{"slug": "more-than-a-quarter-of-belgian-companies-do-not-open-the-door-when-the-ai-agent", "title": "More than a quarter of Belgian companies do not open the door when the AI agent calls", "summary": "Only 2 of 245 tested Belgian and Dutch organizations are fully \"AI-agent-proof,\" and more than one in four is completely invisible to an AI agent, according to the Agentic Readiness Index 2026 published by Ghent-based technology company In The Pocket. The index found 69 percent of organizations are findable by an agent and 60 percent are readable, but just 7 percent let an agent complete a task, with 55 percent of blockages caused by technical issues such as JavaScript-only sites and 43 percent by deliberate barriers like logins, identity checks, or firewalls. In The Pocket COO Hannes Van de Velde said the shift \"from a chatbot to an autonomous, acting agent is accelerating at a pace that most Belgian and Dutch companies and (government) organizations cannot keep up with today.", "body_md": "**Many Belgian and Dutch companies are unprepared for an unexpected visit from an AI agent to their website. Should we let the agent in, and how?**\n\nMore than one in four Belgian companies is invisible to an AI agent, according to the *Agentic Readiness Index 2026* by the Ghent-based technology company In The Pocket, known for the co-development of Itsme and Bancontact Pay. Only two out of 245 tested organizations are fully *‘AI-agent-proof’*. And more than one in four is completely invisible to an AI agent.\n\n## AI agent as a new customer\n\n“With the launch of Muse and Astra, we see how two of the world’s largest tech players are preparing the mass market for an autonomously acting agent. This confirms what our Index is already measuring: the shift from a chatbot to an autonomous, acting agent is accelerating at a pace that most Belgian and Dutch companies and (government) organizations cannot keep up with today,” says Hannes Van de Velde, COO at In The Pocket.\n\nIn the future, consumers will increasingly not act themselves. It is AI agents that perform tasks on behalf of the consumer. Imagine: you soon ask an AI agent to compare your energy contract and switch immediately. The agent sets out, knocks on the door of a series of Belgian companies, and in more than a quarter of cases, encounters a closed door. Of the 245 tested organizations, exactly two allow an agent to fully complete a task.\n\n## Findable, but not executable\n\nIn The Pocket released AI agents on 245 large companies and public service providers in Belgium and the Netherlands, spread across retail, finance, insurance, telecom, energy, HR, healthcare, public transport, and government. Each agent was given realistic consumer tasks: comparing a rate, booking an appointment, completing a purchase, or switching providers. Scoring is based on three levels: findability, readability, and executability.\n\nA sharp waterfall effect occurs on those steps. 69 percent of organizations are findable for an agent, 60 percent are also readable, but barely seven percent actually allow the agent to complete the task. Being found is therefore not the problem. Getting something done is.\n\n## Three levels\n\n*The Agentic Readiness Index* by In The Pocket divides organizations into three profiles. *Invisibles* simply do not exist for an agent. *Informers* are visible and readable but offer no path to action: brochureware for a visitor who actually wants to act. And then there are the *Actors*, where the agent does bring a task to completion. That group currently counts two names out of 245.\n\nThe bar is therefore high. [Beyond the initial hype](https://itdaily.be/blogs/innovatie/agentic-ai-de-hype-voorbij/), the major platforms have set the direction: OpenAI already allowed ChatGPT last year to [compare products and bundle them into a personal buying guide](https://itdaily.com/news/software/openai-chatgpt-shopping-research/), following earlier [an agent that performs tasks independently](https://itdaily.com/news/cloud/openai-launches-ai-agent/). Meta launched Muse earlier this month (not yet in Europe for now), an assistant that sends emails, fills out forms, and makes payments on behalf of the user.\n\nMcKinsey places a market size of 3,000 to 5,000 billion dollars on that agentic commerce by 2030. Gartner predicts that by 2028, nine out of ten B2B purchases will take place via an AI agent, and that by 2029, agentic AI will independently solve eight out of ten common customer service questions without human intervention.\n\n## Unintentionally invisible\n\nMore interesting than the scores themselves is the cause behind them. In 55 percent of cases, the blockage is purely technical. A site built entirely via JavaScript cannot be read by an agent: the same issue that has plagued search engine crawlers for twenty years, now with higher stakes.\n\nIn 43 percent of cases, it is a conscious choice, with a login, identity check, or firewall as a barrier. That ratio is the real news. The majority of organizations that are not ready did not choose to be so. It has to do with technical backlog, and that can be overcome.\n\n## Not every closed door is a mistake\n\nA caveat should be added immediately. Executability is not a neutral score. Gartner advised companies at the end of last year to [block agentic browsers for the time being](https://itdaily.com/news/cloud/ai-browser-too-risky-for-business/), due to indirect prompt injection, phishing, and autonomously placed orders. Those who sell insurance or convert energy contracts have good reasons to keep a human in the loop. And the question of [who decides exactly what an agent can do with customer data](https://itdaily.be/nieuws/business/ai-agents-krijgen-toegang-tot-je-klantdata-wie-bepaalt-wat-ze-mogen/) remains unanswered in most organizations.\n\nA low score on executability is therefore not by definition a shortcoming. But those who did not choose it consciously have also never defended it. That distinction, between a closed door and a door that no one knew was locked, is exactly what the index reveals.\n\nAccording to Van de Velde, the arrival of AI agents will be as crucial a turning point as the arrival of the internet. Those who do not prepare for it now, he says, are not only leaving revenue on the table but are also excluding a portion of their audience.", "url": "https://wpnews.pro/news/more-than-a-quarter-of-belgian-companies-do-not-open-the-door-when-the-ai-agent", "canonical_source": "https://itdaily.com/blogs/software/more-than-a-quarter-of-belgian-companies-dont-open-the-door-when-the-ai-agent-rings-the-doorbell/", "published_at": "2026-09-15 06:18:52+00:00", "updated_at": "2026-09-15 08:03:05.879427+00:00", "lang": "en", "topics": ["ai-agents", "ai-products", "artificial-intelligence"], "entities": ["In The Pocket", "Hannes Van de Velde", "Agentic Readiness Index 2026", "OpenAI", "ChatGPT", "Meta", "Muse", "McKinsey"], "alternates": {"html": "https://wpnews.pro/news/more-than-a-quarter-of-belgian-companies-do-not-open-the-door-when-the-ai-agent", "markdown": "https://wpnews.pro/news/more-than-a-quarter-of-belgian-companies-do-not-open-the-door-when-the-ai-agent.md", "text": "https://wpnews.pro/news/more-than-a-quarter-of-belgian-companies-do-not-open-the-door-when-the-ai-agent.txt", "jsonld": "https://wpnews.pro/news/more-than-a-quarter-of-belgian-companies-do-not-open-the-door-when-the-ai-agent.jsonld"}}