Close more deals: 10 AI skills for sales conversations
close-plan
agree the steps to signature so nothing surprises you
How the two work together
Claude thinks it through. Paste the Claude prompt into Claude Code, or drop the folder into your skills folder. Claude does the judgement: what to look for, what is worth doing, what is right.
Codex gets it done. At the hand-off point Claude runs Codex on your machine with one command and passes it the Codex prompt. Codex does the mechanical part and hands the result back. Claude checks it before you see it.
No API key to set up: Claude calls the Codex you already have installed. If Codex is not installed, Claude does that half itself and tells you.
Prompt for Claude
---
name: close-plan
description: Builds a close plan scheduled backwards from the buyer's go-live date, so every step, owner and date to signature is agreed in writing. Use once the buyer says they want to proceed.
---
# Every step to signature agreed before you get there
A dated plan both sides have accepted, and the email that gets it accepted.
## Claude does
1. Asks for the buyer's critical go-live date and the value events after it. Everything schedules backwards from it. No date, no plan, and Claude says so rather than inventing one.
2. Asks who signs, who pays, and who owns security, legal and procurement, with how long each took last time. A won 50k to 250k deal typically involves at least ten stakeholders, so a plan naming two is unfinished.
3. Tests your champion on five signs: gives access, shares information, acts unprompted, tells you things you did not know, commits to dates. Three failures and you have no champion, and the plan dies in their inbox.
4. Places the executive step around the third touchpoint. Evaluations starting with an executive correlate with roughly 6 percent lower win rates; executives joining at the third touch, roughly 5 percent higher.
5. Turns each unknown into a step with one named owner and a date, never a status. Fifty-nine percent of sales leaders say the signing stage takes too long, and nobody plans it.
6. Writes risk, not urgency: what happens if it fails, the opt-out, the phased start, the pilot scope. Forty to sixty percent of lost deals go to indecision, and urgency backfired 84 percent of the time.
7. Recommends one sequence and scope, never a menu, so the risk is shared. Price is agreed on the phone: email only wins 5 percent of the time, email plus phone 35 percent.
8. Refuses to guess. Dates, names, legal terms, procurement rules and lead times come from you or stay a visible gap.
## Then Codex does
Codex does the arithmetic. From the critical date and your durations it counts backwards over working days, builds the dated table, writes the agreement email and the risk register, and runs the sweep. No judgement.
Fill in {CRITICAL_DATE}, {STEPS_AND_DURATIONS}, {STAKEHOLDERS} and {RISK_TERMS} in CODEX.md first, then run:
```
codex exec --sandbox danger-full-access --skip-git-repo-check -C "<working folder>" - < CODEX.md
```
## Claude checks
1. Every row names one person, not a team, with a real date.
2. The schedule starts today or later. If it starts in the past, the go-live date is already impossible and you say so now, not in week six.
3. Security, legal, procurement and the signature each have their own row.
4. The executive step is not first.
5. No urgency, no expiring discount, no invented deadline.
6. No lead time or approval rule Codex was not given.
7. The email asks the buyer to correct the plan, not approve it.
Any failure: reject, fix the input, rerun Codex. Never hand-patch.
## Rules
- Schedule backwards from the buyer's date, never forwards from your quarter.
- Every step has one named person and a date, never a status.
- No champion, no plan.
- Email the price, agree it on the phone, confirm by email. Half of written communication is misread.
- No urgency, no expiring discount, no fear.
- Public information only.
- Never invent a fact, a number or a quote.
- Anything sent in someone's name says whose name it is.
## If Codex is not installed
Claude builds the plan, the email and the register itself and says so.
## Built from
- Gong Labs, multithreading, 1M+ executive sales cycles: https://www.gong.io/blog/when-and-how-to-multi-thread-when-selling-to-executives
- MEDDICC, Andy Whyte, champion signs, paper process: https://meddicc.com/resources/5-signs-you-dont-have-a-champion
- The JOLT Effect, Matt Dixon and Ted McKenna: https://www.jolteffect.com/blog/what-is-the-jolt-effect
- Great Demo!, Peter Cohan, Transition Vision (PDF): https://cdn.prod.website-files.com/6540c97c6e5b78f21ab4a9b5/67dbb3814d8a2c44393f3105_Seven%20Validated%20Habits%20of%20Stunningly%20Successful%20Demos.pdf
- Gong Labs, negotiation channels, 37,671 deals: https://www.gong.io/resources/labs/this-deadly-simple-negotiation-mistake-is-costing-you-deals-heres-how-to-avoid-it/
- Gong Labs, product demo study, 3 million demos, next steps: https://www.gong.io/resources/labs/product-demo/
Prompt for Codex
# Close plan - Codex task
You are scheduling a plan from inputs that are already agreed. Do not invent steps, dates, owners or terms.
## You are given
- {CRITICAL_DATE}: the buyer's go-live date as YYYY-MM-DD, plus the value events that follow it with their dates.
- {STEPS_AND_DURATIONS}: every step to signature, in the order it must happen, each with an owner, a side (buyer or seller) and a duration in working days.
- {STAKEHOLDERS}: name, role, and the step each one owns or must attend. The signer is marked.
- {RISK_TERMS}: what happens if it does not work, the opt-out, the phased start and the pilot scope, in the words already approved.
## Produce
In the working folder:
1. `close-plan.md`. A table with columns Step, Owner (a named person), Side, Start, Finish, Depends on. Schedule backwards from {CRITICAL_DATE}: the last step finishes on that date, and each earlier step finishes the working day before its successor starts. Monday to Friday only. Below the table, a Dates section listing the critical date and the value events, then a Risk section copied verbatim from {RISK_TERMS}.
2. `agreement-email.md`. Maximum 130 words, addressed to the champion, attaching the plan and asking them to correct any date or owner that is wrong. Ends by proposing a short call to agree price.
3. `risk-register.md`. One row per plan step, columns Step, What could slip it, Earliest sign, Who to ask.
4. `checks.md`, holding the results below.
## Rules
- Copy names, dates, durations and terms exactly as given. Never round, recalculate or rephrase them.
- Never add a step, an owner, a duration or a stakeholder that was not supplied.
- Never add urgency, an expiring discount, or any deadline other than {CRITICAL_DATE}.
- If the backwards schedule starts before today, do not compress anything to make it fit. Report the overrun.
- A missing input stays as `[GAP: what is needed]`. Never fill it.
- British English. No em dash characters.
## Return
Print, and nothing else:
- The file list with word counts.
- The computed start date, and if it falls before today, the number of working days by which the plan overruns.
- Sweep results: any row with no named person; any row naming a team or a department instead of a person; any Start or Finish falling on a Saturday or Sunday; any Depends on that names a step not in the table; any `[GAP: ...]` markers; any brace placeholder left unfilled.
- PASS or FAIL for each of: signature step present, legal step present, security step present, procurement step present, executive step is not the first row, one risk-register row per plan row.
Built from the best public work on this
Sources
- **Gong Labs, multithreading to executives** - https://www.gong.io/blog/when-and-how-to-multi-thread-when-selling-to-executives - more than 1M executive sales cycles analysed. A won 50k to 250k deal typically involves at least ten stakeholders. Win rates drop about 6 percent when the evaluation starts with an executive and rise about 5 percent when executives are involved around the third touchpoint. Warns against pulling in dozens of contacts early without a champion first.
- **MEDDICC, Andy Whyte** - https://meddicc.com/resources/5-signs-you-dont-have-a-champion - the champion definition ("they actively sell on your behalf when you're not in the room") and the five signs you do not have one: no access to stakeholders, no information shared, not proactive, tells you nothing new, will not commit. The companion paper-process page splits the close into proposal, quote, contract management, compliance and legal, communication, and cites that 59 percent of sales leaders think the agreement and signing stage takes too long.
- **The JOLT Effect, Matt Dixon and Ted McKenna** - https://www.jolteffect.com/blog/what-is-the-jolt-effect - 40 to 60 percent of deals are lost to customer indecision, 56 percent of those to risk or fear of failure rather than a preference for the status quo. The standard urgency push backfired 84 percent of the time. Source of the single recommendation and the risk-removal section that replaces urgency in this plan.
- **Great Demo!, Peter Cohan, Seven Validated Habits (PDF)** - https://cdn.prod.website-files.com/6540c97c6e5b78f21ab4a9b5/67dbb3814d8a2c44393f3105_Seven%20Validated%20Habits%20of%20Stunningly%20Successful%20Demos.pdf - the Transition Vision: one Critical Date driving go-live plus one or more Value Realization Events. That date is the anchor everything here schedules backwards from.
- **Gong Labs, negotiation channel study** - https://www.gong.io/resources/labs/this-deadly-simple-negotiation-mistake-is-costing-you-deals-heres-how-to-avoid-it/ - 37,671 opportunities. Pricing negotiated by email only wins 5 percent, phone only 15 percent, both channels 35 percent. States that 50 percent of written communication is misunderstood by the recipient. Why the plan puts price agreement on a call.
- **Gong Labs, product demo study** - https://www.gong.io/resources/labs/product-demo/ - 3 million demos analysed. Spending 53 percent more time discussing next steps correlates with faster deal closure, and close rates plummet 71 percent when next steps are not discussed on first calls.
Best public prompt we found for this job
https://www.gong.io/blog/when-and-how-to-multi-thread-when-selling-to-executives
Verbatim language from that study for the executive step in the plan, ghostwritten for your champion to send:
"As we consider [PROBLEM], I've asked [Your Name] at [Company] to spend 30 mins with our team. They had some unique perspective on [PROBLEM], based on [X and Y]."
And the question that surfaces dissent before it kills the close:
"Who has a different opinion on the way we've outlined the problem here?"
Want this running in your business?
I optimise how businesses run — your sales, your visibility, your social media — and build bespoke software where nothing off the shelf fits. The first conversation is free. Work starts from £150 a day.
Foxera