Ignition 2027 and AI: five questions industrial teams should answer before ICC 2026

1. What is actually known today

Inductive Automation says the ICC 2026 Technology Keynote will preview Ignition 2027 and explain its plan for using AI with Ignition. The same conference program includes sessions and workshops on building agentic AI on the factory floor. That is enough to justify preparation, but not enough to design a production architecture around unannounced details.

The direction is clearer when combined with the MCP Module work. Inductive Automation announced MCP at ICC 2025 as a standard way for Ignition and modules to expose tools, resources and prompts to AI systems. An Early Access build has since been made available for experimentation, while Inductive Automation staff have said the fuller module planned with the 2027 release will add a web configuration experience and first-party tools.

• Separate announced capabilities from assumptions
• Use Early Access only to learn integration patterns
• Avoid committing production architecture to unreleased interfaces

2. Which AI use cases deserve plant-floor access

The first design question is not which model to use. It is what the AI is allowed to know and what consequence follows from a wrong answer. A maintenance explanation, a shift briefing and a production set-point change belong in very different risk classes.

For each use case, define the source systems, allowed context, latency requirement, evidence needed for a recommendation, approval boundary and audit requirement. This prevents a generic AI pilot from turning into an uncontrolled integration layer.

• Advisory answers with cited evidence
• Operator assistance inside existing workflows
• Cross-system incident investigation
• Human-approved actions with explicit policy boundaries

3. How much context should live inside Ignition

Ignition already owns valuable operational context: tags, UDT structures, alarms, historian data, Perspective sessions and project logic. Industrial AI may need some of that context, but it may also need MES genealogy, quality records, maintenance history, visual evidence, engineering documents and enterprise data.

The architecture should therefore distinguish Ignition-side context from enterprise or domain context. A future AI interface from Ignition can become one governed source in a broader operational context model rather than the only source of truth.

4. What remains deterministic

AI should not blur the boundary around deterministic control. PLC, DCS and safety logic continue to own time-critical and safety-critical execution. AI can explain, correlate, predict and recommend, but any path that changes production should inherit explicit permissions, approvals, interlocks and rollback expectations.

That principle remains useful no matter how capable Ignition 2027 becomes. It also gives cybersecurity, controls and operations teams a common language for reviewing AI use cases.

5. What to take to ICC 2026

Arrive with three concrete use cases, not a generic request to “show us AI.” For each one, bring a simple architecture diagram and a list of the information the use case needs. Then use the keynote, workshops and technology ecosystem conversations to validate the access path, governance model and module strategy.

A useful outcome from ICC is not deciding on a model provider. It is leaving with a defensible integration pattern and two or three partner/customer experiments that can be tested against real Ignition environments.

• One advisory use case
• One operator-workflow use case
• One cross-system investigation use case
• A written list of questions for the Ignition 2027/MCP team