Using AI in marketing workflows is completely changing how brands engage with their audience, especially when it comes to customer feedback. We recently ran a campaign for a B2B SaaS client in the project management space where we didn’t just collect customer insights, we used them to generate high-performing AI content. The goal was to prove a direct line between what users were saying and our conversion rates, showing that authentic feedback is the best source for great content creation. The big question for us was, can AI really turn the customer’s voice into marketing that actually works?

Campaign Overview: The “Voice of the User” Initiative

Our client, a B2B SaaS company selling project management software, had a problem we see all the time: a goldmine of customer support tickets, forum posts, and survey answers that were just sitting there, completely unused by marketing. Their content strategy was all internal brainstorming and chasing industry trends, which meant they were missing the nitty-gritty details you only get from direct user feedback. We launched the “Voice of the User” campaign in Q2 2026 to fix this, creating a system to turn all that feedback into content for a new set of team collaboration features.

The whole thing ran for 10 weeks on a $75,000 budget. Our main goals were simple: get 20% more trial sign-ups for the new features and bump up the CTR on our promotional content by 15%. Our bet was that content written in the users’ own language, talking about their specific problems, would resonate way more than our usual marketing copy.

Strategy: From Raw Data to Refined Messaging

Our strategy had a few key stages: collect the data, let the AI analyze it, generate the content, and then get it in front of the right people. We pulled feedback from everywhere, in-app surveys, support chat logs, public review sites like G2 (g2.com), and even some direct interviews with power users. Putting all that data together gave us a really deep, 360-degree view of user sentiment and their biggest headaches.

For the analysis part, we used a natural language processing (NLP) platform we had trained on B2B SaaS jargon. Its job was to spot recurring themes, sort sentiment as positive, negative, or neutral, and pick out the exact phrases people used when talking about collaboration problems. This platform could chew through thousands of feedback entries an hour, a job that would’ve taken our team weeks to do with the same level of detail. The AI then ranked feedback by frequency and emotional intensity, pointing us straight to the things that made users either really happy or really mad. For instance, it quickly flagged “version control nightmare” and “smooth file sharing” as high-frequency terms with a lot of emotional weight behind them.

From there, we fed those themes and key phrases into a generative AI content platform. We tasked it with creating a bunch of different content formats for us, like short-form ad copy, blog post outlines, email subject lines, and video script ideas. The key instruction was to use the *exact* language and problem-solution framing we’d found in the customer feedback. We weren’t asking it to write about “collaboration”. We were telling it to address “the version control nightmare” with a solution for “smooth file sharing.”

Creative Approach: Speaking the User’s Language

Our creative team worked hand-in-hand with the AI’s output. We didn’t treat this like some automated content factory. The AI was our first-draft generator and idea machine, and a very efficient one at that. For example, the AI flagged a common complaint about “time wasted searching for files across different platforms.” That insight was immediately turned into ad copy variations like: “Tired of the file hunt? Our new collaboration hub centralizes everything. Save 3 hours a week.”

We ended up running two main creative tracks:

1. Problem-Solution Focused Ads: These ads quoted or paraphrased the user pain points the AI found, then immediately showed how the client’s new feature fixed that exact problem. The visuals backed this up, showing a transition from a cluttered, frustrating desktop to a clean, organized workspace.
2. Benefit-Driven Content: For longer stuff like blog posts and emails, we went deeper into the benefits of solving those problems. We even used user testimonials (with their permission, of course) that repeated the themes the AI had surfaced. A blog post titled “How ‘Unified Workspace’ Solved Our Version Control Headaches” used language pulled right from user feedback, which made it incredibly relatable.

We also used the AI to rapidly spit out dozens of variations for headlines and calls to action, which let us test different angles focusing on emotional triggers or specific time-saving claims we had pulled from the user data.

Targeting and Distribution: Reaching the Right Ears

We pushed the content out on the channels where our target audience, project managers, team leads, and ops managers in mid-sized tech companies, actually hangs out. That meant LinkedIn Ads (business.linkedin.com/marketing-solutions/ads), Google Search Ads, and targeted emails to our existing lead list. The AI’s insights shaped both our ad copy and our audience segmentation. For example, when feedback from sales-heavy companies kept mentioning “integrations with CRM,” we created a specific ad set just for them with content that screamed CRM integration.

On LinkedIn, we built lookalike audiences from our existing customer list and targeted by job title and industry. For Google Search, we moved beyond broad terms and focused on long-tail keywords the AI had pulled from user queries and support tickets, like “best project management software for remote teams file sharing” or “how to stop losing files in team collaboration.” It’s a much more direct way to find people who are actively looking for a solution to a problem they can articulate.

Results: What Worked and What Didn’t

The campaign gave us some solid insights, and a few results that honestly surprised us. Here’s the breakdown:

Performance Metrics: “Voice of the User” Campaign (10 Weeks)

What Worked:

• Hyper-Relevant Messaging: Our biggest win, by far, was the huge jump in CTR. Ads that hit on specific user pain points in their own words just took off. An ad that said, “Stop the endless email chains for document approvals, our new workflow automates it,” hit a 3.5% CTR, which blew our averages out of the water. This pretty much proved our theory that content based on real feedback just works better.
• Reduced Content Generation Time: The whole ideation and first-draft process which usually bogs our team down for two weeks, was done in three days. That freed up a ton of time for our writers to focus on polishing the content and thinking strategically instead of just trying to fill a blank page.
• Lower Cost Per Conversion: Better relevance meant a higher conversion rate, which brought our costs down. Our CPL dropped by nearly a third, a direct result of more qualified leads clicking and converting. That drop in CPL is where the ROI of this whole AI-plus-feedback approach became impossible to ignore.

What Didn’t Work as Expected:

• Overly Technical AI Output: In the beginning, the AI-generated copy sometimes got way too technical because it was pulling directly from support tickets. The stuff made sense to a developer, but not to a broader marketing audience. We quickly learned to add a human editing step to translate the “tech-speak” into plain English. It was a good reminder that AI is an assistant, not a copywriter.
• Sentiment Misinterpretation on Nuance: The NLP was mostly accurate, but it sometimes got tripped up by sarcasm or complex feedback. A user might write, “Great, another tool to learn, just what I needed,” and the AI would tag it as neutral or positive because it saw the word “great,” completely missing the sarcasm. This meant we had to manually check a small percentage of the feedback to catch these subtleties. It just goes to show that context is everything, and the AI still has its blind spots.

Optimization Steps Taken

We learned a lot on the fly and made a few key adjustments during the campaign:

1. Human-in-the-Loop Refinement: We created a formal two-stage review process. The AI would generate the first drafts and highlight the main themes, but our human content strategists always did the final polish. They were responsible for making sure the tone, brand voice, and clarity were on point. This hybrid approach worked best.
2. Negative Keyword Expansion: We saw that some of our search ad clicks were from people looking for very basic, free tools, which wasn’t our target. By looking at the search terms that led to these low-quality conversions, we built out our negative keyword list and got much better at filtering out the wrong audience.
3. Dynamic Content Personalization: For our email campaigns, we started playing with dynamic content. For example, if a user’s past behavior showed they were interested in “integrations,” the AI-generated email copy would automatically put the section about our collaboration feature’s integrations front and center. We did this by connecting our feedback analysis tool to our marketing automation platform, it took some upfront configuration, but the payoff in engagement was huge.
4. Feedback Loop for AI Training: We set up a continuous loop. Whenever a piece of AI-generated content did really well (high CTR, low CPL), we fed that data back into the model as a “win.” This helped the AI learn what kind of messaging was resonating. We did the same for underperforming content, teaching it what to avoid.

This campaign showed us that using AI to analyze customer feedback isn’t some academic exercise. It’s a practical, high-impact strategy for creating much better marketing content creation. Being able to quickly turn what users are feeling into actual, actionable messaging gives you a real competitive edge, leading to campaigns that people actually click on and a much better return on ad spend.

The future of marketing content that works is all about deeply understanding your audience, and AI provides an incredible lens for that. When you systematically bake customer insights into your AI content workflow, you move beyond generic messaging to speak directly to your customers’ needs. This is what drives real business results. My advice? Start small. Pick one channel, run an experiment, and let the data tell you where to go next.