IntelliTrack™ / AI capabilities
AI assistance is part of IntelliTrack™. Turn a manufacturing question into a conversation with your reports: find and run permitted reports, connect their results, and follow the report runs behind the answer.
Google Gemini, xAI and MCP Agent Access: scheduled for October 16, 2026. Include these capabilities in your evaluation now and plan your deployment around the October release.
AI assistance is included in the base product. You provide the AI account and pay your provider’s charges; FDS does not bill AI usage.
Scheduled release / October 16, 2026
The October release expands IntelliTrack™ AI with Google Gemini and xAI connections, plus MCP Agent Access. Bring these capabilities into your project planning and discuss the workflow you want to evaluate.
Give an MCP-capable agent read-only access to permitted IntelliTrack™ data, so it can use manufacturing evidence alongside information from your other connected systems, such as QMS, ERP and maintenance applications.
Deployment requires a current Annual Maintenance Agreement, an OIDC identity provider and HTTPS. Other systems need their own connections and permissions.
Expand your provider choices beyond OpenAI and Anthropic with Google Gemini and xAI connections. Use your own provider accounts and configure the models, access and outbound-data policy that suit your operation.
Built-in provider connections support assistance within IntelliTrack™. MCP Agent Access is the integration route for external agents.
Google Gemini, xAI and MCP Agent Access are scheduled for October 16, 2026. The report-based AI workflows below are available in the current product.
Start with work people already do
An export gives a chatbot a snapshot. IntelliTrack™ gives its assistant a controlled way to run the relevant reports and keep the sources attached.
Ask in plain language, start with Ask AI on a report, or paste a saved report link. The assistant can use the saved parameters to run that report again and combine supported report results in a conversation.
The answer lists the report runs it used. Open those runs to inspect their scope, values and context before deciding what the result means.
Example questions to evaluate
“Which operations account for the rejects in this report?”
“Compare these report results and explain what changed.”
“What should I check next, based on these records?”
Examples of questions, not promised findings. The answer depends on the available reports, captured data, selected scope and provider response.
Ask from the report you are reviewing. Bring the scope and parameters into the conversation instead of reconstructing them in a separate tool.
Follow the links to the report runs used in the answer. A plausible explanation still needs to agree with your manufacturing record.
Report tools are read-only. The assistant does not change production verdicts, recipes or genealogy through those tools.
For the people who own the system
An administrator adds your provider connection and assigns the model and connection used for each type of AI work. Ordered fallback routes use the connections you configure.
The report’s role checks and limits apply when the assistant runs it. AI is not a route around report access controls.
Review which categories of report context may be sent to the selected provider. Your prompt and permitted context leave the application when provider assistance is used.
Report-call and time limits bound each question. Large results are condensed for the model. Optional budgets track estimated AI spending; review provider billing as well.
OpenAI and Anthropic connections use your own accounts, with Google Gemini and xAI scheduled for October 16, 2026. Your IT team supplies provider credentials, model access and permitted HTTPS connectivity. Verify the selected provider and model during deployment.
Administrators can review reusable AI memory and configure its approval policy. Decide which context is suitable to retain, who reviews it, and how it should be maintained.
Core production, traceability and reporting workflows remain available without provider-based AI.
A useful starting point for documentation
Start with a User Requirements Specification. Use the validation workflow to assess requirements and select relevant tests from the FDS-authored library.
The generator supports editable qualification protocols, a requirements traceability matrix and an optional Functional Requirements Specification. Gaps and proposed responsibilities are made visible for review.
Agree on the released library coverage, intended use and required package with FDS before relying on it for your project.
Generated documents are drafts. They do not certify a system, execute validation or establish regulatory compliance.
Discuss your validation scope →Take the overview to your team
A concise overview for operations, Quality and IT, including Google Gemini, xAI and MCP Agent Access in the release scheduled for October 16, 2026.
The sheet identifies the release date and separately marks longer-term product direction.
October 16 release overview
Beyond the October 16 provider and MCP release, product direction includes monitoring and findings, AI-assisted investigations and containment, curve similarity, vision review, recipe approval advice, and troubleshooting guidance with shift summaries.
These additional capabilities do not yet have confirmed delivery dates here. Existing genealogy and quarantine workflows are available separately from these AI extensions.
Start with evidence you recognize
We’ll discuss the records behind it, the answer you need, your provider and access requirements, and how to evaluate the result. The objective is a better operational decision you can explain.