Quota Whisperer

Reads

call transcript, the email thread

Runs on

an assistant, when I share the transcript

Ask first
- Where do call transcripts come from today? (default: paste the transcript or your notes into the chat after the call)
- How soon after the call? (default: within an hour, when you ask)
- Reply on the existing thread or start a new one? (default: reply on the thread the invite came from)
- Sign-off and tone? (default: first name, plain, no exclamation marks)

Needs
- A transcript or your own notes from the call, in any form. If you have no recorder, two minutes of dictated notes straight after the call are enough.
- Your email, with permission to create drafts. It creates drafts only; it never sends.
- Chat-only version: paste the notes, get the email text back, paste it into your mail yourself.

Build
1. Read: when the person shares a transcript or notes, read them and the last 5 emails on the thread.
2. Decide: pull out three lists: what was agreed, what you owe (with a date if one was said), what they owe. Quote commitments in the words used. If a date was not said, write "by [date?]" rather than invent one. Keep "send the security questionnaire by Friday"; drop "sounds good, talk soon".
3. Deliver: save a draft reply on the thread: a two-sentence thanks, the three lists as short bullets, one proposed next step with a time. The person reads, edits and sends. The routine never sends.
4. Memory: none needed; each call is its own draft.
5. Cost: one model call per call, 30 to 50 a month.

Check
- Paste an invented transcript with two clear commitments and one vague one; the draft must list the two in their wording and mark the vague one with a question.
- Confirm the draft sits in Drafts and nothing was sent.

In my version
- Transcripts come from the call recorder my employer uses; I share the link into the chat after the call.
- Drafts land in Gmail on the original invite thread.

Reads

CRM contacts, public profiles, email reply dates

Runs on

an automation tool that runs an assistant

Ask first
- Who counts as a champion? (default: the main contact on every open deal and every account renewing this year)
- What counts as going quiet? (default: no reply to your last two emails over 10 working days)
- Where should the list land? (default: a morning message to yourself in your team chat tool, otherwise an email)
- Do you have a list of these people today? If not, start with the 20 contacts you emailed most last quarter.

Needs
- A list of people, each with employer and email. A CRM export or a plain list works.
- Their public profiles, read by web search once a day. It reads only public pages; it does not log in as you.
- Your sent mail and replies, read only, for the people on the list only. It does not read other mail.
- Chat-only version: once a week, paste the list and ask "anything changed with these people?"; the assistant searches and reports.

Build
1. Read: each morning, for each person on the list, search their name and employer for a new title, a public post or a news mention in the last 7 days, and check the date of their last reply to you.
2. Decide: flag three things only. Moved: employer or title changed since the last run. Posted: a public post about their work in the last 7 days (keep "we are rolling out a new onboarding flow"; drop a reshare of someone else). Quiet: your last two emails to them unanswered for 10 working days. Anything else is dropped.
3. Deliver: one message, one line per flag, with the evidence in brackets. No message when nothing is flagged. Keep a note of what was flagged and when, so the same post or move is not repeated.
4. Cost: one model call per person per day; 30 people is about 900 calls a month, so cap the list at 40 people.

Check
- Put an invented person on the list whose last reply you set to 15 working days ago; the run must flag them as quiet and nothing else.
- A normal day with nothing new sends nothing; the first real run shows the message at your chosen time.

In my version
- The list comes from the CRM; an automation tool runs the assistant and posts to [Company]'s chat tool.
- About 30 people on the list.

Reads

CRM pipeline, calendar, last week's email

Runs on

an assistant on a schedule

Ask first
- When is your pipeline meeting? (default: Monday 9am; the review arrives two hours before)
- Which deals count? (default: every open deal with a close date in the next 90 days)
- How many "touch today" picks? (default: three)
- Do you have a pipeline list with stage, amount, close date and last activity? If not, keep one in a spreadsheet; the review reads that.

Needs
- A pipeline list with stage, close date and last activity date per deal. A CRM or a spreadsheet.
- Your calendar for the coming week, read only, to see which accounts you already meet.
- Last week's sent and received mail, read only, for the accounts on the list.
- Chat-only version: paste the pipeline list on Monday and ask for the review.

Build
1. Read: Monday early, read every open deal, last week's email per account and this week's calendar.
2. Decide: for each deal write one line: what changed last week (a reply, a meeting, a stage move, or "nothing"). Then pick the three to touch today in this order: close date within 30 days with no activity in 7 days first; then any deal whose last email was from you with no reply; then the largest remaining. Keep "close date 20 Oct, last activity 12 days ago"; drop a deal you meet on Tuesday anyway.
3. Deliver: one message to yourself: the three picks at the top, then the full list grouped by stage. Nothing is sent to anyone else.
4. Memory: store last week's snapshot so "what changed" compares against it.
5. Cost: one model call a week.

Check
- Run it on a Wednesday with an invented pipeline of five deals, one with a close date in 20 days and no activity; it must be pick one.
- Confirm the first real Monday message arrives at the set time.

In my version
- Pipeline from the CRM; lands in [Company]'s chat tool as a direct message to myself.
- About 25 open deals.

Reads

CRM activity dates, email threads

Done when

Closes when the account replies or the deal is marked closed-lost

Runs on

an assistant on a schedule

Ask first
- How long is quiet? (default: 10 working days with no email, meeting or reply)
- Which deals are exempt? (default: close date more than 120 days out, and anything marked on hold)
- Draft nudges, or just the list? (default: both; drafts are never sent)
- Cap per week? (default: 8 nudges)

Needs
- The deal list with last activity date and the main contact's email.
- Your email, with permission to create drafts only. It creates drafts on existing threads; it never sends.
- Chat-only version: paste the list, ask for the quiet ones and a nudge for each, paste the nudges into your mail yourself.

Build
1. Read: midweek, read the deal list and, for each deal, the last email on its thread.
2. Decide: a deal is quiet when the last activity of any kind is 10 or more working days ago and it is not exempt. Rank by close date, soonest first, and stop at the cap. The nudge is two sentences: one that refers to the last thing they said (quoted), one that asks a single question or offers a single next step. No "just checking in".
3. Deliver: save each nudge as a draft reply on the deal's thread, and post the list to yourself with the deal, the days quiet and a one-line reason. Record which deals were nudged and when; do not draft the same deal again within 15 working days.
4. Cost: up to 8 model calls a week.

Check
- Invent three deals: 12 days quiet, 4 days quiet, and 12 days quiet but on hold; only the first gets a draft.
- A week where nothing is quiet produces no drafts and a one-line "nothing quiet" message.

In my version
- Reads the CRM; drafts land in Gmail.
- Cap is 8 and the quiet rule is 10 working days.

Reads

CRM renewal dates, usage summary, email

Runs on

not built yet

Ask first
- How far ahead? (default: renewals in the next 90 days)
- What signals do you have? (default: renewal date and your email thread; usage only if your product reports it)
- Where should it land? (default: a Friday message to yourself)

Needs
- A list of customers with renewal date and main contact.
- Any usage signal you can get as a number per account (logins per week, seats used). Fallback: skip usage and rely on the other two signals.
- Your email with those contacts, read only.
- Chat-only version: paste the renewal list each Friday and ask which look at risk and why.

Build
1. Read: Friday, read every account renewing in 90 days, its usage number for the last 4 weeks, and the email thread.
2. Decide: flag an account when any of these hold: usage down by a third or more against the previous 4 weeks; the main contact left or has not replied in 15 working days; the thread mentions budget, cancel or another vendor. Keep "logins 40 to 22"; drop a holiday week.
3. Deliver: one message listing each flagged account, the signal and the date, with one suggested action per account. Never contact the customer.
4. Memory: keep the previous 4-week usage number per account.

Check
- Invent an account with usage 30 then 18 and a renewal in 60 days; it must be flagged for usage.

In my version
- Not running yet. Usage numbers are the missing input; the CRM and mail parts are in place.

Reads

proposal template, call notes, CRM deal record

Runs on

an assistant, when I ask

Ask first
- Where is your template? (default: paste it once; the assistant keeps it)
- What must never be filled in without you? (default: price, discount, term length, start date)
- Where should the draft go? (default: a document in your drive; otherwise back in the chat)

Needs
- A proposal template, in any form. If none, a five-section outline: problem, approach, scope, timeline, commercials.
- The call notes or transcripts for the deal. Fallback: dictate the key points.
- The deal record for names, seats and dates. Fallback: ask the person for them.
- Chat-only version: paste the template and notes; the draft comes back in the chat.

Build
1. Read: when asked, read the template, every note for that deal and the deal record.
2. Decide: fill each section only from the notes. Any figure or date not in the notes is written as [confirm: price] rather than guessed. Use the customer's own words for the problem section when they said it.
3. Deliver: write the draft to a new document named after the account, and post the list of [confirm] items to the person. Nothing is sent to the customer.

Check
- Give it a template and notes that mention a seat count but no price; the draft must carry the seat count and show [confirm: price].

In my version
- The template is my team's; the draft lands as a document in my work drive.

Reads

calendar, last week's runs, race date

Runs on

an assistant, when I ask on Sunday

Ask first
- Race date and target time? (default: none; aim to finish)
- Runs per week? (default: four, one of them long)
- Hard limits? (default: no run before 6am, none on travel days, long run on the free weekend morning)
- Do you have last week's runs? If not, start from an easy base week.

Needs
- Your calendar for the coming week, read and write for your own calendar only. It reads only your working hours and travel; it does not read event contents.
- Last week's runs with distance and how they felt. A watch export, a note or a sentence from you.
- Chat-only version: say "plan my running week" and list your busy days; copy the sessions into your calendar yourself.

Build
1. Read: Sunday, read next week's calendar for early calls and travel, and last week's runs.
2. Decide: place the long run on the first free weekend morning, two easy runs on days with no event before 8am, one faster session midweek. Keep weekly distance within 10% of last week unless last week was a rest week.
3. Deliver: add each session as a calendar entry with distance and pace in the title. Ask before overwriting an existing entry.

Check
- Run it against a test week with a 6am Tuesday call and a Thursday flight; no run lands on those mornings.

In my version
- Four runs a week, race in early spring; sessions land in my personal calendar.

Reads

a photo

Runs on

an assistant, when I send a photo

Ask first
- Anything you will not eat? (default: none)
- How long to cook? (default: 30 minutes)

Needs
- A photo of the open fridge. Chat-only version: list what you have in a line instead.

Build
1. Read: the photo, and the answers above.
2. Decide: three dinners that use what is visible and need at most one bought item each. Say what to use up first.
3. Deliver: the three ideas back in the chat, each in three lines, with the one item to buy.

Check
- Send a photo with one obviously tired vegetable; the first idea must use it.

In my version
- I send the photo from my phone most weeknights.