AI, Data and Decision-Making: Why Critical Judgment Matters More Than Ever

For more than a decade, many organisations have treated access to information as the key capability in data-driven work. In reality, access is only the starting point. The more valuable skill is the ability to recognise when information is inaccurate, incomplete or misleading.

The widespread adoption of artificial intelligence makes this skill even more critical.

Since 2023, when the discussion around AI literacy became more active, one of the common arguments was that the technology was not mature enough for such a conversation. Today, that argument no longer applies. AI can generate answers, summaries, recommendations, code and analyses, and can increasingly trigger concrete actions.

The question for leaders is no longer only what it can do. It is shifting from “What can AI produce?” to “What happens when AI produces something that looks authoritative, sounds confident and still turns out to be wrong?”

The major difference and future competitive advantage will belong to companies that are able to identify such errors and gaps before they affect decisions.

Ten years ago, this was called data literacy. Before the boom of digital technologies, it was called critical thinking.

Every technology wave highlights a different capability. The internet enabled faster access to knowledge. Search engines developed the ability to collect and synthesise information. AI is now automating synthesis itself.

This means that the next competitive advantage will belong to people and organisations that can assess AI-generated content and decide what to trust, what to question and what to reject.

This is human critical judgment.

It is not developed simply by using AI more often. It comes from deep knowledge of a specific topic, enough to recognise a wrong answer even when it appears correct.

Context and judgment are not the same thing

Context is what we provide to AI: the data it can access, the business logic it can interpret and the parameters that shape its output.

Critical judgment is what people bring when AI makes a mistake. One is a data-related challenge. The other is a leadership issue. Both are essential for the development and stability of a modern business.

Read more: Business Intelligence software- what, for whom and why

The data shows an uncomfortable picture

According to the June 2026 BCG global survey on AI at work, 74% of frontline employees are already regular AI users, while 41% report higher cognitive load despite productivity gains. 

It is important to underline that AI does not automatically reduce cognitive effort at work. For many employees, it can increase it, while at the same time reducing the time available for evaluation, verification and critical thinking.

Similar conclusions appear in other research as well. The study by Dell’Acqua and colleagues, conducted with BCG and published through Harvard Business School research, introduced the concept of the “jagged technological frontier” – the uneven boundary between tasks where AI performs well and tasks where it may fail or hallucinate.

The main risk is that there is no clear signal showing on which side of this boundary a particular task falls. In the research, experienced employees with access to AI performed worse in identifying incorrect answers not because they were careless, but because the output looked plausible.

None of this is an argument against using AI. On the contrary, it is a reason to adopt a more deliberate approach to how AI is used, what standards apply and what is expected from the people working with it.

The risk for leaders is not simply that teams use AI too much or too little. The greater risk is that organisations build AI-dependent processes managed by people who are no longer sufficiently prepared to detect errors.

An AI result can be confident, well formulated and still wrong. At organisational scale, this is not only an operational issue. It becomes a strategic risk.

Read more: 5 Benefits for Companies of Working with a Business Intelligence System

Individual judgment does not scale. Decision architecture does.

Even a leader with strong critical judgment cannot manage or participate in every process. Organisations that will stay one step ahead will not simply be those where every employee has excellent AI judgment. They will be the ones that have built institutional mechanisms through which good judgment can be applied at scale.

Тhis requires a specific decision architecture with clear ownership of AI-generated outputs, review thresholds based on the possible consequences of an error and accountability structures that embed judgment into the process instead of relying only on individual vigilance.

Finance teams spent decades building controls around spreadsheet outputs precisely because they look authoritative, are easy to trust and can spread errors invisibly at scale. AI outputs have the same characteristics, but with much greater speed and volume.

Most organisations have not yet built equivalent controls. They are still at the stage where they assume that if someone has looked at the output, it has been evaluated well enough. This is not a governance model. It is an optimistic assumption.

Read more: UBB Interlease digitizes its reporting with the Qlik Sense BI system [interview]

The management conversation is still catching up

AI governance is not developing at the same speed as AI implementation. AI assistants and agents are already present in more workflows than many management teams realise: supplier selection, purchase requests, pricing actions, regulatory responses and many other areas.

Management teams often focus mainly on the capabilities of artificial intelligence. Yet some of the most important questions remain concrete and operational:

This is the process audit that organisations need to conduct. It clearly reveals the gap between the current management model and the necessary, safer level of AI control.

Critical judgment fails when the data fails

Critical thinking should be applied not only at the level of the final output, but also at the level of the data itself.

An AI system built on incomplete, inaccurate or unstructured data will produce confident and plausible conclusions without a real validation of their quality. The model cannot automatically signal every inconsistency in the data because it works with what it has been given and does not have a view of the full business reality. It does not know what it does not know. That remains a human responsibility.

Leaders who invest mainly in the capabilities of AI models, but not enough in the reliability and practical usability of their data, are building an unstable foundation for management decisions. The model may operate exactly as designed, but if the data is not correct and the organisation cannot trust it, the risk of errors becomes significant.

Data needs to be current and to have clear provenance, so that when a model gives an answer, someone can trace what it is based on and whether the source is still valid.

Ownership of data must also be clearly defined, with a responsible party for potential gaps, rather than looking for someone to blame only after an incident occurs.

It is also essential that data is properly structured for the AI model and not simply transferred from the respective source without considering the use case.

Three actions leaders should take now

From Data to Decisions: How Balkan Services Can Help

At Balkan Services, we help organizations turn data into a reliable foundation for decision-making. As an IT consulting company specializing in business software, data management, analytics, BI and digital transformation, we connect systems, improve data quality, and build environments where information is reliable, accessible, and actionable.

Balkan Services has been implementing business software solutions since 2006 and has completed more than 750 projects, including over 420 in the field of BI systems. Our team follows a proven implementation methodology with clear steps and practical know-how for best practices.

As a long-standing official Qlik partner in Bulgaria with Authorized Reseller status and the market leader in BI solutions in Bulgaria with a 30% market share, we help companies build modern data and analytics environments that provide visibility across the business, support faster decisions and create a stronger foundation for AI initiatives. We believe that successful AI starts with trusted, well-structured and accessible data, not with technology alone. 

For companies that want to use AI, BI and analytics effectively, the foundation is critical: reliable data, clear ownership, connected systems and solutions that support real business decisions.

Whether your goal is better reporting, stronger governance, advanced analytics or preparing your data landscape for AI, our team combines business, technology and industry expertise to help you move from information to insight and from insight to action.

In summary

The organisations that will be ahead in the coming years will not simply be those with the most advanced AI or those that implement AI tools quickly and frequently. The leaders will be the companies that know where human judgment must remain, how it should be applied and what data foundation it needs to rely on.

AI can generate an answer. But without a good foundation, that answer will not necessarily be accurate or worthy of trust.

This remains a human decision. And it starts with critical judgment.

At Balkan Services, we have expert knowledge of business, technology and legislation, and we speak all three languages. We will listen carefully to you and advise you on choosing the right business system for your needs

Source: Qlik’s Blog