Written by Sabine VanderLinden 

As AI agents become increasingly capable of executing work, the defining question for business leaders is changing. It is no longer simply what machines can do. It is what humans must continue to own.

Your-30 second version

• AI agents can now plan, reason, and execute enterprise workflows: Ten human capabilities decide whether that autonomy creates value: judgement, decision-making, systems thinking, problem framing, critical thinking, empathy, ethical reasoning, orchestration, sense-making, and learning agility.
• The gap is measurable: 78% of companies use generative AI in at least one function, yet more than 80% report no material earnings impact, because they automate tasks without redesigning authority.
• The operating answer is human-in-command, not human-in-the-loop: frame the problem, judge the evidence, decide where authority sits, orchestrate humans and agents, and connect through trust. Execution can be delegated. Accountability cannot.

For decades, we have measured human productivity by our ability to execute. How quickly can we process information? How many decisions can we make? How efficiently can we complete tasks?

Artificial intelligence is challenging that equation.

We are entering the age of the agentic enterprise, where AI systems will increasingly plan, reason, coordinate workflows, interact with other agents, and execute actions with varying degrees of autonomy.

An AI agent might assess an insurance claim, retrieve supporting evidence, recommend settlement terms, and initiate payment, as seen in autonomous AI agents transforming insurance workflows. Another might negotiate with suppliers, coordinate customer interactions or continuously optimize operational processes.

Tasks that once required multiple people, departments and handovers could increasingly be orchestrated by intelligent systems.

But here is the paradox.

The more capable machines become at executing work, the more valuable distinctly human capabilities become.

The future of work is not simply about learning to use AI. It is about understanding which responsibilities humans should retain, which capabilities we must strengthen, and where authority must remain firmly in human hands.

I believe ten capabilities will define the human advantage in an agentic enterprise.

Here is my list....

1. Judgement: Knowing What Should Be Done

AI can evaluate alternatives, recognize patterns, and recommend actions. But optimizing an outcome is not the same as understanding whether that outcome is desirable.

Consider an AI agent recommending that an insurer reject a claim because the contractual evidence supports that conclusion, raising the risk of algorithmic denial in insurance.

Technically correct? Perhaps.

But what if the claimant is vulnerable, the policy language is ambiguous or the decision creates consequences beyond the immediate transaction?

Human judgement integrates context, values, experience, and consequences that an optimization function can't always capture.

The essential skill is knowing when an answer is correct, but the decision is wrong.

As AI assumes more analytical responsibilities, judgement becomes a strategic asset.

2. Decision-Making: Owning the Consequences

Decision-making and judgement are related, but they are not interchangeable.

Judgement evaluates what matters. Decision-making commits an organization to a course of action.

Agentic systems will increasingly make operational decisions within delegated boundaries. Humans must determine where those boundaries sit.

Which decisions can be automated? Which require approval? Which must remain reversible? Who intervenes when an agent encounters an unfamiliar situation?

These are questions of authority, not processing power.

In regulated industries, accountability cannot disappear simply because execution has been delegated.

We can delegate execution. We cannot delegate institutional accountability.

The future leader must become exceptionally skilled at allocating decision rights between humans and machines.

3. Systems Thinking: Seeing Beyond the Individual Task

AI agents are often designed around specific objectives.

Reduce claims-processing time. Improve conversion. Optimize pricing. Lower operating expenses.

But optimizing individual activities does not automatically optimize the enterprise.

An agent that improves underwriting profitability might inadvertently reduce accessibility. An agent that accelerates claims settlement could create new fraud vulnerabilities or undermine claim litigation avoidance strategies with agentic AI. An automated customer journey might improve efficiency while weakening trust.

Systems thinking connects these consequences.

It requires understanding how technology, incentives, people, regulation, economics, and customer behavior interact as companies redesign systems around AI agents.

In an agentic enterprise, leaders must anticipate second- and third-order effects rather than celebrating isolated productivity gains. Many organizations still optimize isolated use cases instead of redesigning the operating model.

Agents may optimize existing workflows. Humans must redesign the broader system for scaling impact.

4. Problem Framing: Asking the Questions That Matter

We spend considerable time discussing prompt engineering.

But the more important capability may be problem framing.

AI is becoming remarkably good at generating answers. That does not mean organizations are becoming equally good at asking meaningful questions.

Imagine an insurance executive asking an agent to reduce claims costs by 20%.

The agent might identify opportunities to automate assessments, reduce manual reviews, or tighten settlement criteria.

But perhaps the better question is: How can we reduce the total cost of risk while improving customer outcomes?

Those two questions could produce fundamentally different strategies.

Problem framing requires curiosity, commercial understanding, and the ability to challenge the original objective.

The competitive advantage will increasingly belong to those who define the right problem, not simply those who solve the assigned one.

5. Critical Thinking: Challenging Machine Confidence

AI systems can produce persuasive recommendations with extraordinary speed.

Unfortunately, fluency is not proof. AI tools can assist analysis, but they cannot replace critical thinking skills.

As agent-generated analysis becomes embedded in enterprise decisions, employees must become more sophisticated at questioning evidence, identifying assumptions and recognizing uncertainty.

Where did this recommendation originate? Which data informed it? What alternatives were rejected? What would invalidate the conclusion?

Critical thinking also means resisting automation bias: the tendency to trust a system because it appears sophisticated or consistently performs well. Efficiency gains can be real, but without human engagement they can also invite cognitive atrophy when people stop interrogating evidence for themselves.

The strongest human-AI teams will not be those where people unquestioningly accept machine recommendations.

They will be those where people know when and how to challenge them.

Trustworthy autonomy requires constructive skepticism and a deliberate focus on the AI trust imperative in business.

6. Empathy: Understanding the Human Experience

Machines can increasingly recognize sentiment, personalize communications, and simulate empathetic responses, but AI lacks true empathy and emotional understanding no matter how convincing the mimicry.

But recognizing emotional signals is not equivalent to assuming responsibility for another person's well-being.

Consider a customer dealing with a serious illness, a family recovering from a natural disaster or an employee facing redundancy.

These situations demand more than efficient responses and illustrate why combining AI with empathy in insurance claims is becoming a critical differentiator.

They require sensitivity, dignity, contextual understanding, and sometimes the willingness to depart from a standardized process.

Empathy is also essential to leadership.

As organizations redesign work around intelligent agents, employees will experience uncertainty about identity, relevance, and career progression.

Leaders need emotional intelligence to build trust through those transitions rather than treating workforce transformation as a purely technological exercise.

In that environment, soft skills become more valuable, not less, as routine work is automated and leaders must interpret concerns, challenge outputs, and keep people at the center.

Efficiency determines how quickly an organization responds. Empathy influences whether that response deserves trust.

7. Ethical Reasoning: Defining What Is Acceptable

An agent may be capable of executing an action without that action being appropriate.

That distinction matters enormously.

Should an autonomous system adjust insurance prices using highly granular behavioral information? Should an AI recruiter prioritize predicted productivity over other considerations? Should an agent make decisions that materially affect vulnerable customers, especially as AI reshapes insurance in an autonomous world?

These questions involve fairness, proportionality, transparency, and competing human interests, especially when designing fair and responsible insurance algorithms.

Governments also need strategies to manage how these systems reshape the world, not just firm-level compliance, and decision-makers must decide how to govern such systems in the public interest.

Ethical reasoning is not simply regulatory compliance.

It requires deciding what an organization stands for when several technically feasible options exist.

Human oversight must therefore extend beyond approving outputs.

It must shape the principles, constraints, and escalation mechanisms governing autonomous behavior.

Capability answers what AI can do. Ethics helps determine what it should do.

8. Orchestration: Leading People and Intelligent Agents

Perhaps the most significant new management capability is orchestration.

Traditional managers coordinate people, resources and processes.

Tomorrow's managers will increasingly coordinate hybrid teams comprising employees, specialized AI agents, enterprise platforms, enterprise software and external digital ecosystems as agentic AI transforms industries.

Their responsibilities will include defining objectives, assigning authority, establishing escalation pathways, and ensuring that multiple agents work toward coherent outcomes.

This changes the nature of management.

Done well, deep integration between agents and enterprise systems can create elastic capacity, but only when orchestration is designed intentionally and AI automation is governed as an execution layer rather than treated as magic, as emerging agentic frontier firms in insurance demonstrate.

Instead of supervising every activity, leaders will design environments where work can happen autonomously without losing control.

Orchestration also demands understanding when human intervention adds value and when it introduces unnecessary friction.

The manager of tomorrow will not simply manage people. They will design and govern systems of intelligence.

9. Communication and Sense-Making: Turning Intelligence Into Meaning

When information becomes abundant, clarity becomes scarce.

Agentic enterprises may generate more analysis, recommendations, and operational intelligence than any leadership team can reasonably absorb.

The challenge will no longer be accessing information.

It will be interpreting what matters.

Human communication must connect data to strategic purpose, turn information into actionable knowledge, translate complexity into decisions, and build alignment across stakeholders.

This is particularly important during transformation, when people need to understand not only what is changing but why.

Storytelling, persuasion, negotiation, and the ability to create shared meaning will become increasingly valuable.

Technical know-how alone is not enough when leaders must help knowledge workers interpret AI-driven outputs and act with confidence.

AI can generate a thousand explanations. Leadership must create one compelling direction.

10. Learning Agility: Remaining Relevant as Work Evolves

The final capability is the willingness to continuously learn, unlearn, and adapt.

Agentic AI will reshape job descriptions, organizational structures, and the relationship between expertise and execution.

Technical skills will remain important, but enduring human skills will matter even more because they outlast tools that become outdated faster than traditional training programs can accommodate.

Employees will need to experiment with emerging tools, redesign their workflows, and develop new combinations of human and machine capabilities across the creative process. A growth mindset helps people stay with productive struggle, build focus through the hard parts of learning, and resist outsourcing every difficult step to generative AI.

This demands intellectual humility.

Experience remains valuable, but yesterday's expertise cannot become tomorrow's constraint.

The most adaptable professionals will not define themselves exclusively by the tasks they perform.

They will define themselves by the value they create.

And underpinning all ten capabilities is one essential human quality: curiosity.

Curiosity encourages us to question assumptions, explore possibilities, and challenge the boundaries of what we believe is achievable.

From Human-in-the-Loop to Human-in-Command

These ten capabilities can be distilled into five responsibilities:

FRAME → JUDGE → DECIDE → ORCHESTRATE → CONNECT

Frame the problems worth solving. Judge evidence, context, and risk. Decide where authority belongs. Orchestrate humans and intelligent systems. Connect through empathy, trust and shared purpose.

This is more than a workforce-skills agenda.

AI adoption is broad but uneven: 78% of companies use generative AI in at least one function, yet over 80% report no material earnings from AI initiatives, and most organizations are still learning the hard way that experimentation alone does not create enterprise value-a reminder of the gap between AI facts and fiction in the enterprise.

It is a fundamental redesign of organizational authority.

For boards and executive teams, the strategic question is not merely how many tasks AI agents can automate.

It is whether their organizations are developing the human capabilities needed to responsibly direct increasingly autonomous systems.

That means rethinking leadership development, recruitment, performance measurement, and career progression as part of an AI transformation operating challenge, not merely a talent one.

Scaling impact requires research, disciplined governance, and a venture-clienting approach such as the DIVAAA framework.

We must stop rewarding people exclusively for the volume of work they produce and start recognizing the quality of their judgement, the decisions they own and the outcomes they enable.

The organizations that succeed will not necessarily be those deploying the largest number of agents.

They will be those that establish the clearest relationship between autonomy, authority and accountability.

The future of work is not about humans competing with machines at execution. It is about humans becoming better at the responsibilities that make execution meaningful.

AI may increasingly determine how work gets done.

Humans must remain responsible for why it matters, what boundaries apply, and who bears the consequences.

That is the human advantage.

And it may become the most important competitive advantage of the agentic enterprise.

A question for leaders: If AI agents could execute 80% of your team's current tasks tomorrow, which human capabilities would you invest in today?

Because the answer may reveal whether your organization is preparing its workforce for the future, or simply automating its past.