Everyone is worried about AI, and we get that. But the fears of job loss are, at least so far, overblown. So are the pessimistic studies about most AI pilots failing. (It’s often just their sample and definition.) Instead, the reality is most marketing teams occupy a transitional space: They are trying hard to use AI but the truly transformational use cases seem just beyond grasp—if only they knew the exact steps.

This guide will provide those steps. Everything here is explained in enough clarity that your team can enact most of them with only a consumer-grade LLM subscription. Though this advice does come with a caveat: Every business is different. If there really were one “playbook” for 10x more pipeline, rest assured, we’d all be commenting “playbook” on that influencer’s post. 

Real AI only works if you have real marketing fundamentals:

• Know your customer
• Know your business
• Know what you want

The five use cases in this guide all assume you have figured out at least some of those fundamentals. We hope it gives you all the AI use cases you need to improve this quarter or season, and that the results are big enough to measure in engagement, pipeline, and revenue. 

What’s inside: 

• 5 specific and credible use cases
• Step-by-step instructions
• Pro tips from real pros

Use case 1: Understand how your customer thinks you’re differentiated

AI cannot miraculously tell you who your ideal customer is and how they talk—that’s not in the dataset. But it will take all the materials you’ve produced and tell you how they might react, for far more empathetic marketing. 

What you’ll need:

• A general AI tool
• A written ideal customer persona document
• Labeled transcripts of calls with that persona
• Labeled transcripts of calls with wrong-fit buyers

The goal of this exercise is to approximate how your ideal persona (ICP) would react to something by scanning a vast swath of publicly available reviews and online commentary. You want to figure out:

• How they describe the problems you solve in their own terms
• Their perceptions of how you and your competitors solve them

For this to work, you need to know your ideal customer. Not just who they are, but which segments you’re the most likely to retain and are the most profitable—measured by customer lifetime value (CLV). If you know that, and have call transcripts of those individuals, you can use deep research (available in ChatGPT, Gemini, Claude, Perplexity, and others) to analyze your voice of customer compared to your top three competitors. 

You’ll ask that LLM to take in your persona information and then go scan forums and review sites such as Reddit, Quora, G2, and TrustRadius. You’ll then ask the AI to perform a strengths, weaknesses, opportunities, threats (SWOT) analysis plus a few other methods. By the end, you should have a deck full of information about your differentiation in the eyes of your ideal buyer, to inform messaging and outreach. 

Step-by-step

1. Create a project and enter these instructions
   
   1. When called upon, you will conduct a comprehensive SWOT analysis by examining four key areas: Strengths: Identify internal advantages, capabilities, and resources that give a competitive edge. Consider expertise, assets, market position, and unique qualities: Weaknesses: Analyze internal limitations, gaps, and areas needing improvement. Be honest about resource constraints, skill deficiencies, or operational challenges. Opportunities: Explore external factors that could be leveraged for growth. Look for market trends, emerging technologies, partnerships, or unmet customer needs. Threats: Recognize external risks and challenges that could cause problems. Consider competition, market changes, regulatory issues, or economic factors. Present findings in a clear, organized format. For each category, provide specific, actionable insights rather than generic statements. Prioritize the most significant items and explain their potential impact.‍
2. Upload your materials to its knowledge base
   
   1. A written ideal customer persona document
   2. Labeled transcripts of calls with that persona
   3. Labeled transcripts of calls with people who do not fit that persona‍
3. Customize and enter this prompt ‍
   
   1. Study my ideal persona and try to inhabit how they view the world and environment within their work context. Then scan the internet and especially review sites for instances of them sharing their thoughts and feelings about vendors in our space [enter brief description]. Return a SWOT analysis for our company in winning more market share from these buyers.

Use case 2: Analyze your segments

Business intelligence platforms are complicated. The questions most marketers need to answer with them aren’t. Instead, you should just talk to your data.

What you’ll need:

• Access to download info from your marketing automation platform, ads platforms, and CRM
• A general AI tool

In this exercise, you’ll use a general AI platform to analyze data you’ve exported from your marketing, sales, and advertising systems. (You can also use an AI-native platform like Akkio or NinjaCat to do this much more quickly.) We find this saves marketers from the typical doom loop:

• You have a simple question
• The dashboard doesn’t quite answer it
• The BI team is too busy to help
• You binge YouTube videos try to create your own
• 9 hours later you have forgotten the question

This solution is a bit of a hack, but rest assured, more stable and reliable tools for this are coming. In essence, this allows you to “talk” to your database. Ask it: 

• Which segments are performing? 
• Which segments are underperforming? 
• What channels are most likely to result in a conversion for ABM programs?
• Can you visualize my top metrics in a four-quadrant matrix? (x = CAC, y = NRR)

As with all AI data exercises, trust but verify. This should be a method for leaping to insights that you can then ask the BI team to verify.

Note: This is a great way to write specs for your BI team. It reduces the back-and-forth when you can say, “This is the exact data I want to view.”]

Step by step

1. Write down your top 4-5 questions.
2. Export all the relevant data it would need to answer them.
3. Create an AI project and upload the exported data.
4. Confirm it can understand the data—give it simple, verifiable tests.
5. Ask your questions.

Use case 3: Generate more empathetic messaging

We believe deep research is underused. It can expand your persona and messaging work to tailor it to each segment or account in ways that delight and surprise.

What you’ll need: 

• Persona research
• A general AI tool

Behind every generic buyer pain like “time savings” are five more specific pains. As marketers, we often only see that first pain because that’s what customers tell us on calls. But deep research (a feature available in most major AI tools) enriches our understanding of what’s happening when we’re not in the room.

We have experimented with this at Inverta on client projects and we have noticed that the resulting messaging is much stronger on the first pass. Many clients approve those orchestration plans and content faster as a result. This is because deep research helps us go layers beneath the typical, “The buyer is trying to save money,” to something more like, “Manufacturing heads of finance are really distraught over tariffs and all the complexity of harmonized codes (HC).” The deep researched version lets us do in hours what might have taken weeks. We then verify by reading the citations and forming our own conclusions.

Actual client use case

A large payroll provider was targeting hospitals and healthcare providers with software to help unite their disconnected HR software systems. “Uniting systems” wasn’t quite a compelling reason, though. We used Gemini’s deep research to dig into that pain point and learn that hospitals are at risk of losing funding due to improperly licensed staff. A loss of funding is catastrophic—and existential. This informed our messaging, and the head of sales told us, “This framework captures the hospital space better than any other I’ve seen.” 

Step by step

1. Upload your persona research 
2. Use this prompt:
   
   1. Our buyers tell us they want to [generic pain point]. Please research thoroughly online to understand the full list of pains and motives that might be behind that statement, citing your sources. If you cannot find truthful and accurate information, say so and ask for clarification or guidance.  

Use case 4: Get writing feedback from a synthetic persona

We’ll look back on this time in marketing and think, ‘Wow, we used to do so much guesswork.” That’s because most marketers really don’t know if their writing was any good until the email sends or campaign launches. 

What you’ll need: 

• Persona information
• A general AI tool

Human writing usually beats AI writing. But AI is useful in scenarios where you don’t have a sparring partner to bounce ideas off, and to take the point of view of the customer. In this exercise, you’ll use an AI project to upload all your persona information (you already did this in use case one) and give it instructions to embody that persona. You’ll ask it to take on that type of buyer’s characteristics, knowing the pressures they are under, the jobs to be done, and to speak as they’d speak, citing evidence.

Use the resulting feedback to end those interminable cycles of everyone feeling “the copy isn’t quite there.” In truth, everyone internally may not know. They haven’t talked to enough customers lately and committees are an awful way to make content decisions. A synthetic persona can give every person a lookalike editor to arrive at stronger copy on the first pass.

As the joke goes, “This 32-person committee feels this email lacks a coherent voice.”

Actual client use case

A training software company we work with lost a few people from their marketing team—and they happened to be the two people who could write most clearly. Suddenly, nobody had the bandwidth or depth of knowledge to review all communication. We helped them build a synthetic persona in Copilot to represent their primary buyer, the chief operations officer. This got them producing and reviewing emails again at enough capacity to support the business. Email health scores actually improved over what they were prior.

Step by step: 

1. Create persona (see Use case 3)
2. Run your copy through it
3. Always verify the result

Use case 5: Build a quality content engine

Teams forced to do more should not be spending time on formulaic, time-consuming tasks where each output is low value—like blog articles. You need those assets in aggregate, and you need them fast.

The truth about most blogs is they’re a lightly reskinned version of a prior blog or similar insight. Or a derivative of some other information source, such as original research or webinars. When you think about it that way, the modern content team looks a lot more like researchers and archivists than just writers—and the more they focus on those origins studies and thought leadership, the more help they need automatically writing those derivative articles, ads, and social posts.

That’s why we’re so avid on building your team an AI editor project. It is important that it is shared, and everyone isn’t just using their own instance—you want everyone to benefit from what anyone knows. A shared AI editor can help your team research those reports to produce their own studies and original information. Then it can help produce a “metadata readout” of all the derivative copy that takes so much time: Meta descriptions, YouTube descriptions, LinkedIn posts, emails, and so on.

When you’re doing this well, you’re chaining all of these together and letting them feed each other:

• You feed your competitive differentiation and market research findings into your AI editor project.
• Your AI editor project helps advise on wording and generate derivative outputs.
• You feed the results of those campaigns back into the editor.

Step by step

1. Create a project and name it “AI editor.”
2. Feed it all your persona information.
3. Feed it all your brand, voice, and style instructions.
4. Upload a document containing good examples of each format you want it to write in, properly labeled (copy from content that already exists).
5. Ask it to describe each of those content pieces and write improved instructions for itself—upload those new instructions.
6. Use it for research.
7. Use it for derivatives.

Bonus: How do I staff for this new world?

As work changes, so too must your org chart. Be sure yours changes in ways that make your teams sharper, not less.

A common complaint we hear from content teams hiring junior writers is “They don’t know how to write without AI.” If there’s an AWS outage or they’re challenged live in a meeting, they can’t respond—they don’t know how to think critically. This is an AI risk: Using it in the wrong ways can enfeeble your team and make them dependent, rather than what the most effective teams achieve, which is a team of segmenting assistants that free them to maximize their human creative skills. 

While we don’t yet have an org chart to propose, here’s what we believe to be true about that future AI-powered marketing team:

• Technical skills will still matter—AI sometimes runs into a “last mile” problem where it gets you 95% of the way there, but then you get stuck on a technical issue. For example, having to write your own Excel macro or edit a photo. Non-technical teams get stuck a lot more.
• Expertise matters more—Does your team know enough to accurately judge AI output? Marketers with long experience studying human behavior have an inherent advantage. 
• Soft skills matter more—Hire curious people who are clear communicators. AI is reliant on text or spoken communication, and the more precise your team is about those inputs—and the more curious they can get about why they got the outputs they got—the better they’ll do.
• People do the work, AI assists—If something can credibly be done by an AI and nobody would notice the difference in output, it’s not a task your team should be doing. 
• Humans remain responsible—A human is always accountable even if the AI agents are working autonomously. Just as you nominate one owner of every project, nominate an owner of every AI output. 
• Get creative about hiring juniors—Many junior jobs will disappear, and so will that on-the-job training. But you will desperately miss the diversity of thought they bring, and risk amplifying AI’s biases. Build better training and mentorship programs so you can maintain your multi-generational workforce. 

Which use case will you try first?

This guide aims to get you succeeding with at least one use case, with step-by-step instructions, so there’s no guessing. Again, if one playbook worked for all, we’d all be pushing that button. But your success in AI relies on your team’s curiosity about the outputs they’re getting as well as experimentation. If you aren’t getting the outputs you’re hoping for, great—that’s the first step to getting the ones you want. 

To troubleshoot: 

1. Investigate the inputs

Could it be the quality of your persona information? Segments? 

2. Fine-tune the settings

If you get counterintuitive responses, or the AI won’t seem to stay on topic, look into the settings. Is there a feature for “randomness” that you can turn down? Does removing some instructions fix the issue? 

3. Verify the outputs

If some responses seem wildly counterintuitive to your team, test them on a friendly customer call or with a buyer lookalike. 

Good marketing relies on strong personas

Want rock a rock-solid ideal buyer persona (ICP) foundation upon which to build all your marketing? We’d love to help.