Deals die in the two weeks nobody noticed had passed

I have a call with {name}, {title} at {company}, in 40 minutes.
It is our {first / second / third} conversation.

Here is what exists: {paste last email thread, previous
call notes, and the CRM opportunity record}.

Give me, on one screen:
1. Where we actually left it, in two sentences.
2. The three things they have said matter to them, quoted
   in their words, not summarised in mine.
3. What I promised last time and whether I did it.
4. Four questions I have not asked yet that would move
   this forward.
5. The single most likely reason this deal stalls, based
   only on what is in the material above.

If the material does not support a point, write
"not covered" rather than inferring it.

The line doing the work is the last one. Without it you get a confident paragraph about their “digital transformation priorities” that nobody in the thread ever said, and you repeat it back to them on the call. “Not covered” is the output you want most, because it tells you what to ask. Point three is the second most useful: broken promises are the quietest deal killer and you will not remember them.

Here are my raw notes from a discovery call. They are
half sentences and typos. Do not clean them up into
marketing language.

{paste notes or transcript}

Output four blocks:

SITUATION: what is true about their setup today.
PAIN: what is broken, with the cost they attached to it.
   If they did not attach a cost, say so.
NEXT STEPS: a table with three columns, action, owner,
   date. Owner must be a named person, either me or
   someone they named. "The team" is not an owner. If
   no date was agreed, write "no date agreed" and put
   it at the top of the list.
RISKS: anything I heard that suggests this does not
   close, including things they said politely.

Then write one paragraph I can paste into an email to
them tonight confirming the next step.

“The team is not an owner” is the line that changes the output from a document into something actionable. Unowned next steps are how deals stall while both sides believe the other one is moving. Putting the undated items at the top is deliberate: those are the ones you fix in the confirmation email, tonight, while you still have their attention.

Turn the summary above into a CRM update for the
{company} opportunity.

Constraints:
- Under 120 words. Nobody reads more.
- Written so my manager can read it cold in the forecast
  call and know what happens next without asking me.
- State the stage I think it should move to and the one
  piece of evidence for that, from the call. If the
  evidence is weak, recommend leaving the stage alone.
- Close-date change: only propose one if the customer
  said something about timing. Otherwise say "no timing
  signal, leave date".

Give me the field values separately at the end so I can
paste them: Stage, Next Step, Close Date, Amount.

The evidence requirement is the point. Stage inflation happens because moving a deal forward feels like progress and nothing forces you to say why. Making the model name the sentence that justifies the stage, and refuse when there is not one, gives you a forecast built on things people said out loud. The separated field values at the end matter more than they look: the update gets entered because pasting takes eleven seconds.

Here is every open opportunity in my pipeline, with
last activity date, stage, amount, close date and the
last note. {paste export or connect the CRM}

Today is {date}. Sort them into four buckets:

MOVING: contact in the last 7 days and a dated next
   step exists.
DRIFTING: no contact for 8 to 20 days, or a next step
   with no date.
STALLED: no contact for 21+ days, or the close date has
   passed and the stage has not changed.
FICTION: still open, close date passed by 30+ days, no
   activity in 30+ days. These are not pipeline.

For every deal in DRIFTING and STALLED, give me one
specific reopening line based on that deal's last note,
not a generic check-in. If the note has nothing to work
with, say "needs a real reason to call" instead of
inventing one.

Rank the whole list by amount at risk this quarter.

The buckets are defined in days and dates rather than judgement, so the answer is the same on a good week and a bad one. The FICTION bucket is there because everyone carries a few and nobody volunteers them. The second useful line is the refusal clause: a generic “just circling back” is worse than silence, and being told a deal needs a real reason to call is a genuine instruction, not a failure.

I am a founder running about {15} live deals myself.
My record keeping is email, calendar and a spreadsheet.
There is no CRM hygiene and there will not be one.

Here is my last 30 days of sent mail with these
companies, my calendar, and the spreadsheet.
{attach or connect}

Build me a single table: company, deal size, last
contact, who owes whom a reply, the actual next step,
and one line on whether this is real.

Then tell me the three deals where my own silence is
the blocker, and draft the sends for those three. Match
how I already write to that person, from the thread.

Do not tell me to set up a CRM.

“Who owes whom a reply” is the whole prompt. Founders lose deals to their own inbox more often than to competitors, and the ball is usually in their court on the ones they are most worried about. The last line is not a joke. Every generic assistant answer to this question is “implement a system,” which is a project, and the founder needs three emails sent before lunch.

Here is my pipeline for the quarter, with stage, amount,
close date and the last three notes on each deal.

For each deal, give me a commit level of Commit, Best
Case or Pipeline, and justify it using only these tests:
- Has the buyer said a date out loud?
- Do I know how they buy, who signs, and what else is
  queued ahead of me?
- Have they done work on their side in the last 14 days?
- Is there a written next step with a date?

A deal that fails two or more tests cannot be Commit,
regardless of how the note reads.

Then give me the number three ways: everything, Commit
only, and Commit plus half of Best Case. Tell me which
deals would have to slip for me to miss.

The tests are all observable facts, so the model is not guessing at sentiment, and neither are you. The hard rule about failing two tests is what stops the exercise turning into a re-run of your own optimism. “Which deals would have to slip for me to miss” is the question a manager asks in week ten. Answering it in week two is the whole value.