News

Almost every leader today wants to see an instant productivity leap from adopting artificial intelligence (AI). But the reality is harsh: research and interviews show that roughly 80% to 90% of companies see no tangible return or benefit from AI adoption. The biggest mistake? Companies still treat AI as a pure IT project, completely ignoring business architecture and the human factor.

To look for solutions to this challenge, AI Master Lab brought together a panel of experts: change management expert and organizational psychologist Iveta Bikse, international AI trainer and HR expert Dov Zavadskis, and AI Master Lab partner and business strategist Guntis Kalniņš.

In this article, we've gathered the discussion's key points and practical steps for moving companies from "chaotic experimentation" to profitable, safe AI integration.

1. "Shadow AI" and security risks: what employees are hiding from management

The numbers are striking – around 78% of office employees already use generative AI daily, yet nearly half of them hide it from management. This phenomenon, known as shadow AI, creates enormous risks for companies.

• Uncontrolled tools and data leaks: Employees install various free tools and browser extensions and connect them to company files (such as Excel spreadsheets) without reading the terms of use. This creates a direct threat to the security of sensitive data.
• "Quiet promotion": If an employee gets work done faster thanks to AI, they often fear telling their manager. Why? Because the "reward" for efficiency is usually just a bigger workload, with no pay rise.
• Blind trust in hallucinations: Working under time pressure and without clear guidelines, employees rely completely on AI output, ignoring these systems' tendency to "hallucinate," or make up facts. This can lead to costly mistakes that companies pay for with their reputation.

The solution: Company leadership needs to create a clear AI policy and AI guidelines. Banning tools outright makes no sense. What's needed is a definition of which tools are safe, what data may be entered into them, and how to verify the quality of the output.

2. The middle-manager trap and "techno-stress"

Middle managers currently find themselves caught in the crossfire. On one side, senior leadership demands efficiency and results. On the other, quiet fear and resistance run through the team, as employees worry about losing their jobs.

This creates a new phenomenon – techno-stress and burnout risk. Employees and managers feel confused, unsure what will be expected of them tomorrow. As our experts point out, the "psychological contract" between employee and company is currently being rewritten, and many find that unsettling.

How can a manager get out of this trap?

• Cut back on micromanagement and focus on people: AI is excellent at structuring information and passing on messages, but it cannot offer human support. A manager's greatest value today is the ability to talk about the topics no one else raises, and to help the team adapt.
• Define what WE stop doing: Efficiency doesn't mean working 20 hours a day. It's the manager's job to help employees understand which processes to hand over to AI, and how to put the time gained to good use.
• Be like children – experiment: In this fast-moving period of change, no one is an absolute expert. The best managers are the ones willing to show vulnerability, admit they don't have all the answers, and learn alongside their team.

3. HR transformation: from administrative support to AI champions

The HR department has historically been seen as one of the less technology-driven functions. Yet it is HR that now needs to become the bridge between IT technology and people.

• Change management: Technology rollout accounts for only 10% of success, data for 20–30%, and the remaining 60–70% comes down to people and organizational culture. If HR doesn't get involved in change management, any system rollout will fail.
• Synergy with IT: The HR of the future is no longer just "personnel administration." HR needs to work hand in hand with the IT department. In the US, roles such as Talent Engineer are already emerging – technical specialists who work specifically on talent acquisition and HR data analytics.
• Finding internal "champions": Instead of relying only on outside consultants, HR's job is to identify the company's own AI enthusiasts. Give them a platform (for example, 15 minutes in the weekly meeting) to share real-world examples and inspire the rest of the team.

Watch the full panel discussion

We've added Latvian subtitles to the recording. Watch the full conversation to learn more about:

• Why employees who work with AI in the US earn up to 50% more
• How to tell work done "with AI" apart from work done entirely "by AI"
• What awaits us in the age of autonomous AI agents, and why the Baltic states need to move fast to avoid losing global competitiveness

While our webinar on implementing AI guidelines covered "shadow IT" at length, our experts' conversations with clients reveal three specific problems that go beyond the usual, simplified "allow/block" approach.

Here are three strategic and technical insights from a conversation between our experts, Vija Kalniņa and Jānis Judrups, that will change how you think about security.

1. The "free cheese" trap: model training

The problem isn't that employees use AI – it's which version of it they use.

In the webinar, we highlighted a key distinction:

• Free tools (Public AI): If your lawyer pastes a contract into the free version of ChatGPT or DeepL, that text is most likely stored and used to train the next version of the model. In theory, there is a risk that your organization's confidential information could one day surface as an answer to another user.
• Enterprise / Team plans: In paid versions (API or Team plans), the settings typically ensure that your data isn't used to train the underlying models.

The takeaway: It's not enough for your guidelines to say "use tool X for drafting English-language documents." There needs to be a clear line: public versions of tools may only be used for general information, while sensitive data requires a sandboxed environment that doesn't pass data on further.

2. The "traffic-light principle" for data classification

"How do we control all of it?" is one of leaders' biggest fears. The answer: an organization doesn't need to control the use of the tools themselves – it needs to control which data goes where.

We recommend introducing a simple "traffic-light principle" for employees:

This approach takes the pressure off employees – they don't need to be IT experts, they just need to understand what kind of data they're working with.

3. Human in the loop

Vija Kalniņa highlighted a critical ethics and safety point that often gets forgotten amid the excitement of automation.

No AI decision that affects people (such as candidate selection, benefit approval, or legal analysis) should be fully automated. Guidelines must include the principle that AI is an assistant, not the decision-maker.

Final responsibility always rests with the employee. That means "copy-pasting" a chatbot's output without reading it back and checking the facts counts as a breach of work discipline.

In November 2025, the new "Practical guidelines for implementing artificial intelligence in public administration" were published. This is a long-awaited and important step toward bringing order to a landscape where, according to the data, 65% of public institutions already use AI solutions – often, unfortunately, without a clear strategy or security framework.

This document is a huge step forward for putting Latvia's AI landscape in order. But publishing it is only the beginning of the work.

The challenge: from paper to process

Institution heads, IT managers, and lawyers now face the next critical task: integrating these general guidelines into their organization's specific processes.

The key question every leader faces is: how do you put the necessary controls and security in place without turning everyday work into something bureaucratic, slow, and complicated? AI's potential lies in efficiency, and overly heavy-handed regulation can kill that potential at the root.

From theory to safe action

To help state and municipal institutions learn about and put the new requirements into practice, AI Master Lab hosted an exclusive webinar. It wasn't a theoretical lecture, but a conversation with the experts who actually wrote the content of these guidelines – Jānis Judrups, Dr. sc. ing., and Vija Kalniņa, Dr. iur.

What we covered

We focused on three practical pillars of safe AI governance:

• Building an "allow list": How to practically audit tools, define "red lines," and create a list of AI tools approved for use at your institution, so employees clearly know the rules of the game.
• Risk assessment: How to use the decision trees included in the guidelines to determine, within five minutes, whether a given tool is safe and what requirements apply to it under the AI Act.
• AI culture and people: How do you turn employees into allies for security? On the role of "AI champions," and how to build an environment where innovation isn't hidden ("shadow IT") but actively managed.

Artificial intelligence solutions are no longer just a tool for innovation labs – they've become part of everyday decision-making, data processing, and customer service. But with new opportunities comes a new kind of responsibility. Where does innovation end and a breach of regulatory requirements or ethical boundaries begin?

In the AI Master Lab webinar "AI risk management in organizations: safe and trustworthy AI use – a legal requirement or an ethical duty", Vija Kalniņa and Ivars Krampis-Vanags helped explain how to turn AI risk into a management-level responsibility, rather than leaving it solely to IT or legal.

What we discussed

• AI risk in organizations: how it shows up in practice, and what incidents have already occurred in business settings.
• Accountability: who bears responsibility for the risks created by AI tools – developers, users, or the state?
• Regulatory framework: how the EU AI Act helps or complicates risk management.
• Ethics in practice: how to determine what counts as acceptable versus risky use of AI.
• Where to start: practical recommendations for organizations just taking their first steps in AI adoption.

Speakers

Vija Kalniņa – Doctor of Law, an expert in AI ethics and law, with research experience at Harvard and Boston University. Her work helps organizations adopt artificial intelligence with clear accountability and transparency.

Ivars Krampis-Vanags – a data analytics expert, founder of Excel Know How, and a Microsoft Power BI practitioner with extensive experience in finance and data management.

LIAA (the Investment and Development Agency of Latvia) has launched an ambitious artificial intelligence (AI) support programme for small and medium-sized businesses. It's the first initiative of this scale in Latvia, and it has already generated considerable interest among business owners.

AI Master Lab partners Guntis Kalniņš and Ivars Krampis-Vanags shared insight in the webinar on how to prepare wisely for an application, taking into account both the programme's conditions and the scale of investment planned.

Watch the webinar recording →

How to approach participation in the LIAA AI support programme

• The programme is generous – but not simple. More than €18 million is available through 2029, but applying isn't a copy-paste exercise. It calls for a considered, strategic approach.
• Organization-wide, not personal productivity. The programme isn't meant for individual AI tools or GPT licences. It's about processes at the whole-company level.
• One shot, one project. The LIAA programme currently supports only one application per company. If it fails, there's no second chance. That's why it's critical to test the idea in practice first.
• Don't rush the application. As Guntis puts it: "Don't jump in just because the money is there." Start with preparation and experimentation.
• Experimentation is non-negotiable. Most AI projects fail because of overly high expectations and insufficient data. Experimental sprints help validate an idea before you commit major resources.
• Maturity level determines success. A company's AI maturity – not its general digital maturity – is what decides the outcome. Without prior experience or experimentation, the odds of success are only 5–15%.
• Focus, not generalities. An AI project needs to be specific, with measurable results and a clear business goal – not "improve everything."
• A good idea isn't enough – you need data. Even a brilliant idea won't survive without quality, well-organized data. Data preparation is often one of the hardest (and most expensive) stages of getting ready.
• Apply only once you're ready. Strategy → Ideation → Experimentation → Application. This is the recommended order if you want to improve your odds of securing support.

On April 3, AI Master Lab held the webinar "The evolution of artificial intelligence: what's changed in a year?", where we looked at how the AI landscape in Latvia and around the world has shifted over the past year. Our goal wasn't just to explain the technological changes, but to dig into what they mean for businesses, public administration, and practitioners.

During the webinar we covered examples of generative AI in use, as well as the most common challenges organizations face when trying to integrate these tools into daily operations. Attendees heard real examples from Tilde's experience, an analysis of EU regulation, and advice for companies on how to develop an AI adoption strategy.

Watch the webinar recording →

Key takeaways

• The availability of AI solutions is growing fast; companies are experiencing "FOMO" – the feeling that they can't afford to fall behind.
• Many companies start with tools that help with everyday tasks: reworking text, meeting notes, translation.
• Technology is often "rolled out" without a clear purpose or a real problem to solve.
• There's a lack of AI usage strategy and understanding of data quality.
• The EU AI Act brings many requirements – ethics and data protection need to be considered from day one.
• It's best to start with a small, clearly defined problem – not with "big AI."
• Future competitiveness will depend on the ability to integrate AI wisely, not just experiment with it.