By Craig Cook
AI customer case studies can now draw on every email, proposal and support ticket a client relationship leaves behind. Still, the marketers who win will be the ones who know what never to publish.
Five executives from a national medical device maker sat around a private dinner table. A chef worked the kitchen. Meanwhile, a multi-camera crew circled the room. It caught every toast and every confession between courses. By dessert, we were sure we had magic in the can.
Then came the footage. Terabytes of it.
For most of my career, I believed the secret to a great customer case study was the interview. First, you ask the right questions. Then the customer talks, and the story emerges. That dinner shoot was a high-production testimonial video I bet heavily on. It also taught me how rarely that formula works, and how expensive the lesson can be.
Why customer interviews miss the real story
Customer interviews capture how a project felt. However, they rarely capture what actually happened. People remember the deadline stress, the late nights and the relief at the finish line. In contrast, they forget where they started. They forget the scope that shifted halfway through. They also forget the numbers that proved the work paid off.
Our dinner shoot proved it at scale. We had interviews from several angles and conversations filmed in several places. We also had B-roll of the dinner and the manufacturer’s plants. The production company had not planned enough time for editing. So we took the raw footage in-house. Then we brought in an Emmy-winning editor from CBS to make sense of it.
The finished film was beautiful. Even so, getting there cost us. We blew the budget and missed our dates. We also spent countless hours building a story that, in hindsight, had existed all along. By the time it was done, the story had lost some relevance. Worse, the customer relationship had cooled.
Time to market had become our worst enemy.
That edit was also my first serious test of AI in a creative workflow. For example, it transcribed hours of talk and found the threads of the story. As a result, a question stuck with me. What if AI could help us find the story before anyone picked up a camera?
There was another cost the budget never showed: the customer’s time. Every hour an executive spends rebuilding a project timeline is an hour away from running the company. Likewise, every week spent chasing one more interview is a week the story sits unpublished. In short, interviews capture a voice. On their own, they seldom capture the full arc.
Why AI customer case studies start with the records
The best customer stories already exist. They are scattered across the everyday records of doing business. Every customer relationship leaves a trail, from the first marketing touch to years of service. Read in order, that trail is the case study. The hard part has always been reading it.
First, the trail starts with a webinar, an article that sparked an inquiry, or a conversation that opened a door. Next, it runs through discovery calls, sales meetings, proposals and statements of work. Then sales hands off to the project team. Requirements change. Change orders appear, and each one reveals something nobody fully understood at the start.
After that come project meetings, technical decisions and support tickets. Then come milestones, setbacks and the go-live. Finally, there are weeks of hands-on support and the months or years of service that follow.
That record holds the original business problem, written before anyone had hindsight. It also shows what changed, why it changed, and how those choices shaped the outcome. In fact, it sometimes holds the most valuable piece of any case study. That piece is results measured when they happened, not rebuilt from memory a year later.
The trouble was never a shortage of stories. Instead, each story lived in a dozen separate systems. Different people kept them, in formats never meant to be read side by side. As a result, weaving them into one story took huge skill, patience and time. So marketers did what marketers always do. We built our stories around the easiest source to reach, even though it was the least complete: somebody’s memory.
How AI connects to a company’s customer records
AI can now link directly to the systems where a company’s customer history lives. These include its customer database, its accounting and operations software, and its shared files. The bridge I use is the Model Context Protocol. It is an open standard that Anthropic released in November 2024. In short, it lets AI assistants connect securely to business tools.
For years, marketers have said that content is king. However, I’d argue we crowned the wrong heir. Content is only as valuable as the context behind it. Yet context was the one thing most marketing teams could not easily reach.
So I began connecting AI to our customer relationship software, our operations software, and the shared files that track our work. I also built better habits for recording customer talks. Together, those connections let me map a relationship from its first marketing touch to today.
I stopped asking people to rebuild the story. Instead, I started curating the one we had already written. That shift is the real promise of AI customer case studies. It is part of a wider change in how New York companies are making real money with AI.
What AI should and should not do in customer case studies
In AI customer case studies, the AI works as an archivist, a researcher and a pattern-finder. It should never invent a story or inflate a result. Rather, its value is finding links buried in years of records. Each link comes with a document a person can check before anything is published.
For example, an issue raised in an early sales call may have been fixed eighteen months later. A change order may reveal a major shift in the customer’s business. Similarly, a support ticket may record the fix for a problem nobody mentions in an interview. By then, after all, it is ancient history.
What comes out is the frame of an honest story. It rests on details that interviews alone would likely have missed. As a result, the stories are richer and more faithful to what the customer lived through. Better still, they take a fraction of the time to develop.
Why more data access demands more restraint
The same AI that can read every customer record can also expose information a company has no right to share. Customers hand over their processes, finances, problems, plans and competitive edges. That trust does not come with a license to publish.
So, as I built this approach, I set guardrails. They cover what the AI can see, how it can be used, and what can appear in a published story. Trade secrets stay private. Likewise, confidential financials stay private. No performance number makes it into marketing just because it sits in a customer record. We verify every claim, and sensitive details need explicit approval.
Permission also became part of the process itself. Many of the companies we serve are public or backed by private equity. For them, a case study is a careful disclosure decision, not a casual favor.
Besides, consent is rarely a simple yes or no. For instance, approving a logo on a partner website is different from approving a CEO’s quote in a paid ad. Likewise, being named in a blog article is different from appearing in a video at an industry conference.
That is why we now track exactly what each customer has approved. This covers name, logo, story, approved numbers, employee quotes, photos and video. Just as carefully, we track where each item may appear. We separate editorial content from promotions, paid media and advertising. Permission for one use never quietly extends to another.
Of everything in this process, that may be the most important lesson. Access is a capability. Restraint, however, is a responsibility, and it counts for far more.
What a customer interview is still good for
The customer interview survives. In fact, it finally gets to do the job it does best: capturing emotion and meaning. When the records supply the facts, the interview no longer has to rebuild a timeline. Instead, it can ask how the work felt, what changed for the team, and what the customer would tell others.
The record provides the facts. People provide the heart. When I sit down with a customer now, I have already studied where they started. I also know what changed along the way and what the evidence says they achieved. So I can ask the questions that matter. What did that moment feel like? How did the change affect your team? What surprised you? What would you tell someone facing the same problem today?
No database entry can answer those questions. Similarly, no support ticket will ever capture them. Because I arrive informed, customers spend less time repeating themselves. As a result, they feel understood instead of questioned. Then we can explore what the experience meant instead of arguing over dates.
The marketers who will win the next decade
Modern marketing is usually sold as a race for more. More content, more channels, more data and more automation. Yet most businesses already hold more information than their marketing teams could ever use. Inside it are stories of change, ingenuity, partnership, persistence and real results.
What has been missing is the ability to connect the dots and read the context. Equally missing is the sense to know which stories deserve telling. Just as important is the judgment to know which parts of those stories are ours to share.
Technology alone will not make anyone a better storyteller. That still takes curiosity, empathy and editorial judgment. Done right, AI customer case studies simply give that judgment better raw material. It also matters more now, since what AI tools say about your brand often shapes the sale. It also takes the discipline to separate what we can tell from what we should.
Think back to that dinner table: five executives, a private chef, terabytes of film. The story we spent months chasing was sitting in our own files the whole time. It was simply waiting for someone to read it in order. Ultimately, the marketers who define the next decade will earn access to that context. Then they will prove they can be trusted with it, and turn it into stories worth telling.
Frequently asked questions
How can AI help write customer case studies?
AI customer case studies draw on a customer’s full history. The AI reads a company’s records, project files, support tickets and proposals. Then it surfaces the original problem, what changed, and the results recorded along the way. A marketer still writes the story and checks every claim against the records.
Do you still need to interview the customer?
Yes. The records supply facts and timelines, while the interview supplies emotion and meaning. Because the history is already studied, the interview can focus on how the project felt. It also takes less of the customer’s time.
What customer data should never appear in a case study?
Trade secrets, confidential financial information, and any number the customer has not approved for publication. A figure sitting in a customer record is not permission to publish it. Therefore, sensitive details need written sign-off for each specific use.
What is customer content permission tracking?
It is a record of exactly what each customer has approved. That includes name, logo, quotes, numbers, photos and video. It also records where each item may appear, such as articles, promotions, paid media or ads. Approval for one channel does not carry over to another.
Sources
- Introducing the Model Context Protocol, Anthropic, November 2024
- Anthropic introduces the Model Context Protocol, InfoWorld, November 2024

