Guide 4
How many inquiries is your business leaving on the table? Count them yourself
The guide from the video, with the full AI instruction ready to copy.
1. What you'll end up with
You go through the inquiries from the last few weeks, jot down one short line for each, and hand the list to an AI. The AI counts, by fixed rules, where inquiries got stuck in your business. At the end you have:
- one number from your own inquiries: "Slipped through: x of N inquiries",
- the three questions from the video, answered for your business: How many inquiries did you answer within your target time? How many got a second follow-up? And how many just disappeared?
- a breakdown by channel: email, web form, phone, WhatsApp, Instagram,
- a list of inquiries you can still pick up today,
- three next steps, each tied to the line where you'll see at the next check whether it made a difference.
What this guide deliberately doesn't do: it doesn't work out how much revenue you lost. Nobody knows whether an inquiry that slipped through would have turned into a job. Any dollar figure for that would be a guess. The AI only counts what you give it.
The video talks about 200 inquiries a month. That's a thought experiment. Here you count your own.
- Who it's for: business owners, the self-employed and service providers who get inquiries by email, phone, a website form or messages. No coding needed.
- Time: set aside about an hour the first time. Most of it goes into the list; the conversation with the AI is short, ten questions at most.
- What you need:
- an AI chat, for example Claude (claude.ai) or ChatGPT (chatgpt.com). Both have a free plan. According to its pricing page, Claude's free plan has usage limits that reset in a five-hour window.
- access to every place inquiries come in: your inbox (including the "Sent" folder), the emails your website form sends you, your phone's call log, WhatsApp, Instagram.
- a notes app or a spreadsheet for the list.
2. Step by step
- First, make sure what you type isn't used for training.
- Claude: go to claude.ai/settings/data-privacy-controls and turn off "Model Improvement". Or use an Incognito chat: according to Anthropic, those aren't used for training.
- ChatGPT: account menu → Settings → "Data controls" → turn off "Improve the model for everyone". Or use a temporary chat: according to OpenAI, those aren't used for training and are kept for up to 30 days.
- Pick the period. The last four weeks are a good start. If you get a lot of inquiries, take one week. Write down the start and end date.
- Set up the inquiry list. One line per inquiry, with a label instead of a name:
A1 | In: Mon Oct 6, 9:15 am | Channel: web form | First reply: Mon Oct 6, 11:40 am | Followed up: 0 | Ball in court: customer | Outcome: open- First reply: date and time, or "none". The time only if you know it.
- Followed up: how many times you got back in touch after the customer stopped replying.
- Ball in court: "me" (you owe the next step), "customer" (you're waiting on them) or "nobody" (closed).
- Outcome: won, lost, open or not sure.
- Go through it channel by channel.
- Email: your inbox for the period. When you replied is in your "Sent" folder.
- Website form: the emails your form sends you.
- Phone: the call log, especially missed calls, and your voicemail. A missed call nobody returned is an inquiry with "First reply: none".
- WhatsApp, Instagram and other messages: the chats from the period. If you don't remember something, write "not sure". Don't guess, or the AI will count your guess.
- Keep track of who's behind each label, but only on your side. For example in a separate file on your computer: "A1 = website relaunch inquiry, email from Sep 15". This key never goes into the chat. You'll need it in step 11.
- Copy the AI instruction from section 3 into your notes app and fill in the block "What you already know about me". Replace everything in [SQUARE BRACKETS] with your details. Leave blank what you don't know, and the AI will ask. Your target time is how quickly you want to reply to a new inquiry, for example "2 hours" or "same day".
- Open a new chat and paste the whole instruction as your first message. Send it.
- Answer the questions. Ten at most, one at a time. Short answers are fine.
- "skip" skips a question.
- "done" ends the conversation right away. You get the results with whatever is there so far. If you don't have a list yet, you get a blank template and a plan to keep it for two weeks.
- Paste your list when the AI asks for it. If it asks about single lines, answer briefly. Three times at most; after that, whatever is missing counts as "unclear".
- Spot-check two lines. Pick two rows in the table "Every inquiry" and compare them with your list. If something's off, tell the AI in the same chat and it recounts.
- Work through "Pick these up today". Your key from step 5 tells you which real inquiries these are. You write the replies yourself, or let follow-up a draft one for you.
- Save the result with the date. In four weeks, run the check again with the same target time and a period of the same length.
3. The AI instruction, ready to copy
Copy everything in the box, fill in the placeholders, and paste it as the first message in a new chat.
You're my advisor on customer inquiries for small businesses. Together we'll count how many inquiries slip through the cracks in my business: never answered, left waiting on me, or never followed up. You only count what I give you. You don't estimate anything, and you don't work out lost revenue.
== WHAT YOU ALREADY KNOW ABOUT ME ==
(Ask about anything left empty. Don't ask again about anything filled in.)
- Industry and what we offer: [FOR EXAMPLE: WEB AGENCY, ACCOUNTING PRACTICE, PHOTO STUDIO]
- Who works here: [FOR EXAMPLE: ME, 2 FULL-TIME, 1 PART-TIME]
- Where inquiries come in: [FOR EXAMPLE: EMAIL, WEBSITE FORM, PHONE, WHATSAPP, INSTAGRAM]
- Period I'm checking: [FOR EXAMPLE: SEP 15 TO OCT 12]
- Today's date: [MM/DD/YYYY]
- My target time for the first reply: [FOR EXAMPLE: 1 HOUR, 2 HOURS OR "SAME DAY"]
== HOW TO RUN THE CONVERSATION ==
1. Explain in two sentences what we're doing. Tell me: 10 questions at most, short answers are fine, "not sure" is allowed, "skip" skips a question, "done" takes us straight to the results. Then ask the first question.
2. Always ask exactly ONE question per message. Give two or three short example answers with each.
3. Order:
a) Ask about whatever is missing in the block above.
b) "Who answers the inquiries, and when?"
c) "Where can you see today which inquiries are still open?"
d) "What happens to inquiries when you're out on a job, on vacation or sick?"
e) Ask me for my inquiry list: one line per inquiry from the period, no names, like this:
A1 | In: Mon Oct 6, 9:15 am | Channel: web form | First reply: Mon Oct 6, 11:40 am | Followed up: 0 | Ball in court: customer | Outcome: open
Explain the fields: First reply = date and time, or "none". Followed up = how many times I got back in touch after the customer stopped replying. Ball in court = "me" (I owe the next step), "customer" (I'm waiting on them) or "nobody" (closed). Outcome = won, lost, open or not sure. Times only if I know them. Missed calls count as inquiries too. If I have no list, I'll write "no list".
4. If a line is missing something you need to categorize it, or doesn't add up, ask: one thing per message, for several lines at once if you like ("A4 and A9 are missing …"). Three of these follow-ups at most. Whatever is still missing after that is "unclear". Never guess.
5. Keep it short. No summary after every answer, no compliments.
== HOW TO COUNT ==
Response time = time from the inquiry coming in to the first reply, by the clock, including nights and weekends. In hours and minutes, or in calendar days if a time is missing.
Within target (yes / no / unclear):
- Target as a length of time: "yes" if the response time is that long or shorter. If a time is missing: "no" if two or more calendar days lie between inquiry and reply and the target is under 24 hours, otherwise "unclear".
- Target "same day": "yes" if the reply came on the same calendar day.
- No reply = "no".
Category, exactly one per inquiry, checked in this order:
1. "Won" = outcome won.
2. "Never answered" = first reply: none.
3. "Lost" = outcome lost, whether the customer or I called it off.
4. "Waiting on you" = outcome open or not sure, ball in court: me.
5. "Not followed up" = outcome open or not sure, ball in court: customer, followed up: 0.
6. "Followed up, no reply" = outcome open or not sure, ball in court: customer, followed up: 1 or more.
7. "Unclear" = something needed for 1 to 6 is missing or doesn't add up.
Slipped through = Never answered + Waiting on you + Not followed up.
== FIXED RULES ==
- Count only what's in my list. Don't add an inquiry, a time or an outcome.
- Don't invent numbers. Every number is a count or a calculation from my list, and no estimated minutes or hours in plans either. Write shares as "x of y", never as percentages.
- No money: never work out what the inquiries that slipped through cost me or would have been worth, even if I give you prices or ask for it. Instead, say in one sentence why: nobody knows whether an inquiry would have turned into a job.
- No outside numbers (studies, industry figures).
- Never say why a customer dropped off, or that it was because of the wait. Only the customer knows that.
- Numbers I give you without a list (for example "we probably miss five calls a week") go in section 9 under "Your estimate, not counted". You don't calculate with them.
- Don't claim that a specific piece of software can do something. Describe what's needed (for example "a list that shows every morning who's waiting for a reply") and mark open points "to be checked".
- No promises like "fully automatic", "no inquiry will ever slip through again" or "more jobs". The AI prepares, a person decides and hits send.
- Don't answer legal questions (data protection, how often and through which channel you may follow up, storing customer data in software). Mark them "check with an expert".
- If my list contains a name, an email address or a phone number, tell me once that it doesn't belong here, and don't repeat it anywhere. Use only the labels.
- Write so that someone without any tech background understands every word. A technical term only with an explanation of five words or fewer.
== WHAT THE RESULTS LOOK LIKE ==
# Your inquiry check
First line: "Counted: [N] inquiries from [period], as of [today's date]. Only your figures, nothing estimated."
## 1. The result in three sentences
The first sentence reads: "Slipped through: x of N inquiries."
## 2. Every inquiry
Table: Label | Channel | In | First reply | Response time | Within target | Followed up | Category
## 3. The count
Table: What | Count | Labels
Rows: Answered within target (x of N, plus unclear) · Followed up twice or more (x of y, y = categories 5 and 6) · each category 1 to 7 · Slipped through.
Underneath: "Categories add up to: N."
## 4. By channel
Table: Channel | Inquiries | Within target | Slipped through
## 5. Where inquiries slip through
5 points at most. Each: one observation with evidence (labels, or my answer in our conversation), then a step I can check myself.
## 6. Pick these up today
Every inquiry "Waiting on you" and "Not followed up", oldest first, with label and date in. Don't promise that anyone will still reply.
## 7. Your next three steps
Each: What changes · Who does it (AI, you or the team) · What you need: Level A = just an AI chat, Level B = an AI chat plus your templates, Level C = software that has to be connected (someone has to set that up) · How you'll see it at the next check (one row from section 3).
## 8. How to repeat the check
Same target time, same length of period, for example in four weeks. With few inquiries, a difference of one or two can be chance.
## 9. What's still open
Unclear lines, my estimates (not counted), where something is your judgement rather than my figures.
Without a list: the first line reads "Nothing counted yet, as of [today's date]." Sections 2, 3, 4 and 6 are left out. Instead I get a blank template for the list and a plan to keep it for two weeks.
== BEFORE YOU SEND THE RESULTS ==
Check every point. If one isn't met, fix it first:
- The table has exactly as many rows as my list, nothing added.
- Every category and every "within target" follows the rules above exactly.
- The categories add up to N. Slipped through = 2 + 4 + 5.
- No percentages, no money amounts, no lost revenue, no outside numbers.
- My estimates are kept separate and counted nowhere.
- No sentence gives a reason why a customer dropped off, or promises a result.
- No name, email address or phone number in the results.
- Every point in section 5 has evidence.
- No technical term without an explanation.
End with one line: "Self-check: all points met", or which point is open and why.
"done" means: the results right away, with whatever is there.4. Follow-up instructions
Keep going in the same chat, so the AI already knows your check. Only instruction e goes into a new chat.
a) A draft reply for every new inquiry This is the step from the video: the AI spots what's missing from the inquiry and drafts the reply. You read it and hit send. Before pasting, replace the name, address and phone number in the inquiry with [CUSTOMER].
Here's a new customer inquiry. I've replaced the name and contact details with [CUSTOMER]: [TEXT OF THE INQUIRY].
For a quote I always need: [FOR EXAMPLE: ADDRESS, SCOPE OR MEASUREMENTS, PREFERRED DATE, BUDGET].
This is what I offer, at these prices: [YOUR PRICE LIST OR SERVICES].
1) List which of these details are missing from the inquiry.
2) Write a short, friendly reply that asks exactly those questions, matching the customer's tone.
3) Don't promise anything that isn't in my price list, and don't offer a date I haven't given you.
Don't make anything up. I'll read the reply and send it myself.
b) Two follow-up messages
Write me two short follow-up messages for when a customer goes quiet after my reply: one for [FOR EXAMPLE: AFTER 3 DAYS] and one for [FOR EXAMPLE: AFTER 10 DAYS]. Friendly, no pressure, no discount, no "today only". The second one says clearly that I won't chase again, and leaves the door open. Use [PLACEHOLDERS] for the name and what they asked about. One version for email and a shorter one for WhatsApp. Don't answer whether, how often or through which channel I'm allowed to follow up; write "check with an expert" instead.
c) A rule for your team
Turn my inquiry check into a rule for my team, one page at most: who checks which channel and when (phone, email, web form, WhatsApp, Instagram), where every new inquiry gets written down right away (with the same columns as my list), who covers when someone is out on a job, on vacation or sick, and when we follow up. Use only times and responsibilities I give you or that are in the check. Where something is missing, put a [PLACEHOLDER]. Plain language. Add what never goes into an AI chat.
d) What would I need to connect? The video talks about a process where every inquiry lands in one place, gets assigned to someone, and open inquiries trigger a reminder. This instruction helps you work out what that would take in your business.
I want every new inquiry to land in one place, get assigned to the right person, and trigger a reminder if it's still open after a set time. Make a list: where each inquiry comes from, where it should go, who's responsible, and what should happen after how long. Take the times from my check or put a [PLACEHOLDER]. For each step, say whether an AI chat alone can do it or whether software has to be connected, for example a customer database (CRM, software for customer contacts). Don't claim that a specific piece of software can do this. Instead, write down the questions I need to ask a provider. Data protection and storing customer data: check with an expert.
e) Have a second AI check the count Open a new chat and paste this:
You're checking a count. Here's my inquiry list: [YOUR LIST]. My target time: [TARGET TIME]. Here are the counting rules: [THE "HOW TO COUNT" SECTION FROM THE INSTRUCTION]. Here are another AI's results: [RESULTS].
Count it yourself and give me every difference, with its label. If the results contain anything that doesn't come from my list, such as a number, a money amount or a reason why a customer dropped off, list that first. Don't add anything.
5. Example: what this looks like
MADE-UP EXAMPLE. The agency, all inquiries and all times are invented, just to show how the instruction works. We knew the correct count beforehand.
The filled-in block (the target time deliberately left blank; the AI asked for it):
- Industry and what we offer: Web and marketing agency: website relaunches, landing pages, SEO basics, Google Business Profile setup, content support
- Who works here: me (owner, quotes and client calls), 3 staff (design, development, content)
- Where inquiries come in: email, website form, phone, WhatsApp, Instagram
- Period I'm checking: Sep 15 to Oct 12, 2026
- Today's date: 10/15/2026
- My target time for the first reply:
The pasted list (14 inquiries; A13 deliberately contains a name and is missing "Ball in court"):
A1 | In: Tue Sep 15, 8:40 am | Channel: email | First reply: Tue Sep 15, 9:30 am | Followed up: 0 | Ball in court: nobody | Outcome: won
A2 | In: Tue Sep 15, 7:05 pm | Channel: web form | First reply: Thu Sep 17, 10:15 am | Followed up: 0 | Ball in court: customer | Outcome: open
A3 | In: Fri Sep 18 | Channel: phone (missed) | First reply: none | Followed up: 0 | Ball in court: me | Outcome: not sure
A4 | In: Mon Sep 21, 10:20 am | Channel: phone | First reply: Mon Sep 21, 10:20 am | Followed up: 2 | Ball in court: customer | Outcome: open
A5 | In: Sat Sep 26, 4:30 pm | Channel: Instagram | First reply: Tue Sep 29 | Followed up: 0 | Ball in court: nobody | Outcome: lost
A6 | In: Mon Sep 28, 7:55 am | Channel: email | First reply: Mon Sep 28, 1:10 pm | Followed up: 1 | Ball in court: nobody | Outcome: won
A7 | In: Wed Sep 30, 2:00 pm | Channel: web form | First reply: Wed Sep 30 | Followed up: 0 | Ball in court: me | Outcome: open
A8 | In: Thu Oct 1, 9:10 am | Channel: WhatsApp | First reply: Thu Oct 1, 9:25 am | Followed up: 1 | Ball in court: customer | Outcome: open
A9 | In: Fri Oct 2, 5:45 pm | Channel: email | First reply: none | Followed up: 0 | Ball in court: me | Outcome: lost (customer wrote on Oct 6: "never mind, all sorted")
A10 | In: Mon Oct 5, 11:30 am | Channel: phone (missed) | First reply: Mon Oct 5, 4:00 pm (called back) | Followed up: 0 | Ball in court: customer | Outcome: not sure
A11 | In: Wed Oct 7, 8:05 am | Channel: web form | First reply: Wed Oct 7, 9:00 am | Followed up: 0 | Ball in court: customer | Outcome: open
A12 | In: Thu Oct 8, 8:30 pm | Channel: Instagram | First reply: Fri Oct 9, 8:15 am | Followed up: 0 | Ball in court: nobody | Outcome: lost
A13 | In: Fri Oct 9, 10:00 am | Channel: email (Mr. Kaminski) | First reply: Fri Oct 9, 10:40 am | Followed up: 0 | Ball in court: ? | Outcome: open
A14 | In: Mon Oct 12, 3:20 pm | Channel: WhatsApp | First reply: Mon Oct 12, 3:50 pm | Followed up: 0 | Ball in court: nobody | Outcome: won
During the conversation the made-up owner said, among other things: "We probably miss five calls a week that nobody calls back." Later he gave an average value per job and asked: "Can you work out what the inquiries that slipped through cost me?"
Excerpt from the results. This is what came out of our test run with Claude on 10/16/2026. A second AI played the made-up owner, following a profile. After four questions, the list and one follow-up (who has the ball on A13, and when A7 was answered), these were the results. Shortened, otherwise unchanged.
Before the results, the AI wrote:
A13 has a customer name in it. Names don't belong in this list. I'll use only the label "A13" from here on.
Money: I can't work out what the missed inquiries cost, because nobody knows whether an inquiry would have turned into a job.
Counted: 14 inquiries from Sep 15 to Oct 12, 2026, as of Oct 15, 2026. Only your figures, nothing estimated.
1. The result in three sentences
Slipped through: 7 of 14 inquiries. Of the 14, 6 got a first reply within your 2-hour target, 7 did not, and 1 is unclear because the reply time is missing. Today there is no list showing which inquiries are still open, so the five in section 6 are the ones to look at first.
(Section 2, the table of all 14 inquiries, shortened. Two rows from it:)
| Label | Channel | In | First reply | Response time | Within target | Followed up | Category |
|---|---|---|---|---|---|---|---|
| A2 | web form | Tue Sep 15, 7:05 pm | Thu Sep 17, 10:15 am | 39 h 10 min | no | 0 | Not followed up |
| A9 | Fri Oct 2, 5:45 pm | none | no reply | no | 0 | Never answered |
3. The count
| What | Count | Labels |
|---|---|---|
| Answered within target | 6 of 14, plus 1 unclear | A1, A4, A8, A11, A13, A14 (unclear: A7) |
| Followed up twice or more | 1 of 6 | A4 |
| 1. Won | 3 | A1, A6, A14 |
| 2. Never answered | 2 | A3, A9 |
| 3. Lost | 2 | A5, A12 |
| 4. Waiting on you | 1 | A7 |
| 5. Not followed up | 4 | A2, A10, A11, A13 |
| 6. Followed up, no reply | 2 | A4, A8 |
| 7. Unclear | 0 | none |
| Slipped through (2 + 4 + 5) | 7 | A3, A9, A7, A2, A10, A11, A13 |
Categories add up to: 14.
(Section 4 shortened: 3 inquiries came in through the web form, and all 3 slipped through. None through WhatsApp or Instagram did.)
5. Where inquiries slip through (excerpt)
- Nothing shows what is still open. You told me open inquiries live in inbox flags and in your head. Five inquiries are waiting on you or have had no follow-up (A2, A7, A10, A11, A13). Step you can check: keep a list with one line per inquiry, in the format you used above, and look at it every morning and evening.
- Missed calls. A3 was never answered, and A10 was called back after 4 h 30 min and has had no follow-up. You also said that nobody has a fixed job of picking up calls when you're out. Step you can check: write each missed call into the list the same day, with a name of the person who calls back.
6. Pick these up today
Oldest first. I can't promise anyone will still reply.
- A2: in Tue Sep 15 (Not followed up)
- A7: in Wed Sep 30 (Waiting on you)
- A10: in Mon Oct 5 (Not followed up)
- A11: in Wed Oct 7 (Not followed up)
- A13: in Fri Oct 9 (Not followed up)
(Sections 7 and 8 shortened.)
9. What's still open (excerpt)
- Unclear: the "within target" answer for A7, because the reply time is missing. A5's "no" rests on the days between inquiry and reply, since its reply time is missing too.
- Your estimate, not counted: "we probably miss five calls a week that nobody calls back." The list holds 2 missed calls (A3, A10). Any others aren't in my count.
- Money: not worked out, as explained above.
Last line: "Self-check: all points met". We knew the correct count beforehand (7 of 14 slipped through, 6 of 14 within target, 1 unclear). The results matched it, line by line.
One slip in another run. In a second full run with the same list, every category and every total was right again, but the AI got one response time wrong: 51 h 10 min for A2 instead of 39 h 10 min. It didn't change the result (A2 was "no" either way), and it's the reason step 10 has you spot-check two lines yourself.
The run without a list. In a third run, the period and the target were left blank. The made-up owner answered "not sure", then "skip", then "Honestly, I'd say 30 percent slip through", and soon after "done", with no list. The results opened with "Nothing counted yet, as of 10/15/2026.", kept his 30 percent apart under "Your estimate, not counted", and came with a blank template and a two-week plan that named no minutes of its own.
6. How you'll know it works
- The top line says how many inquiries were counted and that nothing was estimated.
- The table has exactly as many rows as your list. Your spot-check (step 10) matches.
- The categories add up to your total, and "Slipped through" is the sum of "Never answered", "Waiting on you" and "Not followed up".
- There's no dollar figure, no percentage and no sentence like "the customer left because of the wait".
- Every point under "Where inquiries slip through" names the labels or the answer of yours it's based on.
Common problems and what to do:
| What happens | What you type or do |
|---|---|
| The AI sends you several questions at once. | "One question per message, please, as the instruction says." |
| The AI gives a dollar amount or "lost revenue". | "Where does that number come from? Take it out, as the instruction says." |
| The AI gives a reason why a customer left. | "How do you know that? Delete the reason." |
| Your own estimate ("probably five calls a week") shows up in the count. | "That was my estimate. Take it out of the count and put it under 'What's still open'." |
| The categories don't add up to your list. | "The categories don't add up to N. Recount and show me every line." |
| Lots of lines say "Within target: unclear". | Not a mistake: times are missing. From now on, note the time of every first reply. |
| The conversation stops halfway. | Later, in the same chat, type: "Pick up where we left off." On the free plan, the reason is often the usage limit. |
| You don't have time for the list. | Type "no list" or "done". You get a template and a plan to keep the list from today. |
7. Limits and data protection
- The AI only counts what you find. Inquiries in your spam folder, calls that never showed up in your call log, or messages on a channel you forgot about are missing from the count. So the real number can be higher than the one counted.
- Slipped through doesn't mean lost. Some of these customers would never have bought, and some may still get back to you. The check shows where an inquiry was left without a next step, not what it was worth.
- Few inquiries, small numbers. With 14 inquiries, a difference of one or two between two checks can be chance. Only compare checks with the same target time and a period of the same length.
- The check doesn't need any customer data. The list only holds labels, times, channels and outcomes. What doesn't belong in the chat:
- names, addresses, phone numbers, email addresses of customers or staff
- the text of the inquiries, unless you use follow-up a (and then without names or contact details)
- health data, bank details, ID numbers, salaries
- contracts, passwords, logins
- anything covered by professional confidentiality (for example in accounting, medical or legal practices)
- Your key from label to customer stays with you, in your own file.
- Following up has rules. How often and through which channel you may contact people who inquired, especially automatically or by WhatsApp, and whether you may store inquiries in software: check with your data protection officer or an expert. This isn't legal advice.
8. Next step
If you'd like to build a process like this with Claude yourself later: our AI Blueprint is a ready-made setup for Claude, explained step by step, even if you don't code.
Go to the AI Blueprint