How to Use AI to Find New Business

Most owners I talk to do not have a lead problem. They have a searching problem.

The work is out there. It is sitting in public bid portals, city council packets, building permit lists and the local business journal. Nobody has the hours to read all of it every week, so the opportunities get found late, after a competitor already has a relationship with the buyer.

I spent ten years in restaurants and beverage, ending as head of sales and marketing for a brewery, and today I am President and COO of Uniti Med and GQR Healthcare and COO of Nebula, an AI-powered talent sourcing company. In every one of those jobs, the person who found the opportunity first usually won it. This is the first of six areas in my workshop Take Back Your Time with AI, and it is the one that pays for the others.

Where does new business show up first?

Before a project is a bid, it is a signal. A company announces an expansion. A property sells. A building permit gets filed. A planning commission puts a project on its agenda. Someone posts in a local group asking who people recommend. By the time a formal request for proposal appears, the early movers have already had a conversation.

For a contractor, the places to watch are bid platforms, public projects, permits and planning, expansion news, recommendation requests and contractor directories. Other businesses have their own versions. A commercial cleaning company can watch for new office leases, building sales and businesses announcing a move. An accounting firm can watch for new business registrations and companies that just raised money or opened a second location. The question is the same for all of them: what public event usually happens a few weeks or months before someone needs what I sell?

Write those signals down before you open an AI tool. They become the “Focus on” line of the prompt.

What prompt should you use to find opportunities with AI?

This is the research prompt from my contractor workshop. In the original, the business was a commercial mechanical insulation contractor and the territory was Nebraska, Iowa, Kansas, Missouri and South Dakota. Swap in your own business and area, and change the focus list to the signals you wrote down above.

Act as a business development researcher for a [your type of business].

Find current, publicly available project opportunities and early indicators of future work in [your states or service area].

Focus on public bids, school and hospital projects, industrial expansions, permits and planning agendas, facility announcements, and visible contractor requests.

For each result, provide:
- Project and location
- Current stage and date
- Why it may need our services
- Public contact and source
- Recommended next step

Separate the results into:
READY TO PURSUE
EARLY OPPORTUNITY SIGNALS

Verify every result with a public source.
Clearly label any inference.
Do not invent projects, dates, contacts, or scope.

Look at what the prompt asks for. It does not say “find me leads.” It names a role, a specific task, the places to look, the exact fields to return, and the rules. The two-bucket split matters too. Some results are ready to pursue today. Others are early signals worth watching. Mixing them together makes the list feel bigger than it is.

The guardrails at the end are the most important lines. AI tools will fill gaps with confident guesses if you let them. Asking for a public source on every result, and telling the tool to label anything it inferred, is what makes the list usable.

How does this example work, start to finish?

Real opportunities. Public information. AI searched. You decide what is worth pursuing.

What do you do with the list once AI finds it?

Pick one. That sounds too simple, but a list of 25 opportunities usually produces zero calls, because nobody knows where to start. I would rather see an owner pursue one opportunity properly this week than skim twenty.

For a public project, the next move looks like this:

A list does not create revenue. The next action does.

What should the first-contact email say?

Short, specific and easy to answer. The best first email after an AI-found opportunity names the project, asks one real question and makes replying take ten seconds. For a contractor, that question is often whether the scope includes their service. For a cleaning company, it might be whether the new tenant has chosen a janitorial provider for the space yet.

You can ask AI to draft it. Give it the project details you verified and ask for a three-sentence email that references the project by name, asks one question and offers one simple next step. Then read it as if you were the person receiving it, and cut anything that sounds like a mass email.

That number is why the research step matters. AI makes it easy to send a hundred generic emails, and buyers punish exactly that. The point of the prompt is the opposite: a specific project, a relevant message and a clear next step.

What does a person still need to check?

Everything that leaves your building. In the workshop I call this step “what we check,” and it is the step that makes AI safe to use in a real business.

Where should you start this week?

Run the prompt once for your area. Pick one result from the “ready to pursue” bucket, verify it, and send one specific email. Then run it again next Monday. Ten minutes a week of AI research, plus one real follow-up, will put you in front of more early conversations than most of your competitors.

When the opportunity turns into a request for a quote, the next area in this series takes over: How to Use AI to Respond to Quote Requests Faster. For the selling side of what happens after that first conversation, see How to Increase Sales: An Operator’s Playbook.