A friend asked me how I would use AI to find a specific role or contact inside a company without crossing lines that feel sketchy. That is a real problem if you are trying to reach the right person for broadcast, live events, live production, or technical operations.
The goal is not to scrape private information or dig up anything that should not be public. The goal is to use AI as a research assistant to organize public signals from LinkedIn, company sites, and other open sources.
Quick Answer
The cleanest approach I tested was to use DIA Browser by Perplexity with a custom contact search prompt. The prompt tells DIA to look for public information only, prioritize job titles related to the role, and clearly label any email address that is only a pattern-based guess.
For example, when searching Netflix for live events or broadcast contacts, DIA surfaced people with relevant titles such as head of live experiences and broadcast engineer. It also produced possible email patterns, but those should be treated as guesses unless the address is publicly listed.
Why I Used DIA Browser
Normally, I would try this kind of research with ChatGPT, but I wanted a tool that was better suited to searching across pages and organizing web results. DIA Browser, which is made by Perplexity, is useful for this because it can work directly around search and page context.
That said, there is an ethics issue to pay attention to. AI tools can sometimes access or summarize information from websites in ways site owners may not love. Just because a tool can find something does not mean you should use it without thinking.
For this workflow, I kept the search limited to publicly available information and avoided asking for private details. That was the whole point of building the prompt carefully.
The Contact Search Prompt
The useful part of this workflow is not just opening DIA and typing a company name. The useful part is setting up a repeatable prompt that defines the boundaries of the search.
I created a DIA skill called Contact Search. The prompt starts by telling the AI that it is a research assistant and that I will provide a company name. Then it defines the goal: find the best contact who appears responsible for live streaming, broadcast, or live events at that company using only publicly available information.
From there, the prompt tells DIA what kinds of roles to look for. For this example, I used terms like live streaming, broadcast, live events, live production, live experiences, producer, event operations, technical director, broadcast engineer, and live programming.
The prompt also tells DIA to prioritize people who are more likely to make decisions, such as heads, directors, leads, managers, producers, program operations people, or engineering managers.
Where DIA Looks
The search strategy starts with LinkedIn because job titles are often the strongest public signal for finding the right person inside a company. In my setup, I had LinkedIn open and was logged in, which helps DIA work with the context available in the browser.
The prompt also points DIA toward company websites and public social or conference references. The idea is to look for people connected to the relevant work, not just anyone who happens to work at the company.
This matters because a general company contact page usually will not get you to the person who handles live broadcast or technical event decisions. Job titles, team pages, conference bios, and public profiles are much more useful.
Handling Email Addresses Ethically
One important part of the prompt is how it handles email addresses. If DIA finds an email address that is publicly listed, it can include it. If it cannot find one, the prompt asks it to look for a company email pattern and make an educated guess.
That guessed address needs to be clearly marked. In my prompt, I used a field like pattern-based: yes so I know the email was inferred instead of found publicly.
That distinction matters. A guessed email is not the same as a verified public contact. If you use this method, treat those addresses carefully and do not pretend the AI confirmed something it did not.
Example Search With Netflix
For the demo, I used Netflix as the company and asked DIA to find contacts related to live events or broadcast.
DIA returned a small table of top contacts. It found Greg, listed as vice president and head of live experiences, and Valerie, connected to broadcast engineering and live broadcast tech. Those roles matched the kind of search I was trying to run.
It also produced possible email patterns and LinkedIn links. Some of the source URLs looked like they should have been clickable, but they were not clickable during the test. That is one of those DIA quirks you may have to work around by opening the result manually or refining the search.
What Worked And What Did Not
The strongest part of the workflow was role discovery. DIA was good at turning a broad need, like finding someone connected to broadcast or live events, into a short list of people with relevant titles.
The custom prompt also helped keep the results organized. Instead of getting a messy paragraph, the output came back in a table with names, titles, company, email status, LinkedIn URLs, and other contact links when available.
The weak spot was reliability around links and source handling. Some source URLs did not behave the way I expected. Also, because this is AI-assisted research, you still need to verify anything important before acting on it.
- Use DIA to narrow the search, not to replace judgment.
- Treat inferred email patterns as guesses.
- Verify names, titles, and links before reaching out.
- Keep the prompt focused on public information only.
Key Takeaways
- DIA Browser can be useful for finding public company contacts by role, especially when searching LinkedIn-style job titles and company pages.
- A custom prompt makes the search more reliable because it defines the role, seniority level, sources, and output format.
- For broadcast and live events searches, useful title keywords include live streaming, broadcast engineer, technical director, live production, live experiences, and event operations.
- Email patterns should be clearly labeled when they are guessed instead of publicly listed.
- This workflow works best as ethical research assistance, not as a tool for pulling private information.
Watch the Video
The video above for the full walkthrough of setting up the DIA contact search skill, running the Netflix example, and seeing how the results come back inside the browser.
