AI use cases by job: we upgraded Harvard's AI job chart in minutes.
Harvard's interactive chart shows how exposed 911 jobs are to AI. It doesn't tell anyone what to do about it. We used Claude to add two practical AI use cases to every one of those jobs and publish the result as a live, branded tool on our site. Here are the exact prompts, the screenshots and the steps, so you can do the same thing with your own idea.
A simple fix for a simple problem, and a way to think bigger.
This isn't a story about a huge AI project. It's about spotting a small gap in something people already use, then letting AI close it at a scale no person would bother with.
- How to spot a "value gap" in a resource your audience already trusts.
- The exact prompts we used, copied word for word, that you can adapt today.
- Why a screenshot and one worked example beat a long list of instructions.
- How Claude scaled one example into AI use cases for all 911 jobs.
- How to drop the finished tool into WordPress with a single Divi Code module.
- The two features Claude added on its own because it understood the goal.
- How to apply the same pattern to your own industry, data and customers.
Harvard tells you the risk. It stops there.
Harvard Business School Working Knowledge published a strong interactive piece, "Enhance or Eliminate? How AI Will Likely Change These Jobs," based on research by Professor Suraj Srinivasan. Its "How vulnerable is a job to AI replacement?" chart lets you pick a job category and click a bubble to see an automation score for each job title.
It's useful. But when you click a job, here's everything you get: the job title and two scores.
Picture a technical sales rep who sees a 0.62 automation score. The obvious next question is "So what should I do?" The chart has no answer. What's missing is one or two practical, job-specific ways to put AI to work, so the score turns into a plan instead of a worry.
How we built it, prompt by prompt.
Everything below happened in one conversation with Claude, with the Claude in Chrome extension so it could read the Harvard page directly.
Find a resource people already use, and name what's missing
Start with something that already has an audience and credibility. Then write one sentence about the gap. Ours: "It shows how exposed each job is, but not what that person should do with AI."
That sentence becomes the heart of your prompt.
Write the prompt: link, screenshot and one worked example
This is the whole prompt, exactly as we sent it, with a screenshot of the Harvard chart attached:
Why this prompt works:
- It points to the source. The link and screenshot show Claude exactly what to start from.
- It names the gap in one line. "One or two good practical applications of AI specifically for each job."
- It gives one worked example. Two real AI uses for one real job set the tone and level of detail for the other 910.
- It defines "done." Branded, same drop-down behavior, and pasteable into a Divi Code module.
Answer the one question that matters: scope
Claude read the page, found the data behind Harvard's chart and came back with a single question before doing the heavy lifting:
We picked all 911. This is the moment that used to make projects like this impossible. For a person, 1,822 tailored tips is weeks of work. For AI, it's the step that makes the tool worth visiting.
Let Claude build it, then check its work
Claude wrote two AI use cases for every job, built the chart in our brand colors and fonts, and tested it on desktop and phone widths. It also checked its own data against Harvard's: the job counts in all 13 categories matched, and the all-jobs average came out to 0.30, the same mean line shown on Harvard's chart.
It opened on our exact example, with both of our AI uses in place:
Every other job got the same treatment. Here's marketing managers:
Preview it before you publish
The second prompt was four words:
Claude opened the working chart in a side panel so we could click through categories, search and test the AI use cases before anything touched the website.
Paste it into WordPress with one Divi Code module
The output was one self-contained HTML file. To publish it on WordPress with the Divi theme:
- Add a section and a single-column row. Set the section and row margins and padding to 0, and the row width and max width to 100%.
- Add a Code module and paste the HTML in. The chart sets its own max width and padding.
- Point the button where you want leads to go. The "Get your team's AI plan" link is set in one line at the top of the script.
- Add your SEO title, description and keywords, then publish.
All the styling is scoped to the chart, so it doesn't change anything else on the page. The result is live at the AI Job Impact Explorer.
Two features nobody asked for, both right.
Our prompt asked for a copy of Harvard's chart plus AI use cases. Claude understood the bigger goal, which was helping a real person find their job and act on it, and added two features we didn't ask for.
1. A search box for all 911 job titles
Harvard makes you guess which of 13 categories your job lives in, then hunt through a cluster of bubbles. Claude added a search box. Type "Marketing managers" and the chart jumps to Management and highlights that exact job.
2. "Higher score" and "Lower score" buttons
Bubbles are hard to tap on a phone. Claude added two buttons that step to the next job up or down the chart, so you can move through a category one job at a time without aiming at small dots. One tap down from our sales example lands on sales engineers, with their own two AI use cases.
It also added a dashed line for the category average next to Harvard's all-jobs average, so you can see at a glance whether a job is more or less exposed than its peers.
Same data. Far more useful.
| Feature | Harvard chart | CEL AI Job Impact Explorer |
|---|---|---|
| Automation score for 911 jobs | Yes | Yes, same data |
| Filter by job category | Yes | Yes |
| AI use cases by job | No | 2 practical uses for every job |
| Search by job title | No | Yes, all 911 titles |
| Step to next job (mobile-friendly) | No | Higher and Lower score buttons |
| Category average line | No | Yes, next to the all-jobs average |
| Exposure rating in plain English | No | High, Moderate or Low |
| Next step for the reader | None | Link to get a team AI plan |
Think in "times 911," not "times one."
Writing two AI tips for one job is easy. Writing them for 911 jobs is a project nobody starts. That's the shift AI creates: the scale of an idea stops being the reason not to do it.
"I could have downloaded the data and done this myself. It would have taken hours, probably days, to write the tips, and at least a week for a developer to deploy it. This took minutes, and it's far more useful to people now."
Chris Bernard, Founder, Content Engagement Lab
The same pattern works almost anywhere:
- Find public data your customers already care about, such as industry benchmarks, local rankings, regulations or pricing indexes.
- Add the "so what" for every row, one practical next step per item.
- Give AI one excellent example and let it do the other hundreds.
- Publish it as a tool, not a PDF. Interactive pages earn links, time on site and leads.
- Expect AI to improve the idea. When it understands the goal, it fills gaps you didn't think to mention, like search and mobile navigation.
We used the same approach to rank every winery in an entire wine region in one day.
Common questions
What are AI use cases by job?
They're specific, practical ways someone in a given role can use AI in their daily work. For a technical sales rep, that might be using AI with LinkedIn to write custom connection requests at scale. The AI Job Impact Explorer lists two for each of 911 jobs.
Where do the automation scores come from?
The automation and augmentation scores come from Harvard Business School research by Suraj Srinivasan, published in "Enhance or Eliminate? How AI Will Likely Change These Jobs" on HBS Working Knowledge. Content Engagement Lab wrote the AI use cases.
Do I need a developer to build something like this?
No. Claude produced a single self-contained HTML file that we pasted into a Divi Code module in WordPress. Any page builder with a custom HTML or code block can do the same.
Which AI tool did you use?
Claude, working with the Claude in Chrome extension so it could read the Harvard page and its data directly.
Can you build something like this for my business?
Yes. Content Engagement Lab helps businesses find these opportunities and build AI tools, content and automations that bring in customers. Get in touch to talk it through.
See your job's AI use cases, then let's build yours.
Search your job title in the AI Job Impact Explorer. If you have an idea like this for your own business and want help scaling it, we'll show you how.