Just as we think of autonomous vehicles (AVs) as a machine, so too is enterprise AI. With direction and purpose, that AI machine becomes the vehicle that can drive competitive advantage, accelerate innovation, lower costs, strengthen return on investment, and reduce risk, or it can get stuck in the mud. And, if AI is the vehicle, then a “semantic highway” determines how far and how fast it can go. It isn’t the model that makes the biggest difference, it’s whether your vehicle is traveling on a back road or a highway. Most businesses are still bouncing along back roads.
Businesses are hurrying to use AI by automating tasks, adding agents, and looking for areas in every department where it can create efficiency. But behind all the excitement, there is a major challenge to overcome — AI can only go as far as the data beneath it will allow. For most businesses, that data isn’t a paved road; it’s more like a muddy, rutted trail after a storm, with broken systems, duplicate records, baked-in knowledge, and years of digital trash.
This is a big reason so many AI projects get stuck. Neither the models nor the ambition of the team is the issue. Many AI projects get stuck because there is no reliable data path to follow.
“The semantic highway is what I call this missing path or foundation.” says Tirias Research Senior Analyst, Kevin Hein, “It is an integrated, reliable data layer that spans the business and gives AI the end-to-end context it needs to be most effective.”
It is likely the difference between AI going in circles delivering little value or getting results that positively affect the business.
The semantic highway is the data infrastructure that enables just how far and quickly AI can go. Companies that build one increase their chances of success with AI, enable deeper automation, and leverage cross-department intelligence that can turn into a real competitive edge. Those that aren’t are risking spinning their wheels while their competitors are racing across the finish line.
The need
AI can only make processes better if it understands them. As a result, there are real limits to what it can accomplish without clean, solid data. If that data is limited and doesn’t span across the entire business, the benefits of AI will be adversely affected. Companies that don’t have the data and data infrastructure in place yet for AI to work well will soon find themselves at a disadvantage.
A semantic highway linking all core systems, CRM, ERP, HRIS, finance, support, and operations, into a single semantic layer will help to fix this disadvantage. This semantic layer will enable AI to make more contextually consistent decisions across the entire enterprise. And consistent decisions will turn the messy reality of day-to-day operations into something trustworthy.
The semantic highway
The semantic highway takes known industry ideas and combines them into something AI can use. This semantic layer defines what the data means; a data fabric that links distributed data; knowledge graphs showing how things are related; and a unified data model to understand data structures across the business. When combined, this forms a context-rich data backbone that allows AI to understand the entire business. The semantic highway captures every workflow and record that should be part of the decision.
Unified data layers
AI doesn’t just need a place to store data, it needs to be able to understand the meaning, structure, and relationships between the pieces of data it is going to use. A consistent data layer, that spans the organization, is a key piece of the puzzle to delivering actionable AI insights and decision-making effectively.
This layer needs to include both raw data for discovering insights, and structured data to help act on it and move forward. Structured data records the facts. Raw data helps uncover the story behind the facts.
In one effective example, Siemens combined data from dozens of ERP systems and used Celonis process-mining technology to make its manufacturing, procurement, and logistics data visible. Transparency and process efficiency improved dramatically. Having all the data visible in one place can lead to big improvements in not only how effectively AI understands your business, but also how you do business.
In another example, Maersk used IBM’s TradeLens platform to expose data in a structured, shared, and interoperable way to make its global logistics planning easier. The benefits of the exposed unified data backbone were key in cutting down on delays, disputes, and manual reconciliation significantly. Unfortunately, TradeLens itself was difficult for the industry to adopt and ended in 2023.
IBM has since shifted its strategy towards more modular data fabrics, knowledge-graph driven architectures and more vertical-specific data networks. However, its core vision of shared, trusted and real-time data across an enterprise remains and can be implemented using tools like watsonx.data and is key in enabling effective AI decision making.
These efforts demonstrated that AI projects are more effective when a data layer is unified across an organization. This is especially true in particular areas being reworked with AI like analytics, forecasting, and automation. As a result, a unified data layer is a key part of a semantic highway.
AI-Powered Data Hygiene and Automated Governance
Another aspect of the semantic highway is data hygiene. AI cannot operate dependably on inconsistent or contradictory data. Many companies across many fields are now using AI itself for data cleaning, data classification, sensitive-data detection, and governance automation, such as:
- Banks and other financial institutions are using AI to detect problems, sort documents, and cut down on false positives in compliance workflows. This lowers the risk of breaking the law and eases audits.
- The EU is using AI for policy detection and classification to enforce the General Data Protection Regulation (GPDR)
- U.S. Healthcare systems are researching ways to use AI to automate the detection of Protect Health Information (PHI) to ensure compliance with the Health Insurance Portability and Accountability Act (HIPAA).
Automating data hygiene is a common theme in all of these applications. It lowers risks, speeds up audits, and builds trust in what AI delivers.
A critical question to ask when ensuring data hygiene is “Can we prove where all regulated data lives — right now?” Companies automating governance, improving data hygiene, and building this into their semantic highway can answer yes.
Enterprise Knowledge Graphs
When building its semantic highway, an enterprise also needs to maximize the data to which it has access and can leverage. Knowledge is typically spread out over emails, Slack threads, PDFs, spreadsheets, SharePoint, Confluence, proprietary industrial platforms, and even resides in the heads of employees who might leave next year. AI cannot utilize this information if it is hidden from it. Current estimates put unused data, also called dark data, at 55% to 80% of all data generated and stored by organizations.
To address this, many companies are making enterprise knowledge graphs. These graphs show how all of its data is related, what it means, and its context. Some examples include:
- Atlassian’s publicly announced “Teamwork Graph”. It’s a semantic model that links projects, teams, documents, and workflows to make both search and onboarding better.
- IBM’s published research on enterprise knowledge graphs to help people make tough decisions in regulated industries.
- Intuit’s single knowledge graph that powers its AI-based financial advice.
Knowledge graphs have evolved from work in the 1970s but using them throughout an organization and with generative AI is relatively new. Knowledge graphs enable the semantic highway to go beyond being just a data layer and to an intelligence layer.
Process Intelligence
Optimizing data freshness is another aspect of the semantic highway that cannot be overlooked. One way to do this is through process mining and intelligence. Process mapping historically referred to walls covered in sticky notes, whiteboards, and process diagrams. While these resources were informative, much of the data was out of date by the time the information was documented. Companies are now using process mining and process intelligence to determine how work really gets accomplished.
Feeding unified ERP data into a process mining engine allowed Siemens to see procurement, invoicing, manufacturing, and logistics in real time. This enabled problems to be found in a timely manner, fewer mistakes to be made, and enhanced the effectiveness of its global digital transformation program.
Huge reductions in rework, delays, and manual intervention are being seen at companies that use process intelligence tools. Best of all, results are being seen in a few weeks instead of months.
The ROI of a semantic highway
AI isn’t just about automating tasks; it is about demonstrating huge benefits when being used to reveal and redesign the work itself. But to accomplish this, it needs a semantic highway to support it with a structured and connected data foundation.
The quality of the data, a unified data architecture, good data governance, and data structures lead to successful AI implementations. In other words, investing money upfront in data foundations leads to faster, cheaper, and more reliable AI results.
In three years, no one will remember who had the best chatbot wrapper or the coolest copilot in 2025. They will remember which companies they could trust with their data and which ones they couldn’t. To move your business forward with AI, all you need to do is ask yourself: Is there a highway under it or a dirt path? It’s never been easier to close the gap, but you have to start now, not after the next AI cycle passes you by.
