Resource

    🤖 The 2025 Guide to AI in Marketing Operations

    6/16/2025
    by Sara McNamara
    A comprehensive, practical template built for the MOPs pro tasked with both exploration and execution. These are AI-powered processes you can actually set up, govern, and optimize.
     
    ✋ Setting the Stage: Ignore the Hype🔍 Best Practices with AIThe Cost of Using AI vs. Humans📝 Prompting Techniques for Different AI Platforms🗯️ Use Cases🎼 Lead Scoring: Make Sales Prioritization Smarter🖼️ Content Personalization: Serve the Right Message, Every Time📨 Predictive Campaign Modeling: Optimize Before You Hit Send💬 Chatbots and Virtual Assistants: Your 24/7 SDR🆘 AI-assisted Customer Success Bots…in a GOOD Way🚿 Data Cleaning and Enrichment: Fix Your Dirty Data🗺️ Customer Journey Mapping: Reveal the Real Buyer Path🏧 Automated Content Creation: From Zoom to Blog in Minutes🛠️ Tool Recommendations💼 Legal Considerations👥 Scaling AI Usage Across Teams👎 When Not to Use AI💭 Final Thoughts

    ✋ Setting the Stage: Ignore the Hype

    Many AI-company CEOs are speculating that AI will take over all human jobs, or replace salespeople altogether…while the technology may be able to mature quickly, think about the cultural shifts that will have to take place in order for this to happen.
    I don’t see those shifts happening overnight.
    HOWEVER, as operations professionals, we do need to keep up with AI and use it to our advantage. We need to take the IBM approach — learn to run the machines, so we don’t get replaced completely by them. Learn to be as productive as possible with as few people as possible.

    🔍 Best Practices with AI

    There are a few best practices to keep in mind when we talk about AI:
    1. AI hallucinates. Look to verify every metric or statement it makes — it can be get to get you started or to help you think outside-the-box on a complex problem, but it isn’t a super genius. It aims to please, so if it doesn’t know an answer, it will BS and pretend it does. Use, but verify.
    1. Always have a human review content before it goes out to the general public. Because AI hallucinates, and because it can contain biases, it’s important to have human oversight of any external communications. No matter what any CEO tells you or orders you to try to do, AI cannot function completely on its own. This about AI as augmentation, not replacement.
    1. Start small. Run experiments with trusted stakeholders and only expand use cases once they’re working. This is true for basically any major change, by the way — best to battle-test before launching more broadly.
    1. Be sensitive with PII (personal identifiable information) and work with your legal/compliance representatives. AI is not well-regulated right now, but that doesn’t mean that you should throw PII in there or put your company’s sensitive information in there. Work with legal to make sure you understand their stance on current law and what use cases you can run. Yes, sadly, if you work for a public company…this probably means that your use cases will be tiny for now. Sorry. 😬

    The Cost of Using AI vs. Humans

    Here are some examples of the cost of using AI vs. using Humans on tasks:
    Task
    AI Cost
    Human Cost
    Drafting a nurture email
    $0.002–$0.10 per prompt (ChatGPT)
    ~$75–$150/hour (marketer or freelancer)
    Data enrichment per lead
    ~$0.005–$0.02 (Clay, Clearbit, etc.)
    ~$0.50–$2.00 (manual + platform fees)
    Summarizing webinar notes
    ~$0.10 (Descript + GPT)
    ~$40–$60/hour
    Segmenting leads
    ~$10/month (AI-powered scoring)
    ~$1,000–$5,000/month (analyst hours)
    On face value, it looks like using AI is cheaper than human efforts. HOWEVER — most AI still requires human intervention, whether it’s through the prompting phase, iteration phase, and/or review phase. There is no totally-independent AI.
    BUT — AI can help your humans be more productive during their hourly rate of work.
    For these reasons, my recommendation to you: think of AI as augmenting humans, not replacing entire teams.
    “Our content team used to spend 6 hours per blog post across 3 roles. With GPT-assisted workflows, that’s down to 1.5 hours, at a cost of ~$0.20 in prompts vs. $300+ in hourly rates.”
    👆 These are the kinds of success stories you will hear that are legitimate. Focus on use cases like this.
    There is no one-size-fits-all, and each approach has downsides:
    There are hidden human costs:
    • Time to scale: AI is instant, humans require onboarding and ramp.
    • Inconsistency: Human quality varies, AI is repeatable.
    • Interruption cost: Marketers get distracted by small tasks that AI can handle.
    But there is also overhead in AI:
    • Prompt engineering and output validation are still human tasks.
    • There’s a learning curve and governance layer (especially in larger orgs).
    • Licensing cost adds up if AI usage isn’t scoped well.
    The goal of AI usage is to free up human talent for judgment, strategy, and innovation, not eliminate them.

    📝 Prompting Techniques for Different AI Platforms

    Platform
    Prompting Style
    Example Prompt
    Strengths
    Caveats (with Examples)
    ChatGPT
    Instructional (i.e. “Act as…”, “Write a…”)
    “Act as a marketing ops lead. Write a nurture email for a SaaS CFO in healthcare.”
    Great for task workflows, strong structure control
    Can hallucinate data (i.e. invent fake martech tools), or forget role (i.e. switches tone mid-output)
    Claude
    Narrative + long-context
    “Here is a 5-page whitepaper. Summarize key points for a CMO brief.”
    Excellent summarizer with deep context retention
    May return vague outputs (i.e. “the document explores several strategic themes…”) unless prompted tightly
    Cursor
    Code + comment-based
    // write a SQL query to show MQLs by industry last quarter
    Works inline in code, great for structured dev tasks
    Misunderstands ambiguous comments (i.e. “get pipeline velocity”), or writes inefficient queries
    Notion AI
    Editor-style (button prompts)
    “Make this paragraph more concise.”
    Fast for rewriting and polishing content
    Limited control over tone or structure (i.e. “make it snarky” doesn’t register well)
    Jasper
    Template-based with tone/style modifiers
    “Create a LinkedIn post in a confident tone about AI for lead scoring.”
    Easy to produce on-brand copy quickly
    Templates can get repetitive (i.e. multiple ads sounding identical); tough to deviate from format
    Perplexity
    Research-based with citations
    “What are examples of AI use in B2B marketing? Include sources.”
    Great for fact-checking and source-backed summaries
    Won’t write compelling marketing copy; answers (i.e. “According to XYZ, AI can…”) dominate tone
    Claygent
    Structured queries for real-time data fetching
    “Find revenue and headcount for B2B SaaS companies using HubSpot.”
    Great for pulling live data, company intelligence
    Can return zero or outdated results if filters are too niche (i.e. “Series B, 200–300 FTE, EU only”)

    🗯️ Use Cases

    Below, I’m going to take you through some use cases that I’ve set up myself or that others have told me about. Use them as guide posts — not everyone has the budget or justification to run the more complex AI use cases, but becoming familiar is useful so you can think about what your company can justify or budget for.

    🎼 Lead Scoring: Make Sales Prioritization Smarter

    What it is: Using predictive models to evaluate which leads are most likely to convert, based on a mix of demographic and behavioral data.
    Why it matters: Traditional lead scoring often relies on arbitrary point systems. AI looks at historical conversion patterns to assign more accurate scores, automatically.
    How to set it up:
    • Connect your MAP and CRM with tools like MadKudu or 6sense.
    • Feed in conversion data to train your model.
    • Pipe the AI-generated scores back into your CRM for routing and prioritization.
    Example: Use AI to downgrade form-filler leads who never open emails and upgrade silent leads from high-intent companies.
    Case Study: A B2B company using MadKudu saw a 40% increase in sales productivity by focusing only on AI-prioritized leads.

    🖼️ Content Personalization: Serve the Right Message, Every Time

    What it is: Dynamically adapting messaging across channels based on who your user is and what they’ve done.
    Why it matters: Personalized experiences drive better engagement. AI helps deliver those experiences at scale.
    How to set it up:
    • Segment by firmographic or behavioral data.
    • Use dynamic blocks in email/landing pages driven by tools like Mutiny or Persado.
    • Pull copy variants from GPT APIs for on-the-fly personalization.
    Example: A returning healthcare prospect sees healthcare-specific copy, while a finance exec sees financial messaging.
    📖 Case Study: Crabtree & Evelyn used AI-powered personalization to increase return on ad spend by 30% through improved targeting.
    🤔 How you can get started today, even with little budget:
    🤖 Use ChatGPT or Jasper to help you create content variations.
    🤖 Use tags in your marketing automation platform, like product interest, to help determine which ad should display to which person.
    🤖 Use a tool like Vector to target specific people and see which specific people interact with your ads, across platforms.
    🤖 Use a tool like Metadata.io to run B2B ad campaigns using AI.

    📨 Predictive Campaign Modeling: Optimize Before You Hit Send

    What it is: Forecasting your campaign performance before it even launches.
    Why it matters: You don’t have to A/B test your way to effectiveness—AI can model success likelihoods upfront.
    How to set it up:
    • Feed previous campaign metadata and outcomes into a tool like Jasper Predict.
    • Analyze new campaign drafts to see predicted open rates, CTR, and conversions.
    • Use results to fine-tune subject lines, CTAs, and targeting.
    Example: Predict 15% higher conversions if you use a first-person subject line with a “how to” format.
    Case Study: LinkedIn used AI modeling to rank and optimize email campaigns, resulting in higher open and engagement rates across millions of users.

    💬 Chatbots and Virtual Assistants: Your 24/7 SDR

    What it is: AI-powered bots that qualify leads, book meetings, and answer FAQs.
    Why it matters: Live chat is high-converting, but not scalable. AI makes it scalable.
    How to set it up:
    • Choose a bot platform like Drift or Qualified.
    • Train the assistant on common queries and connect to your CRM.
    • Route high-intent users straight to your sales team.
    Example: A bot can distinguish between a job seeker and a decision-maker and personalize the experience accordingly.
    Case Study: A SaaS company implemented Qualified's AI bot and booked 2x more demos without expanding their SDR team.

    🆘 AI-assisted Customer Success Bots…in a GOOD Way

    VentureBeatVentureBeatCloudera built its conversational AI chops by keeping things simple
    Cloudera built its conversational AI chops by keeping things simple

    Cloudera built its conversational AI chops by keeping things simple

    It turns out if you're a data company with modern data architectures at the ready, you have a head start building out internal conversational AI.

    VentureBeatVentureBeat

    🚿 Data Cleaning and Enrichment: Fix Your Dirty Data

    What it is: Using AI to detect duplicates, normalize fields, and enrich records without human intervention.
    Why it matters: Clean data powers every other system. Manual cleanup doesn’t scale. AI-driven enrichment does.
    How to set it up:
    • Identify key fields for standardization (e.g., job title, company name).
    • Use tools like Clearbit, OpenAI, or People Data Labs to auto-correct or append data.
    • Set up real-time enrichment for new leads.
    Example: Normalize “VP of Marketing,” “VP Mktg,” and “V.P., Mktg” into one field—and enrich with LinkedIn URL.
    Case Study: Intel applied AI to structure public web data and enrich contact records—boosting sales intelligence and conversion efficiency.

    🗺️ Customer Journey Mapping: Reveal the Real Buyer Path

    What it is: Visualizing how users interact across multiple touchpoints and using AI to uncover common patterns.
    Why it matters: Traditional funnel models are outdated. AI helps you see the real, nonlinear journey.
    How to set it up:
    • Integrate tools like Heap or Amplitude with your MAP and CRM.
    • Use clustering to group similar behaviors.
    • Recommend next best actions based on AI insights.
    Example: Discover that content engagement spikes after sales calls—and adjust nurture timing accordingly.
    Case Study: A major retailer applied AI journey mapping to personalize recommendations and saw a 25–30% boost in online sales.

    🏧 Automated Content Creation: From Zoom to Blog in Minutes

    What it is: Automating the conversion of recorded content into publishable assets.
    Why it matters: Content creation is slow and expensive. AI can accelerate repurposing without losing quality.
    How to set it up:
    1. Record a session in Zoom.
    1. Sync to Descript for transcription.
    1. Use Zapier to trigger GPT workflows:
        • Draft blog summaries
        • Generate LinkedIn posts
        • Write email copy
    1. Auto-load into CMS or MAP for review.
    Example: A 30-minute webinar turns into 1 blog post, 3 tweets, and 2 nurture emails—automatically.
    Case Study: A marketing agency automated its entire podcast-to-content workflow using Zoom, Descript, and GPT—saving over 20 hours per week in production time.

    🛠️ Tool Recommendations

    Use Case
    Native Tools
    External AI Tools
    Lead Scoring
    Salesforce, HubSpot
    MadKudu, 6sense, Infer
    Content Personalization
    HubSpot, Adobe
    Mutiny, Persado, Jasper
    Performance Prediction
    Salesforce
    Metadata.io, Jasper, OpenAI
    Chatbots & Assistants
    Drift, Intercom
    ChatGPT, Qualified
    Data Cleaning/Enrichment
    HubSpot, Marketo
    Clearbit, ZoomInfo, OpenAI
    Journey Mapping
    Adobe, HubSpot
    Heap, Amplitude, Pathmonk
    Content Automation
    None
    Zoom, Descript, Zapier, OpenAI

    💼 Legal Considerations

    1. Data Privacy Laws (GDPR, CCPA, etc.)
    • You must disclose if you're using AI tools that process personal data.
    • Do not feed customer data — including names, emails, IPs — into public AI tools (i.e. ChatGPT, Gemini) unless you have a business agreement in place.
    • Use anonymization or pseudonymization where possible.
    2. Vendor Contracts and DPAs
    • Ensure all AI vendors are covered by a Data Processing Agreement (DPA).
    • Review their sub-processors and data retention policies.
    • Check if the AI tool trains on your data—some do unless explicitly disabled.
    3. Consent Requirements
    • If your AI-generated personalization relies on user behavior, ensure your privacy policy reflects this use.
    • Cookie-based AI models may require opt-in consent depending on the region.
    4. Content Ownership
    • Clarify whether AI-generated content (especially via third-party tools) is considered “owned” by your company.
    • Some platforms reserve rights to reuse prompts or output unless explicitly restricted.
    5. Bias and Discrimination
    • Be cautious when using AI for segmentation or predictive lead scoring. Biased training data can result in discriminatory outputs.
    • You may be liable under anti-discrimination or consumer protection laws if unfair outcomes occur.
    6. Auditability and Explainability
    • In regulated industries (i.e. finance, healthcare), you must be able to explain AI-driven decisions.
    • Keep records of inputs, outputs, and model parameters where possible.
    7. Marketing Claims
    • If AI generates copy with claims (i.e. “proven to boost revenue 5x”), it still must comply with advertising standards and be verifiable.

    👥 Scaling AI Usage Across Teams

    1. Create a Centralized AI Playbook
    Document approved use cases: content generation, summarization, lead scoring, and more. Include preferred tools (ChatGPT, Jasper, Claude) and how to access them. Maintain a list of standard prompts for common tasks like email writing or persona-based messaging.
    2. Set Up a Prompt Library
    Build a shared Notion, Confluence, or Google Doc with tried-and-tested prompts. Organize them by function — i.e. demand gen, content, operations—and encourage the team to contribute new ones as they iterate.
    3. Assign AI Champions
    Designate “AI captains” per team or function. These folks maintain prompt libraries, train peers, vet new tools, and serve as the go-to for AI-related questions.
    4. Standardize Tools and Access
    Pick a primary AI tool to avoid chaos — i.e. ChatGPT with shared Custom GPTs, Jasper for enterprise use, or Notion AI. Avoid overlapping licenses and scattered knowledge. Use shared folders or libraries for storing Custom GPTs or prompt templates.
    5. Train the Team
    Run onboarding and upskilling sessions to get people familiar with prompt structure and AI quality boundaries. Show when and how to use AI, and when not to. Keep a Slack or Teams thread open for sharing helpful outputs or workflows.
    6. Govern Usage
    Clearly define what not to do—i.e. don’t push PII to public models or publish AI-generated copy without human review. Consider usage guidelines, review policies, and a log of outputs or major deployments.
    7. Create Feedback Loops
    Make it easy to submit high-performing prompts and share learnings. Track impact—time saved, performance boosts—and keep the AI playbook updated based on those results.

    👎 When Not to Use AI

    Scenario
    Why Not
    Handling Personally Identifiable Information
    May violate data privacy laws (i.e. GDPR, CCPA); risk of unauthorized data exposure.
    Producing Unreviewed External Content
    AI can hallucinate or misrepresent your brand if left unchecked.
    Making Legal or Regulatory Decisions
    Requires human legal judgment and understanding of nuanced precedent.
    Scoring Leads Without Good Training Data
    Poor inputs can result in inaccurate or biased outcomes.
    Automating DEI-Sensitive Communications
    AI may generate insensitive or non-inclusive content.
    Replacing Strategic Thinking or Judgment
    AI lacks business context and the ability to weigh tradeoffs effectively.
    Generating Financial Forecasts Without Review
    Can mislead stakeholders if predictions are taken at face value without validation.

    💭 Final Thoughts

    Again, don’t buy into the hype — stay tuned on the latest, but don’t freeze out of overwhelm. If you haven’t worked with AI at all yet, sign up for a free ChatGPT account and take a mini course like the one on Datacamp. If you are a bit more seasoned, use tools like Clay or in-app AI like HubSpot AI scoring to become more familiar with using AI.
    Either way, just get started and experiment…learn to run the machine, so you (well, hopefully) always have a job. 🤷‍♀️
     

    Related Guides

    👩‍🏫 Ace Your Next Marketing Ops Interview (What MOPs Hiring Managers Are Actually Looking For)

    🕵️‍♂️ Intent Data Use Case Template

    Coffee Kitty

    The Marketing Operations Strategist Newsletter

    Join 3,500+ operations professionals. Get actionable MOPs tips every month.