Introducing M-Files Custom Agents: AI That Doesn't Just Read Your Documents — It Acts On Them
AI in Document Management Has Moved Beyond Search and Suggestions
For the last few years, artificial intelligence in document management has mostly meant one thing: helping people find information faster, summarizing a lengthy PDF, or suggesting the occasional metadata tag.
While helpful, this model is inherently passive—AI only helps when a user explicitly stops what they are doing to prompt it.
That is changing. M-Files Custom Agents, now in public beta, represents a major evolution in enterprise document automation. It moves AI from a passive desktop assistant into an active participant in your business workflows—reading content, reasoning about business rules, and taking concrete actions without requiring a human to manually trigger every single step.
For Australian organisations juggling growing document volumes and stretched operational teams, that shift matters. It’s the difference between an AI that answers questions when asked, and an automated agent that quietly keeps your processes moving forward in the background, every day.
What Is M-Files Custom Agents?
M-Files Custom Agents allows system administrators to configure intelligent, agentic AI directly into automated document workflows.
When an object or document reaches a defined stage in your business process, the Custom Agent:
How an M-Files Custom Agent Operates
Reads the document content and properties explicitly provided to it.
Evaluates the task against instructions written in plain natural English.
Executes configured steps—updating metadata, creating tasks, or advancing workflow.
- Reads the document content and related metadata it has been explicitly given access to.
- Reasons about the task using instructions written in plain, natural English.
- Acts by taking actions defined in your prompt—for example, updating properties on the record, creating new linked items (such as follow-up tasks, assignments, or notifications), or transitioning the workflow to the next phase.
Crucially, this requires no custom coding and no brittle rule trees that break every time your internal processes evolve. You define the prompt in M-Files Admin, and the agent executes consistently each time that workflow step triggers.
How It Works in Practice
Configuring an M-Files Custom Agent follows four straightforward principles:
1. Plain-Language Instructions
Prompts are defined using natural language rather than scripting or code. You frame the brief broadly—outlining the situation, what needs to be done, and what the completed outcome looks like.
2. Explicitly Scoped Execution
The agent operates strictly on the object it is running on. Anything else it needs to evaluate must be explicitly provided in the prompt—it never searches or roams across external records on its own.
3. Structured, Actionable Output
Instead of generating loose text summaries, the agent produces structured object properties, assigned tasks, or status changes that feed directly into downstream workflow states.
4. Transparent Audit Trail
Every property update or action initiated by an agent is badged as AI-generated, complete with a click-through reasoning log so staff can inspect how a conclusion was reached.
Where This Saves Time
M-Files Custom Agents helps reduce manual administrative effort across three key areas:
1. Automated Task Creation & Delegation
Instead of someone manually reading through a document to extract follow-up tasks and re-key them into a system, the agent parses the content, creates the necessary sub-tasks, assigns them to the responsible personnel, and links them directly back to the source record.
2. Rule Consistency & First-Pass Compliance Checks
Checking complex documents against internal standards or procedures is time-consuming and prone to human error when teams are fatigued. An agent applies the exact same evaluation criteria every time, catching omissions before human review begins.
3. Smart Triage & Pre-Review Flagging
Rather than skilled managers spending time doing basic document checks or verifying whether prerequisite fields are populated, the agent conducts the initial triage. Reviewers can focus their energy on high-value judgment calls and approvals.
Built With Enterprise Governance at Its Core
Automation is only worth adopting if your organisation can stand behind every decision it produces. M-Files Custom Agents takes a deliberately conservative, governance-first approach:
Straight-Through Processing
For lower-risk administrative tasks (such as metadata classification or initial triage), the agent can execute directly from start to finish.
Human Checkpoints
For consequential determinations, the agent compiles findings and recommendations for human review and final approval.
- Scoped Execution: The agent only works on the object it is running on, and anything else must be explicitly given to it in the prompt. It never searches for external items on its own.
- Human-in-the-Loop Safeguards: Automation does not mean giving up control. You decide which actions the agent can complete automatically and where a human must review and approve before anything takes effect.
- Immutable Version History: If a team member modifies a field originally generated by the AI, M-Files preserves both the AI reasoning and the user modification in the document’s audit history.
M-Files helps you meet regulatory and quality management requirements (such as ISO 9001 and ISO 27001), and Custom Agents provides another tool to help you maintain compliant, well-governed processes.
How Range Information Systems Helps You Get Started
As a certified Australian M-Files Partner, Range Information Systems works closely with your team to safely implement and optimise agentic AI workflows:
- Workflow & Prompt Architecture: Identifying high-friction business bottlenecks and writing robust, reliable natural-language prompts.
- Metadata Structure Optimisation: Refining your metadata vault schema so agents have clean, structured context to reason effectively.
- Governance & Approval Checkpoints: Designing the right balance between straight-through automation and mandatory human review.
- Pilot Testing & Safe Rollout: Validating agent behaviour in staging environments before deploying to live production vaults.
Summary
M-Files Custom Agents represents a genuine leap forward in enterprise information management:
- ✅ Active Execution: Moves AI from a passive search box into an active workflow participant.
- ✅ No-Code Configuration: Configured entirely using plain-English prompts in M-Files Admin.
- ✅ Full Explainability: Every automated change includes a transparent click-through explanation.
- ✅ Strict Governance: Human oversight remains exactly where your compliance policies require it.
In our next deep-dive post, we explore four real-world configuration examples—from auto-triaging safety incidents to self-reviewing policy documents.
Ready to explore M-Files Custom Agents in your organisation?
Contact Range Information Systems to discuss your automation goals, or explore our M-Files Information Management solutions.
🔗 Related Reading: