A Practical AI Automation Workflow

30 Actionable AI Automation Prompts for Digital Marketers

AI automation sounds simple. Connect a few tools, add AI, automate the repetitive work, and save time. Except that's usually where things go wrong. The real challenge is identifying what should be automated, what should remain human-controlled, where AI actually adds value, what information the workflow needs, and what should happen when the automation gets something wrong.

These 30 prompts are designed to help you think through AI automation systematically. They take you from finding repetitive work to designing workflows, deciding where AI belongs, improving existing automations, building marketing systems, and prioritizing what to automate next.

30 Practical Prompts
6 Automation Phases
1 Repeatable Workflow

The Goal

The goal isn't to automate everything. The goal is to remove repetitive work while keeping humans in control of important decisions.

Find repetitive work → Design the workflow → Add AI where useful → Automate execution → Monitor results → Improve

Don't use AI automation simply because you can. Use it where automation creates a meaningful advantage.

Phase 1: Find Automation Opportunities

Before building an automation, find out what is actually worth automating. Not every repetitive task deserves an AI workflow. Some tasks are better handled manually. Some can be automated with simple rules. Others genuinely benefit from AI.

1

Find Tasks Worth Automating

Act as an AI automation strategist for a digital marketing business. Here is a list of tasks my team currently performs: [TASK LIST] For each task, evaluate: * Frequency * Time required * Repetitiveness * Number of manual steps * Dependence on structured information * Need for human judgment * Risk if an error occurs * Potential value of AI * Potential value of traditional automation * Ease of implementation Classify each task as: Automate with rules Automate with AI Partially automate Keep human-controlled Then rank the opportunities based on: Time saved × frequency × business value ÷ implementation complexity Do not recommend automation simply because a task is repetitive. Explain why each recommended task is suitable for automation.
Use When

You have many repetitive marketing tasks but don't know which ones deserve automation.

Give AI

A realistic list of recurring tasks, how often they happen, who performs them, and approximately how long they take.

Get

A prioritized list of automation opportunities.

Do

Start with one high-value, low-complexity workflow rather than trying to automate the entire business.

2

Find Hidden Repetitive Work

Review this description of how I currently run my marketing operation: [PROCESS/DESCRIPTION] Look for repetitive work that I may not have recognized as an automation opportunity. Pay particular attention to: * Copying information between tools * Repeated data entry * Manual reporting * Repeated research * Status updates * Lead qualification * Content formatting * Content repurposing * Notifications * Follow-ups * File organization * Data cleaning * Repetitive analysis * Approval requests * Routine customer communication For each opportunity, explain: Current process → Repetitive step → Possible automation → AI involvement → Human involvement → Expected benefit Do not recommend automating activities that require sensitive judgment without clearly identifying the human checkpoint.
Use When

You feel like you're constantly busy but can't identify exactly where your time is going.

Give AI

A detailed description of your daily or weekly marketing workflow.

Get

A list of hidden repetitive tasks.

Do

Track how often these tasks actually occur before investing in automation.

3

Decide Whether a Task Needs AI

I want to automate this task: [TASK] Determine whether I actually need AI. Compare three approaches: Manual process Rule-based automation AI-powered automation Evaluate each option based on: * Accuracy * Cost * Speed * Complexity * Maintenance * Flexibility * Error risk * Scalability * Human oversight Explain what AI would contribute that a simple rule-based system could not. If AI is unnecessary, say so clearly. If a hybrid approach is better, describe the division of responsibilities between rules, AI, and humans.
Use When

You're tempted to add AI to a workflow but aren't sure whether it's actually necessary.

Give AI

The task, current process, desired outcome, and any tools you're considering.

Get

A practical decision about whether AI belongs in the workflow.

Do

Don't use an AI model where a simple trigger, filter, formula, or automation rule can reliably do the job.

4

Calculate the Potential Value of an Automation

Help me estimate the business value of automating: [TASK] Current process: [CURRENT PROCESS] Frequency: [FREQUENCY] Time required per occurrence: [TIME] People involved: [PEOPLE] Estimated implementation effort: [EFFORT] Estimate: * Hours saved per month * Hours saved per year * Approximate labor value saved * Potential revenue impact * Potential error reduction * Potential response-time improvement * Maintenance requirements * Implementation complexity Do not present speculative revenue as guaranteed. Separate: Direct measurable savings Potential operational benefits Potential revenue benefits Then calculate a simple prioritization score based on expected benefit versus implementation effort.
Use When

You need to decide whether an automation is worth building.

Give AI

Real time and cost estimates wherever possible.

Get

A rough business case for the automation.

Do

Replace assumptions with actual performance data once the automation is running.

5

Identify Tasks That Should Stay Human

Review these marketing tasks: [TASK LIST] Identify which tasks should not be fully automated. Evaluate them based on: * Business consequences of mistakes * Need for strategic judgment * Brand risk * Customer sensitivity * Legal or compliance considerations * Need for context * Relationship management * Creativity * Ambiguity * Reputation risk Classify each task as: Safe to automate Automate with approval AI-assisted only Human-only For every task classified as human-controlled, explain what makes human involvement important. Then identify whether AI can still assist without making the final decision.
Use When

You're designing an automation-heavy operation and want to avoid automating the wrong things.

Give AI

Your marketing tasks and the consequences of errors.

Get

A human-vs-AI responsibility map.

Do

Use this before automating customer-facing, strategic, or high-risk processes.

Phase 2: Map & Design Workflows

Once you've found an opportunity, don't immediately open your automation platform. First understand the workflow. A good automation starts with a clear process.

6

Map a Manual Process Before Automating It

Turn this manual process into a detailed workflow: [PROCESS] Map it as: Trigger → Input → Step → Decision → Action → Output For every step, identify: * What information enters the step * What transformation occurs * What tool is used * What condition must be true * What can go wrong * What happens next * Whether AI is required * Whether human approval is required Then identify: Unnecessary steps Duplicate steps Potential bottlenecks Automation opportunities Human checkpoints Do not automate the process yet. First redesign it into the simplest reliable workflow.
Use When

You have a messy manual process that you want to automate.

Give AI

The actual process as your team performs it today.

Get

A clean workflow map.

Do

Fix unnecessary process steps before automating them. Automating a bad process simply makes the bad process faster.

7

Design the Trigger and End State

I want to automate: [PROCESS] Define the workflow's: Trigger Required inputs Processing steps Decision points Outputs Success condition Failure condition Human escalation condition End state Also identify what should happen if: * Required information is missing * AI output is uncertain * A connected tool fails * Duplicate information is received * The same trigger occurs twice * A human doesn't approve an action * The expected output cannot be generated Present the workflow as a logical sequence that another marketer or automation specialist could implement.
Use When

You know what you want to automate but haven't clearly defined how the automation should behave.

Give AI

The desired workflow and expected outcome.

Get

A detailed workflow specification.

Do

Make failure conditions part of the design instead of adding them after something breaks.

8

Separate AI Tasks From Automation Tasks

Design an automation for: [PROCESS] Separate the workflow into three layers: ### Layer 1 — Automation Logic Triggers, filters, routing, data movement, scheduling and deterministic actions. ### Layer 2 — AI Tasks Classification, summarization, extraction, rewriting, analysis, categorization or other tasks requiring flexible interpretation. ### Layer 3 — Human Decisions Approval, strategy, exceptions, sensitive communication and decisions requiring business judgment. For every step, explain: * Why it belongs in that layer * Input required * Expected output * Failure risk * Next step Then produce the complete workflow in sequence.
Use When

You want to build a reliable AI automation instead of dumping an entire process into an AI model.

Give AI

The process, available tools and desired outcome.

Get

A clean division between automation, AI and human work.

Do

Keep deterministic operations outside the AI layer wherever possible.

9

Design an Automation With Human Approval

Design a human-in-the-loop automation for: [TASK] The automation should handle routine work automatically but require human approval before any consequential action. Define: 1. Trigger 2. Data collection 3. AI processing 4. AI recommendation 5. Confidence or quality check 6. Human approval request 7. Approved action 8. Rejected action 9. Revision path 10. Logging 11. Final confirmation Specify exactly which decisions the human should see and what information should be presented to make the approval quick and informed. Avoid requiring human approval for trivial steps that can safely be automated.
Use When

The task is suitable for AI assistance but shouldn't run completely unattended.

Give AI

The process and the actions that require approval.

Get

A human-in-the-loop workflow.

Do

Keep the approval step focused. A human shouldn't have to manually inspect every tiny operation.

10

Design an Automation Error-Handling System

I am designing an automation for: [WORKFLOW] Create an error-handling strategy covering: * Missing data * Invalid data * AI failure * Low-confidence AI output * Tool/API failure * Timeout * Duplicate trigger * Partial completion * Incorrect output * Human rejection * Unexpected input * Downstream failure For each failure, specify: Detection → Immediate response → Retry behavior → Fallback → Human escalation → Logging Also identify which errors should stop the workflow completely and which can safely be skipped or retried. The objective is to make the automation fail safely and visibly, rather than silently producing incorrect results.
Use When

You're building an automation that needs to run reliably without constant supervision.

Give AI

The complete workflow and tools involved.

Get

An error-handling framework.

Do

Test failure scenarios intentionally before trusting the automation with important work.

Phase 3: Build AI-Powered Marketing Workflows

Now we move from workflow design to actual AI-assisted marketing systems.

11

Automate Lead Qualification

Design an AI-assisted lead qualification workflow for: BUSINESS: [BUSINESS] TARGET CUSTOMER: [CUSTOMER] LEAD SOURCES: [SOURCES] Define the information the system should collect and evaluate. Create qualification criteria for: * Problem fit * Service/product fit * Budget signals * Urgency * Business size * Geographic fit * Decision-making authority * Existing solution * Buying intent Then design the workflow: Lead arrives → Data collected → AI analyzes → Lead categorized → Reason recorded → CRM updated → Appropriate follow-up → Human escalation Create clear rules for: High priority Medium priority Low priority Needs human review Do not allow AI to invent information that is not present in the lead data.
Use When

Your team manually reviews and categorizes leads.

Give AI

Your ICP, qualification criteria, CRM fields, lead sources and historical examples.

Get

A lead-qualification workflow.

Do

Review AI classifications against real leads before allowing automated routing to become fully autonomous.

12

Build an Automated Content Brief Workflow

Design an AI-powered workflow that turns a content idea into a structured marketing brief. CONTENT TYPE: [TYPE] AUDIENCE: [AUDIENCE] TOPIC: [TOPIC] The workflow should: 1. Receive the topic 2. Identify the target audience 3. Determine the intended outcome 4. Analyze search or audience intent where data is available 5. Identify important questions 6. Define the content angle 7. Create an outline 8. Identify required sources or evidence 9. Suggest internal links 10. Produce a final brief 11. Route the brief for human approval Specify which steps require AI, which are deterministic, and where human review should occur.
Use When

Your content team repeatedly creates briefs using the same process.

Give AI

Your editorial guidelines, audience information, content examples and workflow requirements.

Get

A repeatable content-brief automation.

Do

Use approved examples and clear editorial rules to keep AI outputs consistent.

13

Automate Content Repurposing

Design a content-repurposing workflow for: ORIGINAL CONTENT: [CONTENT] TARGET AUDIENCE: [AUDIENCE] CHANNELS: [CHANNELS] Create a workflow that converts the original content into channel-appropriate assets. For each output, define: * Purpose * Format * Required transformation * Tone * Length * CTA * What information must remain unchanged * What can be adapted * Human review requirement The system must not simply copy and shorten the original content. Each output should be adapted to the behavior and expectations of its specific channel.
Use When

You're repeatedly turning one article, video, webinar or report into multiple marketing assets.

Give AI

Original content, brand guidelines, audience, channels and examples of acceptable outputs.

Get

A scalable repurposing workflow.

Do

Review the first batch carefully and use the best examples as future quality references.

14

Automate Customer Feedback Analysis

Design an AI workflow that analyzes incoming: [REVIEWS / SUPPORT TICKETS / SURVEYS / COMMENTS] The workflow should automatically: 1. Collect new feedback 2. Remove obvious duplicates 3. Classify topic 4. Identify sentiment cautiously 5. Extract recurring problems 6. Identify feature requests 7. Identify praise 8. Detect urgent issues 9. Group similar feedback 10. Produce a periodic summary 11. Route important issues to the appropriate team Create categories relevant to: Product Customer experience Marketing Sales Support Potential business opportunity Include a human review step for serious complaints or ambiguous cases.
Use When

Customer feedback is spread across multiple channels and nobody has time to analyze it consistently.

Give AI

Sample feedback, existing categories and escalation rules.

Get

A structured feedback-analysis system.

Do

Use the aggregated insights to improve products, content, offers and customer experience.

15

Build an Automated Marketing Report

Design an automated weekly marketing report for: BUSINESS: [BUSINESS] CHANNELS: [CHANNELS] GOALS: [GOALS] The report should collect relevant data and summarize: * Key performance indicators * Significant changes * Wins * Problems * Underperforming areas * Important trends * Possible explanations * Recommended actions Separate: What the data shows What we think it means What we should investigate Do not allow the AI to present assumptions as facts. Define which metrics should be included and which should be ignored because they create noise. Then design the workflow: Data collection → Validation → Analysis → AI summary → Human review → Report delivery
Use When

You spend too much time manually preparing recurring marketing reports.

Give AI

Your KPIs, reporting sources, business goals and example reports.

Get

A repeatable reporting workflow.

Do

Automate data collection and summarization first. Keep strategic interpretation under human review until the system proves reliable.

Phase 4: Improve Existing Automations

An automation isn't finished when it works once. It needs to be monitored, tested and improved.

16

Audit an Existing AI Automation

Audit this automation: [WORKFLOW] Evaluate it across: * Trigger reliability * Data quality * Workflow complexity * AI necessity * Prompt quality * Output consistency * Error handling * Human oversight * Cost * Speed * Maintenance * Scalability * Security/privacy considerations * Business value Identify: What works What is unnecessary What is fragile What could fail What should be simplified What should be monitored Then give me the 5 highest-priority improvements. Do not recommend rebuilding the entire system unless the existing architecture is fundamentally flawed.
Use When

An automation works but feels unreliable, complicated or expensive.

Give AI

Workflow description, prompts, examples of outputs and performance data.

Get

An automation audit.

Do

Fix the biggest reliability problems first.

17

Improve Weak AI Outputs

Here are examples of outputs produced by my AI automation: [OUTPUTS] Here is what I actually wanted: [EXPECTED OUTPUT] Analyze the difference. Identify whether the problem comes from: * Missing context * Poor instructions * Ambiguous instructions * Incorrect input * Missing examples * Poor output structure * Conflicting requirements * Excessive freedom * Incorrect workflow placement * Lack of validation Then rewrite the AI instruction/prompt to improve consistency. Create: Improved prompt Required inputs Output format Validation rules Failure behavior Do not simply make the prompt longer. Make the instructions clearer and more operational.
Use When

Your automation technically works but produces inconsistent AI results.

Give AI

Bad outputs, desired outputs and the current prompt/instructions.

Get

A more reliable AI component.

Do

Test the improved prompt against multiple real examples rather than judging it from one successful output.

18

Reduce Unnecessary AI Steps

Here is my AI automation: [WORKFLOW] Analyze every AI call and determine whether it is actually necessary. For each AI step, identify: * Purpose * Input * Output * Why AI is being used * Whether a rule could replace it * Whether another existing AI step could perform the task * Cost implications * Failure risk Then recommend: Keep AI Replace with rule Combine with another step Remove entirely The goal is to make the workflow simpler, cheaper and more reliable without reducing its business value.
Use When

Your automation has accumulated too many AI steps.

Give AI

The full workflow and each AI prompt.

Get

A simpler automation architecture.

Do

Remember: fewer moving parts usually means fewer things that can break.

19

Find Bottlenecks in an Automation

Analyze this automation: [WORKFLOW] Find potential bottlenecks in: * Processing time * API/tool limits * Human approval * Data availability * AI generation * Manual handoffs * Duplicate processing * Error recovery * External dependencies For each bottleneck, estimate its impact and identify whether the best solution is: Remove Simplify Parallelize Batch Cache Add a fallback Change the trigger Add human capacity Leave unchanged Explain your recommendation.
Use When

An automation worked initially but is becoming slow or difficult to scale.

Give AI

Workflow steps, volumes, timing data and dependencies.

Get

A bottleneck analysis.

Do

Fix the bottleneck that limits the entire workflow rather than optimizing random individual steps.

20

Create an Automation Quality-Control System

Design a quality-control system for this AI automation: [WORKFLOW] Define validation checks for: * Required fields * Data accuracy * Output completeness * Formatting * Brand compliance * Duplicate content * Unsupported claims * Incorrect classification * Missing information * Unexpected AI output For each check, define: What is checked → How it is checked → Pass condition → Fail condition → Action Identify which checks can be automated and which require human review. The goal is to prevent bad AI output from silently moving to the next stage.
Use When

Your automation produces customer-facing or business-critical outputs.

Give AI

Workflow, expected outputs, examples of mistakes and brand rules.

Get

A quality-control layer.

Do

Build validation into the workflow instead of discovering errors after publication or delivery.

Phase 5: Create Automated Marketing Systems

Individual automations are useful. But the real leverage comes from connecting multiple processes into a system.

21

Build an Automated Lead-to-Meeting System

Design an AI-assisted lead-to-meeting workflow for: BUSINESS: [BUSINESS] IDEAL CUSTOMER: [ICP] LEAD SOURCE: [SOURCE] The workflow should cover: Lead capture → Data enrichment → Qualification → Categorization → Personalized response → CRM update → Follow-up → Meeting booking → Reminder → Human handoff For every stage specify: * Trigger * Input * AI task * Rule-based task * Output * Human involvement * Failure condition * Data that should be recorded Do not allow AI to make commitments, pricing decisions, or sensitive representations that require human authorization.
Use When

Leads currently require significant manual processing before reaching sales.

Give AI

Your sales process, qualification criteria, CRM fields and communication guidelines.

Get

A lead-to-meeting automation blueprint.

Do

Test it with a controlled group of leads before expanding it.

22

Build an Automated Content Production Pipeline

Design a content production workflow from: Idea → Research → Brief → Draft → Editing → Fact checking → SEO review → Approval → Publication → Distribution → Performance review For each stage specify: * Trigger * Input * AI task * Automation task * Human task * Required output * Quality check * Next stage Identify which stages can run automatically and which should require approval. The workflow should prioritize: Accuracy → usefulness → brand consistency → efficiency not simply publishing volume.
Use When

Your content team repeatedly moves the same type of content through the same production stages.

Give AI

Your editorial process, guidelines, tools and examples of approved content.

Get

A complete content-operations workflow.

Do

Automate administrative movement between stages before trying to automate creative judgment.

23

Build an Automated Customer Follow-Up System

Design an AI-assisted follow-up workflow for: CUSTOMER/LEAD TYPE: [TYPE] TRIGGER EVENT: [EVENT] The workflow should determine: * What happened * What the customer likely needs next * Whether follow-up is appropriate * What message type is suitable * When follow-up should occur * When follow-up should stop * When a human should intervene Create branches for: Positive response No response Negative response Question requiring human help Purchase completed Unsubscribe/opt-out The system must avoid excessive or inappropriate follow-up.
Use When

Follow-ups are being handled inconsistently or forgotten.

Give AI

Customer journey, communication rules, triggers and examples.

Get

A controlled follow-up system.

Do

Set clear stopping conditions. Automation should make follow-up more reliable, not more annoying.

24

Build an Automated Competitor Monitoring Workflow

Design an AI-assisted competitor monitoring workflow for: MY BUSINESS: [BUSINESS] COMPETITORS: [COMPETITORS] Monitor only information that could influence marketing decisions, such as: * Major product launches * Pricing changes * Positioning changes * New content * New offers * Major campaigns * Important messaging changes * Significant customer feedback patterns The workflow should: 1. Collect relevant information 2. Remove obvious duplicates/noise 3. Classify the change 4. Summarize what changed 5. Explain why it might matter 6. Compare it with our current position 7. Flag changes worth human attention 8. Ignore insignificant changes Do not create alerts simply because something changed. Prioritize meaningful competitive developments.
Use When

You want to monitor competitors without manually checking everything they publish.

Give AI

Competitor list, monitoring sources and your strategic priorities.

Get

A focused competitor-monitoring system.

Do

Define what counts as "important" before setting up the workflow.

25

Build an Automated Marketing Operations Assistant

Design an AI marketing operations assistant for my business. BUSINESS: [BUSINESS] TEAM SIZE: [TEAM] MARKETING ACTIVITIES: [ACTIVITIES] The assistant should help with: * Task intake * Information extraction * Task categorization * Prioritization * Content requests * Reporting * Meeting summaries * Follow-ups * Campaign status * Research requests * Internal notifications For every capability specify: Trigger → Input → AI action → Output → Human approval → Destination Identify which functions should be fully automated, which should be approval-based, and which should remain human-only. Also identify the minimum data the assistant needs to work reliably.
Use When

Your marketing operation has many recurring administrative tasks spread across different tools.

Give AI

Your workflows, tools, team responsibilities and recurring requests.

Get

A blueprint for an AI-powered marketing operations assistant.

Do

Start with 2–3 high-frequency use cases instead of attempting to build a giant AI assistant immediately.

Phase 6: Measure, Scale & Prioritize

The final question isn't: "What else can I automate?" It's: "What should I automate next?"

26

Measure the ROI of an Automation

Evaluate the performance of this automation: AUTOMATION: [AUTOMATION] Before automation: * Time per task: [TIME] * Frequency: [FREQUENCY] * Error rate: [ERROR RATE] * People involved: [PEOPLE] After automation: * Time per task: [TIME] * Frequency: [FREQUENCY] * Error rate: [ERROR RATE] * People involved: [PEOPLE] * Automation cost: [COST] * Maintenance time: [TIME] Calculate and explain: * Time saved * Cost impact * Error reduction * Throughput improvement * Payback period * Ongoing maintenance burden Separate measured results from assumptions. Then determine whether the automation should: Keep Improve Expand Replace Retire Explain why.
Use When

You want to know whether an automation is actually delivering value.

Give AI

Before-and-after performance data.

Get

An automation ROI assessment.

Do

Measure real results rather than assuming that "automated" means "better."

27

Find Automation Opportunities From Team Feedback

Analyze these comments from my team: [TEAM FEEDBACK] Identify recurring complaints about: * Repetitive work * Manual data entry * Slow processes * Reporting * Follow-ups * Approvals * Information retrieval * Content production * Communication * Tool switching * Administrative work Group similar problems together. For each group, identify: Problem → Frequency → Business impact → Possible automation → AI role → Human role → Complexity Prioritize problems that are both frequent and costly. Do not recommend automation for complaints that are actually caused by unclear processes or organizational problems.
Use When

Your team knows where the pain is but hasn't translated those frustrations into automation opportunities.

Give AI

Team survey responses, meeting notes, interviews or feedback.

Get

An automation opportunity backlog based on real operational pain.

Do

Fix process problems first where possible. Automation shouldn't hide a broken operating model.

28

Create an AI Automation Roadmap

Create a 90-day AI automation roadmap for: BUSINESS: [BUSINESS] CURRENT PROCESSES: [PROCESSES] AVAILABLE TOOLS: [TOOLS] TEAM: [TEAM] BUSINESS GOALS: [GOALS] Rank potential automations using: * Business impact * Time saved * Frequency * Implementation complexity * Error risk * Data readiness * AI suitability * Maintenance requirements * Dependency on other systems Organize the roadmap into: Month 1 — Quick Wins Month 2 — Core Workflows Month 3 — Connected Systems For every automation include: * Objective * Expected benefit * Dependencies * Human involvement * Success metric * Risk * Recommended sequence Do not recommend building complex systems before the foundational workflows are reliable.
Use When

You have many automation ideas but need a practical implementation sequence.

Give AI

Your current processes, tools, team capacity and business priorities.

Get

A prioritized 90-day automation roadmap.

Do

Build the simplest valuable automation first and use what you learn to inform the next one.

29

Find the Best Automation to Scale

Here are the automations currently running in my marketing operation: [AUTOMATIONS] For each one, evaluate: * Usage * Reliability * Time saved * Business impact * Error rate * Maintenance effort * Cost * Dependency on humans * Scalability * Potential to connect with other workflows Then identify: Automations worth expanding Automations worth improving Automations that should remain as they are Automations that may no longer be worth maintaining Finally, identify the 3 strongest candidates for becoming foundational systems that other workflows can connect to. Explain your reasoning.
Use When

You already have multiple automations and need to decide which ones deserve further investment.

Give AI

Your automation inventory and performance information.

Get

A scaling strategy for your automation stack.

Do

Scale proven workflows, not merely interesting ones.

30

Find the Next Best Automation

Act as my AI automation strategist. Here is my current marketing operation: BUSINESS: [BUSINESS] TEAM: [TEAM] PROCESSES: [PROCESSES] TOOLS: [TOOLS] CURRENT AUTOMATIONS: [AUTOMATIONS] BIGGEST BOTTLENECKS: [BOTTLENECKS] TIME CONSTRAINTS: [CONSTRAINTS] BUSINESS GOALS: [GOALS] Identify the 5 automation opportunities that could create the greatest practical impact. Evaluate each opportunity based on: * Time saved * Frequency * Business value * Revenue impact potential * Error reduction * AI suitability * Implementation effort * Maintenance effort * Data availability * Human oversight required Do not recommend automation simply because the task is technically possible. For each recommendation provide: 1. The process to automate 2. Why it matters 3. Current manual workflow 4. Proposed automated workflow 5. Where AI should be used 6. Where rules should be used 7. Where humans should remain involved 8. Required inputs 9. Expected output 10. Key risks 11. Success metric 12. First implementation step Then answer one final question: If I can automate only ONE process in the next 30 days, which one should I choose? Give me one answer and defend it.
Use When

You have lots of automation ideas but limited time, budget or technical resources.

Give AI

Your current processes, tools, existing automations, bottlenecks, team capacity and business goals.

Get

A prioritized automation strategy.

Do

Build the first workflow, measure the result, and use the lessons to decide what comes next.

The AI Automation Workflow in One Simple System

You don't need to use all 30 prompts at once. Use them according to where you are.

Phase 1 — Find Opportunities

#1, #2, #3, #4, #5 → Find repetitive work, identify automation opportunities, determine whether AI is actually needed, estimate value, and protect tasks that need human judgment.

Phase 2 — Design the Workflow

#6, #7, #8, #9, #10 → Map the process, define triggers and outcomes, separate AI from automation logic, create human checkpoints, and design error handling.

Phase 3 — Build AI-Powered Workflows

#11, #12, #13, #14, #15 → Automate lead qualification, content briefs, content repurposing, feedback analysis, and marketing reporting.

Phase 4 — Improve Existing Automations

#16, #17, #18, #19, #20 → Audit workflows, improve AI outputs, remove unnecessary AI, find bottlenecks, and add quality control.

Phase 5 — Build Marketing Systems

#21, #22, #23, #24, #25 → Connect lead management, content production, customer follow-up, competitor monitoring, and marketing operations.

Phase 6 — Measure & Scale

#26, #27, #28, #29, #30 → Measure ROI, find automation opportunities from team feedback, build a roadmap, scale proven workflows, and choose the next best automation.

But There's an Even Simpler Way to Remember It

When thinking about AI automation, ask five questions:

1. What is wasting time?

Use: #1, #2, #3, #4, #5

Find repetitive work and determine what is actually worth automating.

2. How should the process work?

Use: #6, #7, #8, #9, #10

Design the workflow before choosing tools.

3. Where can AI help?

Use: #11, #12, #13, #14, #15

Use AI for tasks that benefit from interpretation, classification, generation, extraction or analysis.

4. How do I make the automation reliable?

Use: #16, #17, #18, #19, #20

Audit, simplify, validate and improve the system.

5. What should I automate next?

Use: #21, #22, #23, #24, #25, #26, #27, #28, #29, #30

Connect systems, measure results, identify new opportunities and prioritize the next automation.

AI Automation Doesn't Mean "No Humans"

One of the biggest mistakes marketers make with AI automation is assuming that the goal is complete autonomy. It isn't. A better model is: Rules handle predictable work. AI handles tasks requiring flexible interpretation. Humans handle judgment, strategy and important decisions.

The Three Levels of AI Automation

You can also think about automation in three levels.

Level 1 — Assist

AI helps a human complete a task.

Examples:

  • Summarizing research
  • Drafting emails
  • Categorizing leads
  • Creating content briefs
  • Analyzing feedback

Level 2 — Automate With Approval

AI performs the work but a human approves the important output.

Examples:

  • Lead qualification
  • Content publishing
  • Customer responses
  • Campaign recommendations
  • Competitor alerts

Level 3 — Automate

The workflow operates without routine human intervention because the process is predictable and the consequences of mistakes are manageable.

Examples:

  • Data movement
  • Notifications
  • File organization
  • Report collection
  • Simple classification
  • Routine administrative actions

The objective isn't to reach Level 3 everywhere. The objective is to use the right level for each process.

The Real Point of These Prompts

These aren't designed to help you automate everything. They're designed to help you answer better questions: What should I automate? Does this task actually need AI? Where should humans remain involved? What happens when the workflow fails? How do I know whether the automation is working? Which automation should I build next?

That's the difference between adding AI to a business and actually designing an AI-powered marketing operation. Don't automate work just because automation is possible. Automate work because the result is better.

Download All Prompts (PDF)
Scroll to Top