promptprompts7 min read

Operations Process Automation Prompt Template

A reusable AI prompt template for operations managers automating repetitive processes — covering process mapping, bottleneck identification, SOP documentation, and exception handling, with placeholder fields for any team or toolset.

Use case: AI-assisted process automation scoping and documentation for operations teams

Template objective#

This prompt template gives operations managers a consistent starting point for engaging an AI agent across four distinct phases of process automation: mapping the existing process, identifying where it breaks down, documenting the resulting standard operating procedure, and defining how the agent handles exceptions.

Each variation is designed to be reusable across different processes without modification to the core logic — only the bracketed placeholder fields need to change. Teams that build a library of these prompts reduce their prompt-engineering overhead substantially when rolling out automation to a second or third process.

Prerequisites#

Before using any variation of this template:

  1. Identify the specific process by name and define its start and end points.
  2. Know the team size involved and the frequency with which the process runs.
  3. Document which tools and systems are involved (ERP, ticketing system, spreadsheets, etc.).
  4. Estimate the average cycle time from trigger to completion.
  5. Identify the person who currently owns the process and can validate outputs.

Core prompt template#

Copy the block below and replace each [PLACEHOLDER] with your specific values.

You are an operations analyst helping to automate business processes at [COMPANY NAME].

Process name: [PROCESS NAME — e.g., "vendor invoice approval", "employee onboarding", "monthly inventory reconciliation"]
Process owner: [TEAM OR ROLE THAT OWNS THIS PROCESS]
Team size: [NUMBER OF PEOPLE INVOLVED]
Process frequency: [HOW OFTEN IT RUNS — e.g., "daily", "weekly", "per purchase order"]
Tools currently used: [LIST TOOLS — e.g., "NetSuite, Slack, Google Sheets, Jira"]

Objective: [WHAT THE PROCESS ACCOMPLISHES WHEN DONE CORRECTLY]
Current pain points: [KNOWN ISSUES OR REASONS FOR AUTOMATING]

Analyze and respond to the following task: [INSERT SPECIFIC TASK FROM VARIATIONS BELOW]

Variation 1: Process mapping prompt#

For producing a structured map of an existing manual or semi-automated process before designing automation:

You are an operations analyst at [COMPANY NAME].

Process to map: [PROCESS NAME]
Process owner: [TEAM OR ROLE]
Team size involved: [NUMBER]
Frequency: [DAILY / WEEKLY / PER EVENT]
Tools used: [TOOL 1], [TOOL 2], [TOOL 3]
Current average cycle time: [TIME — e.g., "3 hours", "2 business days"]

Map this process from trigger to completion. For each step, provide:
- Step name
- Who performs it (role, not individual name)
- Input required (data, document, system state)
- Action taken
- Output produced
- Tool used
- Approximate time taken

Format the output as a numbered list of steps.
After the step list, identify: how many steps are candidates for automation based on being rule-based, repetitive, and not requiring human judgment.

Context to apply: [ANY KNOWN PROCESS DETAILS, SYSTEMS, OR CONSTRAINTS TO INCORPORATE]

Variation 2: Bottleneck identification prompt#

For diagnosing where a process slows down, fails, or produces rework:

You are an operations analyst at [COMPANY NAME].

Process: [PROCESS NAME]
Process map: [PASTE PROCESS MAP OR DESCRIBE THE STEPS]
Team size: [NUMBER]
Known pain points: [LIST ANY KNOWN ISSUES — e.g., "approvals sit for 2+ days", "data entry errors in step 4", "unclear handoff between finance and ops"]
Volume: [HOW MANY INSTANCES PER WEEK/MONTH]

Analyze this process for bottlenecks. For each bottleneck identified:
- Location: which step or handoff point
- Type: capacity constraint / approval delay / data quality issue / system friction / unclear ownership
- Impact: estimated time lost per instance
- Root cause hypothesis: what is causing the delay or failure
- Automation potential: whether an AI agent could address this bottleneck, and how

Prioritize bottlenecks by impact. Output as a ranked list with a brief remediation recommendation for each.

Variation 3: SOP documentation prompt#

For generating a draft standard operating procedure from a mapped process, suitable for team training and agent configuration:

You are a technical writer and operations specialist at [COMPANY NAME].

Process name: [PROCESS NAME]
Process owner: [TEAM OR ROLE]
Audience for this SOP: [WHO WILL FOLLOW IT — e.g., "new operations coordinators", "the AI agent being configured to run this process"]
Process map or description: [PASTE MAP OR DESCRIPTION]
Tools used: [TOOL LIST]
Edge cases known: [LIST KNOWN EXCEPTIONS OR SPECIAL CASES]

Write a standard operating procedure (SOP) for this process.

SOP structure:
- Purpose: one sentence on what this process achieves
- Scope: who and what this SOP applies to
- Preconditions: what must be true before the process starts
- Step-by-step instructions: numbered, with tool and action for each step
- Decision points: where the process branches based on a condition
- Exception handling: how to handle [EXCEPTION 1], [EXCEPTION 2]
- Escalation path: who to contact if the process cannot complete normally
- Success criteria: how to confirm the process completed correctly

Tone: [INSTRUCTIONAL TONE — e.g., "clear and direct, use active voice, written for someone doing the task for the first time"]

Variation 4: Exception handling prompt#

For defining how an AI agent should handle cases that fall outside the normal process flow:

You are configuring an AI agent to handle [PROCESS NAME] at [COMPANY NAME].

Normal process flow: [DESCRIBE OR PASTE THE STANDARD PROCESS]
Tools the agent has access to: [TOOL LIST]
Agent authority level: [WHAT THE AGENT CAN DO WITHOUT HUMAN APPROVAL — e.g., "read and write to the ticketing system, send Slack notifications, but cannot approve payments or modify financial records"]

Define exception handling logic for this agent. For each exception type:
- Exception name: [EXCEPTION TYPE — e.g., "missing required data", "approval timeout", "system unavailable", "value outside normal range"]
- Trigger condition: the specific input or system state that indicates this exception
- Agent response: what the agent should do (retry, pause, notify, escalate, log and continue)
- Notification target: who receives an alert and via which channel
- Timeout or retry logic: how many retries, how long to wait, what to do if retries fail
- Logging requirement: what to record about the exception and its resolution

Exception types to define:
1. [EXCEPTION 1 — e.g., "required approval not received within 24 hours"]
2. [EXCEPTION 2 — e.g., "source data file missing or malformed"]
3. [EXCEPTION 3 — e.g., "output value exceeds threshold requiring human review"]
4. [EXCEPTION 4 — e.g., "dependent system API returns error"]

After defining each exception, output a summary exception handling matrix.

Customization guidance#

Process specificity. The most important customization is an accurate description of the tools involved. Agents configured without specific tool names tend to produce generic automation designs that don't reflect your actual stack. Be explicit: "NetSuite via REST API", "Jira Service Management ticket creation", "Google Sheets with named range 'VendorList'".

Authority boundaries. The Variation 4 exception handling prompt includes an authority level field — this is the most important field in the entire template set. Define clearly what decisions the agent can make autonomously versus which require human sign-off. Agents with under-specified authority boundaries either over-escalate (creating friction) or under-escalate (taking actions they should not).

Iterative SOP refinement. The SOP output from Variation 3 should be reviewed by the current process owner before it is used to configure an agent. Process owners routinely catch implicit steps that were not described in the original briefing — steps they do automatically that they did not think to mention.

Implementation guidance#

These prompts are designed to be used sequentially: map the process first (Variation 1), identify what to fix (Variation 2), document the corrected process (Variation 3), then define how the configured agent handles failure cases (Variation 4). Running them in order builds each output on the previous one.

For a full operations automation deployment workflow that uses these prompts as input to agent configuration steps, see the Operations Workflow Automation Blueprint.

For guidance on how to assess which processes in your organization are suitable for agent automation before starting this template sequence, the AI agent use cases in operations guide covers common patterns and readiness criteria.

Tags:
promptoperationsautomationprocess-improvement
← Back to Prompt Templates