I spent two months interviewing 40 to 50 CROs about how they use AI in their sales work. Not what they post about it, but what they still had running a few months later. Four use cases survived, and I was surprised by how few of them had anything to do with writing emails.
That's the first thing most of us reach for, and it turns out to be the smallest job you can give it.
You can watch the video above or read on. Everything in it is written out here, including the prompts.
First, Your AI Has No Memory of Your Business
When you open a new chat and start typing, you're living inside that one conversation. Once you run out of room in it, it closes, and everything you explained about your business goes with it. So you explain it again next week, and you pay tokens for the privilege.
This doesn't mean you're using AI wrong. A plain chat was never built to remember you.
What fixes the memory problem is working inside a project that's connected to a folder on your computer:
- Create a folder on your computer for one client or one area of your business.
- Put your real files in it: contracts, past proposals, client data, call transcripts.
- Create a project in Claude or ChatGPT and point it at that folder.
- Work inside that project instead of starting loose chats.
Now AI can read all of your files, and what it learns gets saved for every conversation after that one. All four of the use cases below depend on having a project set up first.
A plain chat
No memory. It fills up, it closes, and you start from nothing. You explain your business again every time.
A project on a folder
It reads your real files and saves what it learns. Every answer after that starts with your business already in context.
The second thing to set up is your connectors. In Claude or ChatGPT you go to customize and connect the tools your work already lives in. For sales that means your email, your calendar, your drive and your CRM, and there are connectors for HubSpot, Zoho and Monday already sitting in the directory. I keep Zoom connected as well, which means I can pull a transcript from a client call straight into an analysis without going to find the file. There are more than 2,000 connectors now, and you need about four of them.
Use Case 1: How to Find Your Ideal Client With AI
Knowing precisely who you're selling to roughly doubles your rate of closing. Most of us have a rough idea of our ideal client and a prospect list that doesn't quite match it.
So don't start by asking AI to find you leads. It has no idea what a good client means in your business until you show it. Start with the ones you already closed. Who are your three best clients, why did you close them and not someone else, and what was it about the deal that resonated?
If you have call transcripts, feed them in. And if you don't have best clients yet, that's fine. You can let the AI interview you about the client you want and research that category itself.
"Here are my three best clients: [names, what you sold them]. Ask me 10 clarifying questions, one at a time. Then write my ideal client profile and save it to memory for all future lists, prospecting, presentations and communications."
The part that matters most in there is save it to memory for all future lists. That's the difference between a profile you use once and a profile that shapes every list you pull from now on.
"Find 20 businesses in my area that match this profile. For each one: the owner's name, website, and one thing happening there right now. Put it in an Excel file and prioritize the prospects according to their best fit to my ICP."
Two details in that prompt are doing the work.
The ranking. When you have a month end target, you can start with the prospects who best match your profile instead of working down the list alphabetically.
The one current fact for each prospect. A recent hire, a new location, an award they just won. That detail is what makes your email read like you actually looked at their business, and it is the difference between a message someone answers and one they delete.
A precise client profile makes every step after it easier. Better prospects, better emails, better replies.
Ask AI where its information came from. If you don't teach it what a good source looks like, it will go looking anywhere and hand you a confident list built on very little. So ask for sources on every research task, every time.
Use Case 2: How to Never Miss a Follow-Up
78% of buyers go with whoever responds first (Lead Connect). When someone needs a service now, speed and consistency win the deal more often than a better pitch does.
And up to 70% of a salesperson's time goes to admin instead of selling (Salesforce, State of Sales). Those two numbers sitting next to each other are the problem. Speed is what wins the deal, and admin is what fills your day.
You're not behind on follow-ups because you're disorganized. You're behind because nobody has time to check, so what you want is something that checks for you.
Here is how to set up the daily briefing:
- Connect your email and your calendar in the connectors menu.
- Go to scheduled tasks and create a new task.
- Paste in the briefing prompt below and set it to run every weekday at 9am.
- Choose to skip approvals, so it sends without waiting on you.
Mine runs every weekday at 9am. Because your email and calendar both live in the cloud, the task runs whether your laptop is open or not, and that is the part most people assume isn't possible.
"Every morning at 9am, review my calendar for the day and prep me for each meeting with prospect research by sending me an email in my inbox. Then review my sent emails from the last 7 days. Which clients or prospects have I not spoken to in 3 days? For each one, remind me what we last discussed and draft a follow up response."
What arrives in my inbox is short. Who I'm meeting, whether I'm already prepped, which threads are still live, and who's gone quiet on me. It tells me a prospect hasn't replied in eight days and that there's a half written follow-up already sitting in my drafts, and I didn't write any of it.
Use Case 3: Your Data Is Worth Nothing Without Insights
When I was a sales analyst for AB InBev, the biggest brewer in the world, I spent hours and hours every day building out files to deliver insights. I wish I had had Claude back then, because the same insights can now be delivered in about 15 minutes.
Using AI to analyse your own sales data is my favourite of the four use cases, and almost nobody does it. Most small businesses are sitting on months of data that nobody has really looked at, because looking at it properly used to take a week you didn't have.
A B2B owner uploaded six months of cold outreach, about 5,000 emails, into Claude and asked one question. What do the people who said yes have in common? It found three patterns he'd missed after months of staring at the same data, and his reply rate went from 2.8% to 5.9% in three weeks.
"Here is my outreach history: who replied, who didn't, who bought. What do the people who said yes have in common that the others don't?"
"Turn what you found into a set of rules for how I should write my next 20 emails."
Ask AI for rules rather than templates. Rules stay yours, and templates start to sound like AI wrote them, which your reader can usually tell.
If I had to choose between AI as an analyst and AI as my email writer, I'd take the analyst every single time. It's much better at telling you which of your emails already worked than at writing the next one. It does write emails too, but you're the one who knows your audience and the intent behind the message, and that part it can't do for you.
A living dashboard. You point it at a folder, ask AI to build the view, and publish it as an artifact. You get a link you can send to a partner or a client, and it updates as you add data. One I built from a Google Analytics export and a campaign report took about 15 minutes.
Use Case 4: How to Write Proposals and RFPs in Minutes
I was talking to an artist recently who told me she doesn't have time to fill out all the RFPs she gets, and she knows that if she doesn't do an amazing job on them she won't win them. So she sends fewer of them. Her calendar was making that decision for her.
With AI she now fills out 70 to 80% of each one and then makes it her own. She sends far more than she used to, and she still puts the finishing touch on every single one.
The competitor research on its own used to take hours. The client, the project, the industry, and who else is bidding. That's about 15 minutes now.
"Research [business name] in [city]. Find one concrete thing happening there right now that connects to what I sell. Research industry and competitors as well, position my solution in relation to the current challenges. Show me your sources."
"With everything you know about this lead, the company, the proposal terms, priorities and my proposal template, create a winning business case and RFP."
The best thing you can do here is teach AI from the proposals you already won:
- Give AI your winning proposals and the ones nobody responded to.
- Tell it which was which.
- Ask AI what the difference was between them.
- Ask it to apply those winning criteria to your next proposal.
Now AI is learning your actual winning criteria instead of general proposal advice.
That's what training AI looks like. You teach it what good means in your business, and it gets better over time from its own mistakes. You can point the same habit at contracts, so if someone sends you one to review, read it yourself and then ask AI to review it as well.
What This Actually Gives You Back
When I say five to ten hours a week, I'm not making that number up. It's what owners who are actively using this and building memory tell me:
- Customer emails with AI: "Saved me around 5 hours a week"
- Email templates: "What used to take 20 minutes takes 2", on 10 to 15 emails a week
- Sales follow-ups: from 15 minutes per email to filling in blanks, "5 to 6 hours a week back in my pocket"
- 3,000 rows of messy CRM data standardized "in an afternoon"
- 20 competitor pricing pages analyzed and summarized, "would've taken me hours"
- Research synthesis: "what used to take 2 hours took 20 minutes"
None of that came from a better prompt.
The Actual Difference
The small business owners who are getting hours back aren't the ones with clever prompts. They're the ones who stopped chatting and started delegating, one small task at a time.
So pick the small task, the repetitive one where you already know you're losing the most time, and let it run for a week before you add anything else.
AI is here to augment you rather than replace you. You're still the one with the judgement and the read on your own market, and only you are the genius in your business. But if it takes the boring half off your plate, you get those hours back for the work that brings in revenue and clients.
Get all four prompts in the Small Biz AI Hub
Every prompt from this article is inside the community, along with live weekly AI office hours where we work through your real business questions together. Real people, real support, and discussions that are always ongoing. Free to join for a limited time.
Would you rather have it done for you?
Book a free AI diagnostic. In 30 minutes we find where AI actually pays off in your business, and you leave knowing which of these four to set up first.