Enterprise Ontology for AI Reinvention
By G. Sawatzky, embedded-commerce.com
August 27, 2025
The Problem with Today's AI Adoption
- Most organizations use AI to automate existing processes (chatbots, ML optimization).
- This treats AI as a mere productivity tool, leading to three key risks:
Three Key Risks
- Commoditization: If everyone automates the same processes, efficiency becomes a standard expectation, not a competitive advantage.
- Disruption Vulnerability: Optimizing current processes can blind a company to new ways of creating value.
- Strategic Drift: Focus on operational efficiency can obscure changes in fundamental value propositions and customer needs.
The GenAI Divide: 95% Failure Rate
MIT Project NANDA study (2025): "The GenAI Divide: State of AI in Business 2025" revealed a 95% failure rate for generative AI pilots in enterprises to deliver measurable returns.
Common Reasons for Failure
- Vague objectives
- Lack of deep business integration
- AI systems failing to learn or adapt to nuanced operational contexts
- Technology-first approach without understanding core business value
- Missing human accountability
The Business Essence: Uncovering Your "DNA"
- DEMO (Design & Engineering Methodology for Organizations) provides a framework for understanding an organization's core.
- Separates stable essence from variable implementation.
- Focuses on two essential types of human actions:
Two Essential Human Actions
- Production Acts (P-Acts): The actual work that creates new business value (manufacturing a product, providing a service).
- Coordination Acts (C-Acts): Communication between people to reach agreements and make commitments (promising to deliver, accepting a request).
DEMO reveals the "business DNA" by mapping these essential acts, ensuring technological changes enhance core value creation.
The Complementary Role of Ontologies
- Enterprise Ontology (DEMO): Reveals what coordination must happen between people to create value.
- Formal Ontology (Sowa): Enables formal, logic-based knowledge representation for robust machine reasoning. Provides structured, interpretable foundation for meaningful AI analysis.
- Domain Ontology (ORM): Provides structured, interpretable models for AI reasoning. Object-Role Modeling is well-suited for neuro-symbolic integration, giving LLMs verifiable long-term memory.
Strategic insight: Successful AI transformation often requires all three approaches working together.
An Introduction to DEMO
- DEMO identifies three types of acts performed by human actors:
Three Types of Acts
- Original Acts (O-Acts): True value-creating actions that bring about new commitments and facts. Require human accountability.
- Production Acts (P-Acts): Execution of work that changes the state of the world.
- Coordination Acts (C-Acts): Communication to request, promise, state, and accept commitments.
- Informational Acts (I-Acts): Processing information (calculating, querying, analyzing). Crucial but don't create new facts or commitments.
- Documental Acts (D-Acts): Physical handling of information carriers (archiving, printing, sending). About the form of information, not content.
Key DEMO Insight for AI
Only Original Acts (O-Acts), driven by human actors, create business value and commitments. I-Acts and D-Acts are support activities for O-Acts.
Implications for AI Transformation
- What can be safely automated: I-Acts and D-Acts (information processing, document handling).
- What needs strategic enhancement: O-Acts (value-creating human commitments).
- Even when AI agents perform coordination acts, they must operate on behalf of, and in the interest of, human stakeholders.
This distinction ensures human intent and accountability remain central to all value-creating transactions.
ProMatch Inc. Example: A Simple Transaction
Transaction: Matching a Client with a Service Provider
| Step |
Act Type |
Actor |
Essential Act |
| 1 |
C-Act (rq) |
Client |
Requests a service provider match |
| 2 |
C-Act (pm) |
ProMatch |
Promises to find a suitable match |
| 3 |
P-Act (ex) |
ProMatch |
Executes the matching process (O-Act) |
| 4 |
C-Act (st) |
ProMatch |
States a match has been found |
| 5 |
C-Act (ac) |
Client |
Accepts the proposed match |
The P-Act (Execute matching process) is the core Original Act where ProMatch creates value. C-Acts are essential communication steps that make this P-Act possible.
A DEMO-Guided AI Transformation
ProMatch Inc. Case Study
- Current AI Usage: Automated matching and content generation (efficiency-focused).
- The Problem: No competitive advantage. Competitors have similar algorithms, leading to commoditization and reduced client loyalty.
DEMO Analysis Reveals Core Value
- Need Qualification: Understanding what clients actually need, not just what they ask for.
- Relationship-Based Matching: Identifying ideal fit based on working style and cultural alignment.
- Connection Facilitation: Building authentic relationships.
Current AI supports informational acts, not Original Acts. The solution: design AI to enhance human acts, not replace them.
Enhanced AI Strategy for ProMatch
- AI analyzes business data to surface patterns a human consultant might miss, providing "deeper understanding of business context."
- Human role remains to:
- Interpret emotional and cultural signals
- Make judgment calls
- Take accountability for the final outcome
By reducing processes to their essence, we're freed from the shackles of existing operations. This creates an opportunity to envision how a business might be invented from the ground up using AI, autonomous agents, or neuro-symbolic AI to fundamentally redefine value creation.
The Human Accountability Test
- As AI becomes more sophisticated, a critical question emerges: Who is responsible when things go wrong?
Two Types of AI
- Autonomous AI: AI systems making business commitments without human oversight. Invalid act—a machine cannot be held accountable.
- Authorized AI: AI systems executing pre-approved human commitments within defined parameters. Valid act—the human remains responsible for the outcome.
The accountability test: If a human is responsible for the outcome, it is a valid act. If an AI is, it is not. This ensures human authorization chains remain intact and AI acts as an intelligence amplifier, not a decision replacement.
The Future of AI and Work
- In the coming years, information processing, document generation, and routine coordination will be fully automated.
- What remains, and what will become your competitive advantage, is the essence of human contribution:
Uniquely Human Capabilities
- Human commitment-making
- Strategic judgment and intuition
- Authenticity verification and trust-building
- Ethical decision-making with stakeholder impact consideration
- Creative problem-solving in novel situations
Organizations that thrive will use AI to amplify these uniquely human capabilities, creating authentic relationships at scale while preserving accountability and trust.
Key Takeaways
- Real AI transformation requires understanding your enterprise's essence, independent of current implementation.
- DEMO separates stable essence (O-Acts) from variable implementation (I-Acts, D-Acts).
- Three complementary ontologies: Enterprise (DEMO), Formal (Sowa), Domain (ORM).
- Automate support activities (I-Acts, D-Acts); enhance value-creating activities (O-Acts).
- Maintain human accountability: Authorized AI, not Autonomous AI.
- Focus on amplifying uniquely human capabilities, not replacing them.
References
- Dietz, Jan. Enterprise Ontology: Theory and Methodology. Springer, 2006.
- Project NANDA, MIT. "The GenAI Divide: State of AI in Business 2025." July 2025.