Knora 4.2: AI-FDE Loop Automates Ontology Engineering for Enterprise AI Agents

Knora's three product releases trace a continuous evolution path. In November 2025, the focus was unified ontology combining semantic graphs, behaviors, logic, and agent reasoning. In March 2026, Knora 4.0 validated end-to-end business chains in manufacturing. Version 4.2 shifts to AI-FDE — not proving ontology works, but solving how complex scenarios deliver faster and at scale.

Ontology has become a popular enterprise AI concept, but "building a business world model" does not equal completed deployment. Real projects still require data and scenario sorting, business semantics definition, executable capability construction, and operational closure. Data, documents, rules, expert experience, permissions, and responsibility boundaries do not appear automatically because an ontology platform exists.

FDE (Forward Deployment Engineering) handles this middle work: translating vague problems, expert experience, data, and documents into verifiable, runnable, governable systems. One end connects to the real business world; the other end connects to unified ontology, Action/Logic/Skills, agents, automation, and governance. FDE must define problems and semantic boundaries, build ontology and capabilities, and handle permissions, exceptions, feedback, and final acceptance. Complexity does not disappear — it is understood, absorbed, and engineered.

Knora 4.2: AI-FDE Loop in Five Steps

AI-FDE is not "FDE that understands AI" but rather turning FDE's engineering tasks into platform capabilities. Knora's multi-year foundations include unified ontology, Action/Logic, Onto-Skills, Knora Claw, and platform execution/governance. Version 4.2 extends these into the delivery process.

The full loop organizes into five steps: upgraded ontology auto-construction, text auto-extraction, Skills auto-construction, Knora Claw × Onto-Skills Agent, and Automation Trigger. Business feedback re-enters ontology and agent for the next iteration. Human review, permissions, governance, and critical acceptance run throughout.

Live Demo: Procurement Intelligent Control

Step 1 — Ontology Auto-Construction: Business description documents and connected data resources feed a construction agent. Personnel, product, batch, material tables participate directly in modeling. The new auto-construction is an interactive agent: pause, supplement information, continue building. Output includes not only entities, relations, events, but also Action and Logic, plus mapping between original business tables and ontology models. Structured data then enters the ontology.

Step 2 — Document Rule Extraction: Procurement control rules often live in policy documents, not databases. 4.2 reads documents while understanding the target ontology schema, then extracts. The demo rule document yielded 14 rules; results export for FDE verification and revision before writing to ontology, avoiding treating model extraction as formal business fact.

Step 3 — Skills and Agent: Describe the skill to build; AI organizes resources, third-party tools, and scripts per Skills spec. After Skills publish, Knora Claw gains queryable ontology objects, executable Actions, and Onto-Skills. For tasks like "check for non-compliant procurement and analyze related personnel/equipment relationships," the agent autonomously invokes capabilities, queries ontology data, unfolds reasoning chains, and produces analysis. Automation Trigger then fires Actions on ontology data changes, moving beyond user-initiated queries.

These capabilities do not eliminate FDE. AI suits describable, executable, verifiable, correctable engineering tasks; problem definition, goals, key semantics, risks, responsibility boundaries, and final acceptance remain human. The shift: work once dependent on coding, scripts, and platform proficiency automates, freeing FDE for true business judgment.

Delivery Outcomes

4.2 emphasizes four results: faster entry into business validation, lower engineering barrier, easier scaling of complex scenarios, and沉淀 (accumulation) of ontology, rules, Skills, and context as reusable assets. Complex scenario delivery shifts from reliance on few experts toward human-AI collaboration.

Enterprise AI's Next Step: Understanding What Is Happening

Past enterprise AI focused on three things: answering questions, finding knowledge, generating content — all "human discovers problem, then asks AI." But decisions impacting operations involve judgment chains: why did night-shift yield drop, which step failed, which batches affected, what if unhandled, what action next.

To answer, AI must understand the enterprise operating world. Enterprise ontology solves three basics: what exists, how they relate, what rules govern operation. Products, orders, customers, suppliers, equipment, processes, materials organize first; then relationships — order-to-product, product-to-process, process-to-equipment, material-to-supplier; finally exceptions, risks, investigation, action rules enter one semantic space. Scattered information across databases, systems, documents, and minds becomes continuously understandable enterprise operating state.

Same Ontology Capability Across Five Real Business Chains

1. Production Quality

Difficulty is not missing SPC, MES, or equipment data — it's that data, rules, and expert experience haven't formed sustainable quality decision capability. A yield anomaly investigation passes through: detect anomaly, lock time window, locate impact scope, correlate investigation, counter-evidence verification. AI takes each new evidence round, decides next query, organizes facts, hypotheses, support, and rebuttal into evidence chains — not one-shot data dump to conclusion.

Results: Single defect investigation from 2 hours to 15 minutes (87.5% time reduction); repetitive analysis workload down 72%; risk batch identification accuracy from 68% to 92% (+24 percentage points); one factory recovered +5 million RMB annual yield loss vs. historical year-over-year.

2. Operations Tracing (Dealer Claim Example)

When problems cross production into orders, warehousing, transport, after-sales, analysis chains lengthen. Parts, suppliers, warehousing, packaging, transport, dealers, order invoices, claim materials originally scattered across systems. Ontology restores these objects into one continuous business chain; AI locks objects, restores chains, locates anomaly nodes, identifies impact and responsibility. Logic handles business judgment; Action decides next — human intervention or direct improvement/follow-up tasks. Analysis goes beyond "where is the problem" to continued judgment and action along the business chain.

3. Professional Business Analysis

Handles structured data, documents, professional rules simultaneously. Platform first understands final report template and output requirements, then reverse-decomposes analysis steps from overall task, organizes business systems, knowledge bases, local materials into one task chain, invokes Logic, Skill, Action around task, combines professional calibers for rule verification. Humans intervene anytime to review/modify; every conclusion retains source chain. Result: manual sorting and writing time reduced 80%+.

4. R&D Testing (Automotive DVP Example)

Regulatory requirements, risk points, historical cases, verification targets, test tasks, prototype resources first linked via ontology. AI then completes test item generation, duplicate test identification, round strategy, prototype resource planning. While ensuring verification completeness, typical DVP planning cycle compressed from ~1 week to 1–2 days, simultaneously reducing duplicate tests, omissions, unnecessary prototype occupancy.

5. Cost Management

Lowest purchase price ≠ lowest true business cost. Procurement decisions must track through market quotes, inquiry/comparison, purchase orders, inbound batches, inventory occupancy, production consumption, actual cost-out. True cost includes purchase price plus inventory occupancy impact, quality/cost-out impact, capital/turnover impact. Logic continuously judges price reasonableness, inventory health, cost-out anomalies, supplier performance changes; Action turns analysis into alerts, rechecks, approvals, or business system write-backs, moving procurement from local low-price to enterprise total cost.

From "Human Finds AI" to Proactive AI

All scenarios converge: AI not just waits for questions, but continuously understands enterprise state, proactively starts work when key changes appear. Enterprise data changes → system identifies trigger conditions → agent auto-investigates and analyzes → forms decision recommendations → Action drives execution → results flow back to enterprise state.

Yield anomalies trigger quality investigation; fulfillment risks trigger affected order tracing; inventory or raw material price changes trigger alerts; raw material anomalies further correlate batches and suppliers. Prerequisite: AI already understands objects, relationships, states, rules.

Final loop: Perceive → Understand → Analyze → Decide → Act . Enterprise ontology describes operating world; semantic data intelligently understands changes; agent handles investigation, tracing, judgment, planning; decision and Action convert conclusions to recommendations, alerts, tasks, system operations; new execution results return to enterprise state. Different scenarios don't need new AI from scratch — on unified ontology and business semantic foundation, the same intelligent capability continuously enters production, operations, professional analysis, R&D, cost and other business chains.