Tim Berners-Lee's Vision Is Finally Arriving — 25 Years Late

In 2001, Tim Berners-Lee published a vision in Scientific American for what he called the Semantic Web — an internet built not for human eyes, but for autonomous software that could navigate, interpret, and act on information independently. It largely didn't happen on his timeline. The web remained stubbornly human-centric.

According to MIT Technology Review, that vision is now materializing — not through the Semantic Web's original architecture, but through large language models and AI agents capable of browsing, reasoning, and executing tasks across the open internet. The agentic web, as it's being called, is no longer a research concept. It is an emerging operational reality.

For Canadian mid-market companies, this is the AI development most worth paying attention to right now.

What "Agentic" Actually Means in Practice

An AI agent is software that doesn't just respond to prompts — it pursues goals. Given an objective, an agent will plan steps, use tools, browse websites, call APIs, and complete tasks with minimal human intervention between start and finish.

Practical examples already in use or in active development include:

  • Procurement agents that research suppliers, compare pricing, and draft purchase orders
  • Sales development agents that identify prospects, research their businesses, and personalize outreach
  • Operations agents that monitor dashboards, flag anomalies, and trigger predefined responses
  • Customer service agents that handle multi-step support issues end-to-end, not just answer FAQs

The distinction from earlier automation tools is important. Robotic process automation (RPA) followed rigid, pre-scripted rules. AI agents can handle ambiguity, adapt mid-task, and make judgment calls — which dramatically expands what can be automated, but also what can go wrong.

The Business Opportunity for Mid-Market Companies

Most mid-market companies are still extracting value from basic AI applications: content generation, summarization, simple chatbots. Agentic AI represents the next order of magnitude in operational leverage.

Consider a Canadian professional services firm managing 200 client accounts. Today, account managers manually compile status reports, track deliverables, and coordinate follow-ups. An agentic system could monitor all of that continuously, surface issues proactively, and draft client communications — compressing hours of administrative work into minutes.

Or consider a regional manufacturer sourcing components across multiple suppliers. An agent with access to supplier portals, internal inventory systems, and email could handle routine reordering, flag supply chain risks, and escalate only the decisions that genuinely require human judgment.

The companies that build toward this now — by cleaning up their data, documenting their processes, and piloting narrow automation use cases — will have a structural advantage when agentic tools mature into reliable enterprise-grade products over the next 18 to 36 months.

What You Need to Get Ready

Adoption readiness for agentic AI comes down to three things:

1. Data infrastructure. Agents are only as effective as the data they can access. Fragmented systems, inconsistent records, and siloed databases will limit agent performance severely. Investing in data hygiene and integration now pays dividends regardless of when you adopt agentic tools.

2. Process documentation. Agents need to follow defined workflows. If your best practices live only in senior employees' heads, they cannot be automated. Documenting core operational processes is a prerequisite — and a useful exercise in its own right.

3. Governance and guardrails. This is non-negotiable. Agents with unchecked permissions can make costly or embarrassing mistakes — placing orders, sending communications, or accessing systems in unintended ways. Any agentic deployment requires clear scope boundaries, human review checkpoints, and audit logs.

The Realistic Timeline and Next Steps

The agentic web is real, but it is still early. Current agent systems are capable in controlled environments and unreliable in open-ended ones. Enterprises deploying agents today are doing so carefully, with narrow scopes and significant oversight.

For most Canadian mid-market companies, the practical window to begin experimenting is now, with the expectation that reliable, scalable agentic workflows will be achievable within two to three years.

The leaders who will capture the most value are not necessarily the first to deploy — they are the ones who use this period to build the foundations that agentic systems require: clean data, documented processes, and a culture comfortable with measured AI adoption.

Berners-Lee's vision took a quarter century to arrive. Now that it's here, the companies with the clearest internal infrastructure will move fastest.