
Unlocking AI value starts with the operating model
You’ll have heard by now: despite unprecedented investment in AI, most companies are struggling to realize meaningful value.
- Yet more than 80% report no significant earnings impact
- And nearly three-quarters struggle to scale beyond pilots
AI adoption is widespread. Transformation is not. The constraint isn’t technology. It’s the operating model.
AI isn’t just improving work. It’s changing it
Most organizations are applying AI to operating models designed for a different era – layering intelligent tools onto processes built around human labor.
But AI introduces a fundamentally different type of worker. AI agents don’t just assist. They can reason, decide, execute multi-step workflows and improve through feedback. That breaks a core assumption that has shaped organizations for decades: Humans do the work. Technology supports them.
When that assumption changes, optimization is no longer enough. You don’t improve the model – you redesign it.
The real shift is that services are becoming software. For decades, software automated tasks. AI is now automating entire services.
Activities that once required teams – customer support, underwriting, financial analysis – are increasingly delivered through AI-driven workflows.
In industries like insurance, multi-step processes such as: risk submission, underwriting, policy issuance and claims handling are collapsing into real-time, end-to-end execution. What previously required multiple teams, systems and handoffs can now be orchestrated by AI, with humans stepping in only at critical moments. This isn’t just efficiency. It’s a restructuring of how value is created.
Why organizations are struggling
Most organizations are still approaching AI through incremental optimization – adding copilots, automating individual tasks and looking for productivity gains within existing structures.
But those structures were designed for human execution. Traditional organizations are, at their core, information routing systems. Hierarchy exists to move context, decisions and priorities across the business.
AI changes that. For the first time, coordination doesn’t have to sit in layers of management – it can sit in the system itself. The problem is the question organizations are asking.
Most ask: ‘Where can we apply AI?’
The better question is: ‘If AI did the default work, how would we redesign this outcome?’
A task-based mindset anchors you to the current model. An outcome-based mindset allows you to reinvent it. When AI is forced into legacy structures, the result is incremental improvement – not transformation.
What needs to change
To unlock real value, organizations need to rethink how work is done, how teams are structured and how value is created.
1. Work shifts from tasks to workflows
Instead of people executing tasks across functions, AI operates end-to-end workflows.
Humans focus on:
- Judgment
- Exceptions
- Risk and escalation
2. Organizations shift from functions to outcomes
The traditional model:
Marketing → Sales → Operations → Service
This is replaced by end-to-end workflows aligned to outcomes (e.g. ‘acquire customer’, ‘resolve claim’). These workflows cut across functions and are increasingly orchestrated by AI.
3. The role of humans changes
As AI takes on execution, people move to the ‘edge’:
- Handling ambiguity and novel situations
- Making ethical and high-stakes decisions
- Providing context AI cannot fully capture
New roles emerge:
- AI workflow owners
- Agent trainers and evaluators
- Human-in-the-loop reviewers
- Exception managers
One individual, enabled by AI, can now deliver the output of entire teams.
4. Economics fundamentally shift
AI drives:
- Near-zero marginal cost of execution
- Faster service cycles
- Mass personalization
As a result:
- Value chains compress
- Intermediaries are removed
- Advantage shifts to those who control data, models and customer access
AI doesn’t just change how work gets done. It changes who captures value.
How to start: a practical playbook
Most organizations don’t struggle with understanding this shift. They struggle with where to begin. The answer is neither small pilots nor enterprise-wide transformation. It’s somewhere in between.
1. Apply the Goldilocks principle
Don’t automate a task. Don’t attempt to redesign the entire business.
Instead, select a bounded, high-value workflow:
- End-to-end enough to matter
- Contained enough to move quickly
Examples:
- Acquire a customer
- Resolve a claim
- Onboard a merchant
2. Redesign from zero
Assume AI does 70-80% of the work by default.
Then design:
- The new workflow
- Where humans intervene (judgment, exceptions, risk)
3. Define the human–AI contract explicitly
Be clear on:
- What AI owns
- What humans own
- What triggers escalation
This is where most transformations fail – because it’s left implicit.
4. Measure outcomes, not activity
Shift metrics away from tasks and productivity.
Focus on:
- End-to-end cycle time
- % of workflow automated
- Exception rates
- Outcome quality
5. Scale by replication
Once a workflow works:
- Replicate the model
- Don’t just roll out the technology
What the future organization looks like
The shift can be summarized simply:
[table “” not found /]
Instead of: ‘This is the marketing team’
You get: ‘This is the customer acquisition workflow – run by AI and governed by humans.’
Final thought
The organizations that will win with AI won’t be the ones deploying the most tools. They will be the ones that redesign how work actually happens.
The starting point is simple:
- Pick one workflow
- Redesign it with AI as the default
- Define where humans add value
- Measure outcomes, not activity
Then do it again. AI doesn’t transform companies. Operating model change does. AI just makes it unavoidable.
