Shift2new at OEB Global 2026

Herculano Rebordão brings a real case of reliable AI in professional training to Berlin

There are conversations about artificial intelligence that stop at the promise — and there are conversations that start from practice. It is the latter that Shift2new brings to OEB Global 2026 in Berlin, through a presentation by Herculano Rebordão, CEO of Shift2new, regarding one of the most critical — and least honestly discussed — issues in applying AI to professional training: how to ground AI in reliable knowledge .

OEB Global, held annually in Berlin since 1995, is the most established and internationally respected European conference on digital learning and technology-supported training. The 2026 edition takes place from December 2nd to 4th, at the InterContinental Berlin , gathering more than 2,000 participants from over 70 countries and more than 270 speakers from all sectors of education and training — universities, governments, companies, startups, and international organizations.

For Shift2new, having a presence on the OEB stage is not just participating in an event. It is an opportunity to contribute to the global conversation about the future of learning from a concrete perspective, rooted in real implementation experience in contexts where information accuracy is non-negotiable.

---

The 2026 theme: "Sovereign Learning" and the issue of trust

The 2026 edition of OEB takes place under the theme "Sovereign Learning: Trust, Agency & the Education Reset" .

OEB’s own framing is clear: artificial intelligence has ceased to be an emerging technology to become a force that is redefining how knowledge is produced, verified, and applied. The question the conference poses is not if AI will transform education and training — that is taken for granted — but how to ensure that this transformation equips people to operate and govern intelligent systems responsibly.

It is precisely at the intersection of trust (can we trust what AI produces for training purposes?) and agency (how do we keep the learner in control of their learning?) that Shift2new's contribution to this year's program is situated.

---

The panel: Future Workforce Learning, AI Augmented Capability & Skills Intelligence

Herculano Rebordão's presentation is part of a benchmark panel on the future of learning at work:

"Future Workforce Learning, AI Augmented Capability & Skills Intelligence"

The panel explores how artificial intelligence is transforming how organizations understand, develop, and apply the skills of their workforce — addressing practical models of continuous upskilling, human-AI collaboration, and learning integrated into daily work. The focus is on creating future-ready environments where technology accelerates human capabilities rather than replacing them.

It is a panel with an international and multidisciplinary composition, bringing together perspectives from academia, research, and practice:

- Herculano Rebordão — CEO, Shift2new (Portugal) - Eran Gal — Head of Graduate Program, Faculty of Instructional Technologies, Holon Institute of Technology (HIT) - Funke Saraki — Researcher, Ballsbridge University - Katarzyna Kowalik — Learning & Content Strategist, Independent Expert

---

Shift2new's presentation: grounding AI in reliable knowledge

Within the panel, Herculano Rebordão's presentation is titled:

"Grounding AI in Lifelong Learning: A RAG-Based Assistant for Knowledge-Intensive Workplace Training — Design, Implementation and Early Evidence"

The central question the presentation poses is direct: can RAG-based AI make workplace learning more reliable? And the answer does not stem from theory — it comes from a real implementation case, including the design that supports it, the approach to response accuracy, and the first evidence of impact on learning that the implementation is allowing to gather.

Why RAG matters

Most applications of generative AI to training face a structural problem: language models, by default, generate plausible but not always correct answers — the phenomenon commonly known as "hallucination." For entertainment or brainstorming content, this imprecision is tolerable. For knowledge-intensive professional training — and particularly in regulated sectors such as banking, insurance, healthcare, or maritime contexts — it is unacceptable. A learning assistant that invents a compliance procedure, a safety regulation, or a technical data point is not a useful assistant: it is a risk.

RAG (Retrieval-Augmented Generation) is the technical approach that directly addresses this problem. Instead of letting the AI model answer only from what it "learned" during its training, RAG forces it to ground every answer in specific, controlled documentary sources — an organization's own manuals, regulations, procedures, and validated content. The result is an AI whose answers are traceable, verifiable, and anchored in trusted knowledge.

But — and this is where practical experience differs from theory — implementing RAG well is not trivial. The quality of answers depends on design decisions that go far beyond just "turning on" the technology: how to structure the knowledge base, how to ensure the precision of information retrieval, and how to design the experience so that the learner maintains control and critical thinking. It is these decisions — and what is learned from making them in practice — that the presentation focuses on.

Design, precision, and learner agency

The presentation is organized around three axes:

Design — the architectural decisions that determine whether a RAG implementation produces a reliable learning assistant or a sophisticated source of misinformation.

Precision and early evidence — how reliability is measured, validated, and guaranteed in a context where errors have real consequences, and what early evidence the implementation is allowing to gather.

Learner agency — the axis most aligned with OEB's "Sovereign Learning" theme: how to design the experience so that AI reinforces the learner's autonomy and critical thinking instead of replacing them. An AI that always provides the answer can, paradoxically, make the learner more passive. The issue is not just technical — it is pedagogical.

---

Why this matters for organizations

The relevance of this presentation goes far beyond academic interest. For any organization considering applying AI to its training — and there are more and more — the issues addressed in the session are exactly those that determine the success or failure of these projects.

The question most organizations ask is "which AI tool should we use?". The right question, and the one Shift2new's presentation poses, is "how do we ensure that the AI we use is reliable, grounded in trusted knowledge, and respects the autonomy of the learner?". The difference between the two questions is the difference between an investment that generates value and one that generates risk.

This is, in fact, the approach that defines Shift2new's work: technology is the means, not the end. The value lies in the method, the rigor of implementation, and the knowledge of how to combine tools at the service of concrete learning objectives — especially in sectors where reliability is non-negotiable.

---

Session details

Event: OEB Global 2026 Location: InterContinental Berlin, Germany Panel: Future Workforce Learning, AI Augmented Capability & Skills Intelligence Format: Presentation Panel Date and time: Friday, December 4th, 2026, 12:00–13:15 (Berlin time) Room: Tiergarten Speaker: Herculano Rebordão, CEO of Shift2new Presentation: Grounding AI in Lifelong Learning: A RAG-Based Assistant for Knowledge-Intensive Workplace Training — Design, Implementation and Early Evidence

---

We will be in Berlin — come talk to us

If you will be at OEB Global 2026, this is an open invitation. In addition to Friday's presentation, Herculano Rebordão will be available during the three days of the event to discuss the challenges and opportunities of applying AI to professional training — whether regarding RAG implementation, modernizing e-learning production chains, or specific training challenges in regulated sectors.

If your organization is thinking about how to apply AI to training reliably and methodically — rather than by trial and error — Berlin is a great place to start that conversation.

To schedule a meeting during OEB, or to learn more about Shift2new's work in this area, contact us at www.shift2new.pt

See you in Berlin.