Orient launches meaning layer for enterprise AI systems

Orient has launched what it calls a Meaning Layer for organisations and AI systems. The Stockholm-based startup says the platform is designed to create a shared understanding between humans and machines.

The launch enters a growing debate over how businesses should manage AI systems that can generate content, make recommendations and, in some cases, take autonomous actions. Orient argues that, for many organisations, the main issue is no longer whether AI can produce answers, but whether those answers are based on information that is current, trusted and understood in the same way by staff and software.

More companies are preparing to use AI agents in day-to-day operations. Orient cited Gartner figures showing that 17% of organisations have already deployed AI agents, while more than 60% expect to do so within the next two years.

That trend has sharpened concerns about how AI systems interpret instructions and source material. In business settings, the issue goes beyond the risk of high-profile failures and extends to routine work such as drafting reports, assessing research, summarising meetings and supporting operational decisions.

Shared understanding

Orient says its system sits between an organisation's information and the actions taken by AI tools. Rather than treating individual documents as the sole source of truth, the platform is intended to build a live model of what a business knows, what it has decided, what remains uncertain and what can be acted on safely.

The system draws on internal sources including documents, meeting notes, presentations, spreadsheets, transcripts, messages and customer records. It links outputs back to underlying evidence and highlights whether information is current, disputed or incomplete.

Its approach reflects a wider problem facing large organisations as AI-generated material multiplies. Teams often work across emails, chats, slide decks, notes and automated summaries, leaving the reasoning behind decisions scattered across multiple systems.

That fragmentation can create gaps between what an organisation believes, what individual teams understand and what an AI tool retrieves. If an AI agent acts on an outdated presentation, an incomplete note or a disputed internal claim, the result may still appear coherent even if it is wrong.

Two products

Orient has split the offering into two parts. Orient for Humans is described as a workspace for teams to review information, compare evidence, identify contradictions and retain the rationale behind decisions. Orient for Machines is intended to give AI agents structured organisational context before they respond or act.

The company argues that AI systems often fail not because the underlying models are weak, but because they are working from fragments. In that view, the challenge for enterprises is less about producing more data and more about preserving meaning as information moves between employees and software.

Industry research has highlighted the scale of the problem. Orient pointed to a European Broadcasting Union- and BBC-led study that found 45% of AI assistants misrepresented source content, with 20% of outputs containing major accuracy problems, including hallucinated details.

It also cited Zapier research on so-called workslop, which found that 58% of workers spend more than three hours a week revising or redoing AI outputs before they can be used. A Stanford and BetterUp study of 1,150 US desk workers found that 40% had received AI-generated workslop in the past month.

Spinout roots

Orient was spun out of Swedish workplace learning platform Learnster and received about €500,000 in initial funding before becoming an independent company focused on enterprise AI infrastructure. The startup says its team has experience in reasoning systems, language models and agent-based software.

The company argues that a missing layer exists in enterprise software. Traditional systems have helped businesses create, store and retrieve information, while newer generative AI tools have made it easier to produce text and analysis at scale. But neither necessarily preserves why a decision was made or whether the evidence supporting it still holds.

For firms trying to move from AI assistants to more autonomous systems, that distinction matters. A tool that can answer a question is one thing; a system that can make or influence business decisions requires visibility into priorities, assumptions and exceptions that may never be clearly set out in one place.

Mikael Larsson, Chief Executive Officer and Founder of Orient, said the company believes AI has solved generation, but not understanding.

"AI has solved generation. It hasn't solved understanding. Organisations don't need another tool that produces more content. They need a way for people and AI to work from the same shared understanding. Enterprise AI does not fail because models lack intelligence; it fails because organisations lack shared understanding. We believe the next competitive advantage won't come from generating more information. It will come from preserving meaning as work moves between people and increasingly autonomous AI systems," said Mikael Larsson, Chief Executive Officer and Founder of Orient.