Key Takeaways
- Maks Giordano argues that AI should extend human capability without displacing human judgment, accountability, or democratic choice.
- Enterprise adoption is moving faster than governance, impact measurement, and workforce planning.
- Transparent oversight, clear decision rights, and meaningful routes for redress can reduce operational and public resistance.
After three decades of helping companies pursue faster and more creative routes into the future, Maks Giordano has turned his attention to a harder question: How much authority should people hand to artificial intelligence?
At TEDxBerlin, Giordano’s answer is grounded less in opposition to AI than in a distinction between capability and legitimacy. Systems can process information, generate content, recognize patterns, and recommend actions at a scale that people cannot easily match. They cannot independently establish the values that should govern those actions, nor can they accept social or legal responsibility when an outcome causes harm.
That distinction has immediate relevance for business leaders. OpenAI’s ChatGPT, Google’s Gemini, and Microsoft’s Copilot are becoming interfaces through which employees research subjects, draft communications, write software, and interpret organizational knowledge. Their outputs can shape decisions even when the systems do not formally make those decisions.
However, calling a product a “copilot” does not automatically keep a person meaningfully in command. Human oversight can become little more than a checkbox if employees lack the time, expertise, authority, or source information required to challenge an automated recommendation.
Adoption data suggests that the gap is widening. Forrester reported in its 2025 assessment that more than 70% of surveyed organizations had generative or predictive AI in production. Yet relatively few were measuring financial impact or operating with mature governance and workforce strategies. That creates a familiar enterprise problem: deployment is visible and fast, while responsibility remains scattered.
Trust figures add another wrinkle. University of Melbourne and KPMG research found that only 46% of people globally said they would trust AI systems, while 70% believed regulation was warranted. Separately, IDC and SAS reported that 78% of organizations said they fully trusted AI, but only 40% had invested enough in governance, explainability, and ethical safeguards to demonstrate that trustworthiness.
The mismatch matters. Confidence expressed by executives is not the same as evidence that a model performs reliably across different users, languages, regions, or operating conditions. Nor does internal enthusiasm settle whether customers, employees, regulators, and citizens consider a use legitimate.
In lower-risk settings, errors may mean an awkward email or a weak presentation. In hiring, lending, insurance, healthcare, public benefits, policing, or defence, the consequences can be much greater. A person affected by an AI-assisted decision may reasonably want to know what information was used, who approved the process, and how to contest the result.
These are not theoretical concerns. The OECD recorded nearly 8,600 reported AI incidents and hazards between January 2022 and August 2025, including more than 3,100 involving government, security, and defence. Its analysis also identified public resistance in 76% of government AI use cases across 11 policy functions.
Resistance, then, is not merely a communications obstacle. It can delay deployment, weaken adoption, generate legal exposure, and erode confidence in the institution using the technology. A technically accurate system may still face rejection if people see its process as opaque or its purpose as illegitimate.
For enterprises, Giordano’s argument points toward practical governance rather than broad restraint. Organizations can classify AI uses by potential impact, assign accountable owners, document data and model limitations, test outputs before deployment, and preserve escalation paths for disputed decisions. Employees also need room to disagree with a system without being penalized for slowing an automated workflow.
A small but important detail is procurement. Buyers often evaluate model accuracy, integration, security, and cost, yet give less attention to contestability and responsibility. Who investigates an incident? Can the organization explain a recommendation to an affected person? What happens when a vendor changes the underlying model?
AI can expand what people are able to do. Giordano’s warning is that expanded capability should not quietly become delegated authority. For business and government alike, the durable approach is likely to keep values, accountability, and final decision rights in human hands.
⬇️