Skip to main content
THEN

Human-AI Decision-Making: How to Make Decisions When AI Is Your Subordinate

Published: 18 Nov 2026 · Last updated: 18 Nov 2026 · Article by Lynn, Learning Hub Thailand Content Team

Hello, Lynn would like to start with the pressure that leaders face today.

AI helps brainstorm and propose options more quickly, so leaders are expected to make decisions faster.

Lynn likes to view this simply: think of AI as a highly capable, fast subordinate who has just joined the team. We can delegate tasks, ask it to gather information, and propose options, but we would never let a new team member make major decisions alone without review—and the manager remains accountable for the team's output regardless.

The core of this is that humans must remain in the equation. If we let AI take the lead without humans keeping up, we might gain speed but lose quality and trust. Today, Lynn will explore how leaders can make sound decisions alongside AI.

When AI assists in decision-making, who should have the final say?

Humans should have the final say, especially on critical matters. While AI excels at gathering information and proposing options, weighing values and taking accountability for outcomes remain human responsibilities. A survey by Resume Now in mid-2026 revealed that 97% of workers still rely on human judgement first—both their own and their colleagues'—before turning to automated systems.

Interestingly, when opinions between AI and colleagues conflict, most people still choose to trust humans. Resume Now found that 72% would follow their colleague, with only 28% following AI. People trust AI most for research and drafting tasks, whilst strategic and planning tasks receive the least trust. This reflects a clear understanding of where AI should assist and where humans should decide.

A sound principle is therefore not to let AI decide on our behalf, but to let AI propose data and options while humans make the final decision and take responsibility. Especially for high-risk decisions or those affecting people, a human must always remain at the final checkpoint.

When does AI improve decision-making, and when does it lead us astray?

AI is immensely helpful for tasks within its capabilities, but can lead people to make errors when tasks exceed them. Research by Harvard Business School in collaboration with BCG, surveying 758 consultants in 2023, found that for tasks within AI's capabilities, consultants using AI completed 12% more tasks, 25% faster, with 40% higher quality.

However, for tasks too complex for AI, the results reversed. For every 100 people, the group using AI answered correctly about 19 fewer times than the group that did not. Researchers refer to this as the 'jagged frontier', meaning AI's competence varies across tasks. Between two tasks of seemingly equal difficulty, AI might perform exceptionally well on one yet fail at the other, and this is not easily distinguishable from the outside.

In this study, the task AI handled well was generating new product ideas, whilst the task that led people astray was analysing sales channel and company brand data to recommend profit-enhancement strategies. The takeaway for managers is that a confident-sounding answer from AI does not make it correct. One must first verify whether the task truly falls within AI's strengths; the more a task requires connecting disparate data points or weighing human impact, the more humans need to think for themselves.

The traps of over-relying on AI in decision-making

The biggest trap is trusting AI to the point of ceasing to question it. When AI delivers a polished answer quickly, people tend to accept it without verification. The more frequently it is used, the more complacent people become, until one day they no longer apply their own judgement. This risk is the exact opposite of refusing to use AI altogether.

Figures confirm this is a real issue. A survey by Kolmogorov Law of 500 US workers using AI at work in September 2026 revealed that 65% do not verify AI outputs every time, and 42% admitted to using AI answers even when they suspected they were incorrect. Meanwhile, only 22% reported that their organisation had a written policy requiring AI outputs to be reviewed before being used in work.

The preventive measure is designing verification into the process itself, rather than leaving it to individual diligence. Data from Resume Now showed that 91% of workers expect disclosure when AI has assisted on a task to any degree. Making AI usage transparent and auditable therefore safeguards both decision quality and team trust. Training people to know when to trust AI and when to trust themselves is a core skill for decision-making in the AI era.

Recommended Courses

Human-AI Decision-Making

1-day In-house Training course · Managers, Supervisors

Practise questioning AI recommendations and verifying before trusting, ensuring humans remain the decision-makers and take ownership of outcomes.

View course details

Which tasks can AI decide, and which must humans decide themselves?

Use two questions, just as you would when delegating to a new team member: First, if the decision is wrong, is it easily reversible? Second, how significantly does this decision impact people or finances? The harder it is to reverse and the greater the impact, the more the decision must be retained by humans. Lynn breaks this down into three levels below.

Decision levelExamplesAI's roleHuman role
Easily reversible, low impactDrafting routine inquiry emails, prioritising daily tasksDecide and execute autonomouslySpot-check periodically
Reversible, but impacts team workflowSelecting new tools, adjusting workflowsPropose options with rationaleMake the decision and explain it to the team
Difficult to reverse or directly impacts peopleHiring, performance evaluations, restructuring, major budgetsAssist with data gathering onlyMake all decisions directly and take full accountability

This table aligns with what professionals already feel. Data from Resume Now showing that people trust AI significantly for information retrieval and drafting, but trust it much less for planning and strategy, reflects this exact tiering. The benefit of establishing clear rules like this is that teams do not have to guess every time how much they should trust AI, and managers know precisely what needs reviewing.

The Thai Context: Organisations are using AI, but decision-making rules are still lacking

Thai organisations are facing a concerning gap where AI adoption is outpacing governance. A survey by NECTEC of 574 Thai organisations under the AI Readiness 2026 project, as reported by Krungthep Turakij, found that 53.7% are already using AI, but 46.2% have not yet assessed risks or established AI governance, and 73.4% have not implemented AI ethics principles.

This means many organisations are already letting AI assist in thinking, but have yet to agree on who reviews, who decides, and on what principles those decisions are made. You can start addressing this immediately without waiting for major system-wide rollouts. Team leaders simply need to define clearly which tier of the table above each type of decision falls into, providing a practical foundation for governance within the team.

A Workplace Scenario: The unquestioned AI report

Lynn would like to share a common scenario drawn from observations across multiple organisations, rather than any single individual.

One team used AI to summarise quarterly sales and suggest which advertising channels should have their budgets reduced. The report was neatly presented with comprehensive figures; everyone in the meeting agreed and swiftly approved it. Several weeks later, they discovered that the input data had entirely omitted distributor sales, meaning the channel that had its budget slashed was actually their top revenue generator.

High-performing teams use AI in the same way, but the difference lies in the leader. Before approving, the leader always asks three questions: Is the data AI used complete? What are the consequences if this recommendation is wrong? And who on the team knows this topic best and should have the chance to challenge it first? These simple questions allow the team to maintain AI's speed whilst keeping humans firmly in charge of the decision.

Reflection Questions

Take a moment to examine decision-making within your organisation:

  • In key organisational decisions, are humans still the final decision-makers, or are you beginning to let AI take the lead?
  • Do team members feel confident questioning AI's outputs, or do they simply accept them because they are fast and well-presented?
  • Does your team know which tasks play to AI's strengths, and which exceed its capabilities?
  • Are there clear rules defining which decisions AI can make independently, and which require human review?
  • If an AI recommendation proves wrong, does the team know who is held accountable?

Next Action: Getting Started Today

If you want to make better decisions alongside AI, Lynn recommends starting with small, actionable steps:

  • 👉 Try defining clearly within your team which tasks AI can propose, but where humans must always review and make the final call. This simple step ensures that people remain in the decision-making equation.

Lynn believes that great leaders in the AI era are not those who make decisions the fastest, but those who decide in a way that keeps their people aligned and trusting. 💛

Assess the readiness of leaders in your organisation

Before planning your decision-making approach, Lynn would like to invite you to explore how your organisation's leadership development system currently stands.

Learning Hub Thailand offers the Thailand LDM Assessment to help you see where your organisation's leadership and people development systems stand. It takes around 5 minutes; try it at the Thailand LDM Assessment to identify key areas for enhancement first.

Develop leaders to make decisions alongside AI

Decision-making alongside AI is an essential modern leadership skill. Learning Hub Thailand offers courses designed to develop this capability.

Frequently Asked Questions

Decision-making in the AI era: who should have the final say?

Humans should make the final call, particularly on decisions that are critical or affect people. While AI excels at gathering data and proposing options, weighing values and taking accountability for outcomes remain human responsibilities. Data from Resume Now shows that 97% of workers still value human judgment over AI, and 72% choose to trust colleagues over AI when opinions conflict.

What types of tasks can AI decide autonomously?

Tasks where mistakes are easily reversed and have minimal impact on people, such as drafting general enquiry emails or prioritising daily tasks, can be decided by AI with periodic spot checks. However, for tasks that are difficult to reverse or directly impact people—such as hiring, performance appraisals, or major budgets—humans must make the decision, using AI solely to gather data.

Can AI make decision-making worse?

Yes, if applied to tasks outside AI's capabilities. Research by Harvard Business School in collaboration with BCG found that for tasks within AI's capabilities, people using AI produced work of 40% higher quality. However, for tasks that were too complex for AI, the AI-assisted group answered correctly about 19 fewer times per 100 compared to those who did not use AI. Leaders must therefore assess whether a task truly falls within AI's strengths before trusting its recommendations.

What is the pitfall of using AI to assist decision-making?

It is over-relying on AI to the point of stopping questioning it. When AI provides an answer that looks plausible and arrives quickly, people tend to accept it without checking. The more frequently it is used, the more people trust it, ultimately bypassing their own judgment. The prevention is to design verification and disclosure of AI usage directly into the workflow, rather than leaving it to individual discretion.

How can leaders train their teams to make decisions alongside AI?

Start by establishing clear guidelines on which decisions AI can make autonomously, which ones AI proposes while humans decide, and which ones humans must decide entirely on their own. Then, train the team to ask key questions before accepting AI recommendations: Is the data complete? What are the consequences if it is wrong? Who should verify it? Additionally, clearly disclose when AI is used—something employees already expect, with Resume Now survey findings showing that 91% want transparency around AI usage.

References