Skip to main content
Blog

How I Use AI in My Own Business: The Workflow That Preps My Sales Calls

Part 1 of 5 in my series How I Use AI in My Own Business: real systems I run at Hub Media, what broke, and what you can copy.

I sell AI systems to business owners, so it’s fair to ask what I actually use myself. This series answers that one build at a time, starting with the newest: the workflow that books my sales calls and preps me for them.

The problem: a speed-to-lead gap on a speed-to-lead offer

My main entry offer is the AI Opportunity Audit. A business owner fills out a short form on hub.media, we talk, and I come back with the automations worth building first.

When I looked at that form with fresh eyes, here’s what happened after someone hit submit: a thank-you message, then “David reviews,” then nothing. No booking step. The owner waited for me to email them, and then we traded messages to find a time.

That’s embarrassing for someone who writes about how fast law firms should respond to a new lead. Interest is usually highest right after someone asks. I was making people wait anyway.

The second problem was prep. Before a discovery call I want to know what the business runs on: its website platform, how it takes leads, whether it’s running ads, what its reviews say. I’d planned on about 15 minutes of manual digging per call. Sales trainer Jeb Blount, who argues hard against over-researching cold calls, still says that for a discovery meeting you should “absolutely do your research.” I agree. I just didn’t want to do it by hand.

What I built in one afternoon

I built the fix in one afternoon, with Claude as my coding partner and n8n, a workflow automation tool, as the engine. It went live on November 19, 2026.

The pieces:

  • The form: Gravity Forms Basic on my WordPress site, with a short PHP snippet that sends each request to n8n instead of a paid add-on.
  • Booking: a Calendly widget on the thank-you page, so the call gets booked in the same visit.
  • Research: n8n gathers public information about the business, with SerpApi handling the search pieces.
  • The brief: an OpenAI model turns that research into one page.

How a request flows through it

  1. Someone requests an audit. n8n logs the lead and emails me right away.
  2. They book on the spot. The thank-you page shows my calendar. No back-and-forth.
  3. The research runs when they book. n8n reads their homepage for the tools they use: site platform, forms, chat, booking, call tracking and ad pixels. It pulls their Google rating, review count, newest reviews and how often the owner replies. Then it checks for active Meta and Google ads.
  4. A one-page prep brief lands in my inbox, usually within about a minute of the booking.
  5. The day before the call, if the booking still stands, I get a mystery-shop checklist: submit their website form, call after hours, and see what happens.
  6. If they don’t book within 24 hours, a follow-up email is saved as a Gmail draft for me to review and send.
  7. Cancellations are tracked, so I don’t prep for a call that isn’t happening.

One detail shows how this stuff works in real life. For the ad check, I use SerpApi instead of Meta’s official Ad Library API. Meta’s API documentation says ads that didn’t reach any location in the EU “will only return if they are about social issues, elections or politics.” Ask it about a Sarasota dentist’s regular Facebook ad and you get nothing back. Plan around what each tool actually does, not what its marketing page suggests.

The guardrails

Prep is low risk, since nobody sees the brief but me. I still set rules, because the habits you build on low-risk work carry over to the high-risk work.

Nothing goes to a prospect automatically. The one message that could go out on its own, the 24-hour follow-up, is saved as a draft. I read it and hit send myself. It’s the same rule as the inbox agent I built for another business owner: AI drafts, a person approves.

The AI only uses facts it was given. OpenAI’s own help center warns that ChatGPT “can even make up things like quotes or citations.” So the model gets the research and nothing else, and it’s told to write “Not found” for anything missing. A brief that says “Google rating: Not found” is useful. A brief that invents a rating is dangerous, because I’d walk into the call repeating it.

Public business information only. It reads the homepage, the public Google profile and public ads. No logins, no personal social accounts, no purchased data. If you put customer information into AI tools, check which account and settings you’re using first.

It doesn’t decide anything. The brief tells me what to look at. What to recommend is my job on the call.

What broke on the first live test

The first real booking surfaced two problems.

First, the AI account ran out of credits. The research ran fine, the step that writes the brief failed, and I got nothing. Now, if the AI step fails, the workflow emails me the raw research instead. It’s not pretty, but I still walk into the call prepared.

Second, the steps were in the wrong order. Booked status was saved after the research ran, so when the AI step failed, the booking never got recorded. Now booked status is saved first, then the research runs.

Neither fix was clever. Both came from watching a real run instead of assuming it worked. Expect a first-week surprise from any automation, and make sure a failure tells you something instead of going quiet.

What it changed

The booking gap is closed: anyone who asks for an audit can grab a time in the same visit. And most of the roughly 15 minutes of prep I’d planned per call now happens before I open my inbox. That’s my estimate, not a measured result. I’ll share real numbers in this series once there’s enough history to count.

A smaller version you could set up this week

You don’t need n8n or a developer to get most of the value. If you take calls with prospects, try this:

  1. Put a booking link on your thank-you page or confirmation email. This alone fixes the wait. Most scheduling tools make it a copy-and-paste job.
  2. Write down your prep checklist. What do you always look up before a call? Their website, their reviews, whether they advertise, who you’re talking to.
  3. Turn it into a prompt. Paste the public information you find into an AI tool on a business account and ask for a one-page summary, with the instruction: “Use only what I gave you. Write ‘Not found’ for anything missing.”
  4. Automate one step at a time. Once the checklist is stable, tools like n8n or Zapier can start it from each new booking. n8n’s Calendly trigger, for example, listens for events created and canceled.

Start with the booking link. It’s the step most businesses skip and the one prospects notice.

For professional firms, the same pattern works for consultations: intake form, instant booking, a short brief before the meeting. See how we approach AI for professional services, or read more about our AI implementation work.

This is the call it books

If you request an AI Opportunity Audit, this is the workflow you’ll meet. You’ll book a time on the spot, and I’ll show up having already looked at how your business gets and handles leads. We’ll talk through what’s working, where things slip, and the 2–3 automations I’d build first.

Next Friday, part 2: the morning brief my AI writes every workday, and what it isn’t allowed to do.

Frequently asked questions

What should I research before a discovery call?

Look at what the business runs on and how it handles leads: its website platform and forms, whether it has chat or online booking, whether it runs Google or Meta ads, and its Google rating, newest reviews and how often the owner replies. Then test the experience yourself, for example by submitting their website form or calling after hours.

Can AI prepare for sales calls for me?

Yes, for the research and summary. A workflow tool such as n8n or Zapier can start when a call is booked, gather public information about the business and have an AI model turn it into a one-page brief. The judgment on the call, and what you recommend, stays with you.

How do I stop AI from making things up in a prep brief?

Give the model only the research you gathered, tell it to use nothing else, and have it write ‘Not found’ for anything missing. OpenAI’s own help center warns ChatGPT can make up things like quotes or citations, so check anything you plan to repeat on the call.

Do I need a developer to automate sales call prep?

Not to start. A booking link on your thank-you page and a written prep checklist you paste into an AI tool on a business account cover most of the value. Automating the research from each new booking with a tool like n8n or Zapier is the next step once the checklist is stable.

Should AI send follow-up emails to prospects automatically?

Not in Hub Media’s workflow. The follow-up for someone who requested an audit but didn’t book is saved as a Gmail draft, and David reviews and sends it himself. Keeping a person between AI and anything a prospect sees is a simple rule that prevents most mistakes.