Inspired by themes originally highlighted in “The Human Element” by Dan Fornelius, CIA, published in the Internal Auditor Magazine (The IIA), December 2025, and expanded here for leaders navigating AI adoption across industries.

AI is now part of everyday work. What once felt experimental has quietly become the operating system of the knowledge economy.

Drafting content, analysing massive datasets, and orchestrating multi‑step workflows through emerging agent‑like capabilities.

But as AI gets faster, smarter, and more capable, there’s one thing forward‑looking leaders recognise:

AI can analyse, automate, and accelerate – but only people can interpret, persuade, influence, build trust and set direction.

Competitive advantage is shifting from who has the most powerful model to who can pair it with the most human organisation.

The next decade will be won by organisations that are human‑led and AI‑enabled.

And at the centre of that model is a simple principle:

Human-in-the-Loop (HITL)

Humans set the goal.

AI does the heavy lifting.

Humans make the call.

Keep that blueprint in mind as we explore how work is changing.

AI Adoption is Moving Fast

AI is no longer confined to data teams or experimental pilots. It now appears across nearly every function:

• Customer service
• Finance and reporting
• HR and talent
• Marketing and product
• Operations, supply chain, logistics
• Audit and Assurance

At the same time, a new generation of agentic AI has emerged.

Systems capable of planning, reasoning, and executing multi‑step tasks end‑to‑end. These agents already play a visible role in finance and reporting, assisting with large‑scale data ingestion, validation, and analysis.

But for all this acceleration, there’s a deeper issue at play:

Doing more isn’t the same as doing what matters.

And that’s exactly where the human element becomes the true competitive advantage.

What AI Does Brilliantly (and What It Doesn’t)

AI excels at the kinds of work that once consumed human hours:

• Scanning millions of data points in seconds
• Identifying patterns and anomalies
• Performing multi‑step tasks via AI agents
• Drafting summaries, reports, and communications
• Automating repetitive or rules‑based tasks

But AI has hard limits:

• It doesn’t understand context
• It cannot navigate politics, culture, or relationships
• It cannot build trust
• It cannot make ethical trade‑offs
• It cannot persuade a hesitant executive

And it cannot look a stakeholder in the eye and say:

“Here’s what this means — and here’s what we should do next.”

The Human Element: The Five Capabilities That Matter Most

To thrive in an AI‑enabled world, organisations must invest not only in models, data, and platforms, but in people.

The following capabilities form the backbone of human‑centred AI adoption.

1. Critical Thinking

AI produces answers. Humans must decide whether those answers are correct, meaningful, or dangerous.

The NIST AI Risk Management Framework emphasises precisely this: human accountability, explainability, and oversight across the full lifecycle:

Govern – Map – Measure – Manage.

Micro‑habit: For any AI‑generated insight, require three plausible explanations and one recommended decision. If it can’t be justified, it shouldn’t be actioned.

• Empathy

AI can draft emails; it can’t read a room.

Understanding stakeholders’ goals, pressures, and language, is what translates technical capability into organisational movement.

The OECD AI Principles echo this, positioning AI as human‑centric and values‑driven.

Micro‑habit: Before key conversations, complete a quick empathy map: What does this leader care about? What risk or opportunity will resonate? What language will land?

• Influence & Communication

AI can draft a message. Only a person can deliver it with nuance and conviction.

Influence is the ability to turn insight into action.

Something every organisation needs as AI produces more data, more recommendations, and more “noise.”

Micro‑habit: Use a one‑page “So what? / Now what?” brief for any AI‑supported recommendation. If it can’t be explained crisply, it isn’t ready.

• Ethical Judgment & Courage

Courage is the willingness to challenge an algorithm’s output, pause an automation pipeline, or escalate when “something doesn’t feel right.”

The EU AI Act mandates human oversight for high‑risk systems and places accountability firmly with people,

Micro‑habit: Create visible override thresholds (when humans must step in) and celebrate constructive, well‑reasoned pauses.

• Creativity & Strategic Thinking

AI can generate ideas. Only humans can determine which align with purpose, values, long‑term goals, and societal impact.

OECD and NIST frameworks both reinforce the need for human‑led vision and future-oriented thinking.

Micro-habit: Begin each AI initiative with a decision memo: desired outcomes, metrics, guardrails, and clear stop conditions.

Human‑in‑the‑Loop (HITL): A Simple Operating Model

HITL keeps humans in charge across the lifecycle:

1. Define (Human): Objective, risk appetite, and success criteria.
2. Design (Human-guided): Data sources, privacy/fairness expectations, mandatory review points.
3. Deploy (AI): Process, generate, analyse.
4. Decide (Human): Review and approve high‑impact recommendations before action.
5. Monitor (Human): Track performance, bias, drift, and incidents; refine or roll back.

This mirrors NIST’s Govern–Map–Measure–Manage and complements prescriptive approaches like the EU AI Act’s risk‑based obligations.

The throughline: You don’t need to choose between speed and safety. The HITL model gives you both.

Why Soft Skills Deliver Hard ROI

Faster decisions: Clarity and trust collapse cycle time from insight to action.

Higher adoption: Empathy and influence reduce resistance and increase sustained use.

Lower risk: Critical thinking and override discipline prevent small issues becoming major incidents.

Stronger brand: Transparent communication builds customer and regulator confidence.

A Practical 60 Day Playbook as a Starting Point

Days 1–10: Inventory & Intent

• Map where AI already touches your workflows (including embedded AI inside your existing SaaS tool).
• For each use case, draft a short decision memo outlining the intended outcome, the metric that defines success, and the guardrails.
• Identify where human checkpoints are required before actions affect customers, compliance, or financials.

Days 11–30: Build the Habits

• Roll out the simple “So what? / Now what?” one‑pager for any AI‑assisted recommendation to strengthen reasoning and accountability.
• Train teams in:
  • The Three Explanations Drill (critical‑thinking muscle).
  • Empathy Mapping (clear, audience‑aligned communication).
• Define and publish override thresholds and escalation paths:
  • Positioning courageous, well‑reasoned human intervention as a control, not a failure.

Days 31–60: Measure & Refine

• Track:
  • Action Rate (how many recommendations lead to decisions within 30 days)
  • Time‑to‑Decision
  • A simple Message Clarity score from senior leaders
• As a leadership team, review incidents, overrides, or drift.
  • Strengthen guardrails using the NIST‑aligned “Measure → Manage” cycle.

Ongoing

• Continually train (and actively encourage) teams to challenge AI outputs.

A Global Landscape, A Shared Direction

Despite different regulatory approaches, global norms are converging around similar principles:

EU (EU AI Act 2024/1689): Rights‑preserving, human‑centric; transparency, safety, mandatory oversight

US (NIST AI RMF 1.0): Governance‑framework, risk‑management emphasis; HITL assurance.

OECD AI Principles (47+ jurisdictions): Human well‑being, fairness, and meaningful oversight.

Africa: Rapidly evolving; alignment with OECD and NIST norms to embed transparency, fairness, oversight.

Middle East (UAE): Innovation‑forward with ethical guardrails and accountable governance.

Cayman Islands/Caribbean: Principles‑based alignment with international expectations, especially in finance.

Canada (AIDA): Risk‑based federal approach, broadly aligned with EU‑style principles.

UK: Sector‑led model; innovation‑friendly, non‑statutory principles.

Singapore: Model AI Governance Framework, globally referenced.

China: Algorithmic regulation (recommendation systems, deep synthesis), prioritising security and transparency.

Different frameworks, same message:

Design for trust. Keep people accountable.

Conclusion: The Future Is Human‑Led, AI‑Enabled

The goal is not to tame AI until it is harmless.

It is to design for trust so you can move faster.

Trust comes from clarity of purpose, human accountability, and communication that treats every stakeholder as an intelligent partner.

AI is not replacing humans. It is redefining what makes humans valuable.

The organisations that win will let machines do more of the heavy lifting, and people do more of the thinking, connecting, and deciding.