— Free field guide · No sign-in
Five AI workflows sales teams can put to work now.
Five places sales work gets stuck—and for each one, a working instruction file, sample data to test it with and the output it produces.
Everything here runs in the AI tools your organization already has—ChatGPT, Claude, Gemini or Copilot. The examples use fictional businesses, people and numbers. And if you lead the organization, you don’t have to run any of this yourself: the Monday email hands the 30-day test to your sales leader.
The operating rule
AI does the first pass. Your team makes the call.
Give it only information your organization allows. Keep the source material and missing facts visible. Nothing is sent, priced or saved until the right person checks it. Every instruction file below has that rule built in.
01 · Follow-through
Right after the call
Leave the call with the follow-up nearly done.
Good calls lose momentum while notes sit in a notebook, the CRM waits for an update and the promised follow-up goes unwritten.
What goes in
- Call notes or a transcript your team is allowed to use
- Known account and contact details
- Any owner or date actually stated in the conversation
What AI prepares
- A concise meeting recap
- Suggested deal stage, with the evidence behind it
- An action list that leaves missing owners and dates blank
- Suggested CRM updates and a draft follow-up email
The demonstration
All businesses, people and numbers are fictional
What went in
Synthetic call notes, as messy as the real thing
Call notes — every business, person and number below is fictional. call w/ Dana P., Hartley's Home & Garden, Tues ~40 min, her office wants a "spring push" — garden centers open mid-April, wants presence before that BIG issue: last co-op campaign (fall, thru the old rep) — she never got a single report. "I have no idea if it did anything." brought it up twice budget — she said "probably $2,500 a month, maybe more if I can show my brother it works" (brother = Rob, co-owner, handles the books) mentioned they're redoing the outdoor kitchen display area in March, thinks that could be a story? asked what we have for younger homeowners, didn't get into it she wants numbers monthly. non-negotiable for her NEXT: she said call her after the 15th once she's back from the buying trip also — invoice thing from Nov still unresolved? she thinks she overpaid, check w/ billing
What came out
Output from an AI assistant given the instruction file below and the notes above, trimmed for length. Your wording will differ; the shape should not.
Recap
- Met with Dana P. at Hartley’s Home & Garden, in person, about 40 minutes.
- Dana wants a spring campaign in market before garden season opens in mid-April.
- Trust concern raised twice: the fall co-op campaign, run through a previous rep, never produced any reporting.
- Budget indicated at “probably $2,500 a month,” possibly more if results can be shown to Rob, the co-owner who handles finances.
- The store is rebuilding its outdoor kitchen display area in March; Dana suggested it could be a story.
- Dana asked about options for reaching younger homeowners; not discussed in depth.
- Monthly reporting is a stated requirement, not a preference.
- A November invoice question is unresolved; Dana believes she overpaid.
Deal stage
Discovery — budget and timing were discussed but nothing was proposed or agreed (“she said call her after the 15th”).
Actions
| Action | Owner | Due | Source |
|---|---|---|---|
| Call Dana | You | After the 15th — no exact date stated | “she said call her after the 15th once she's back from the buying trip” |
| Check the November invoice with billing | You | — | “invoice thing from Nov still unresolved… check w/ billing” |
| Bring options for reaching younger homeowners | — | — | “asked what we have for younger homeowners, didn't get into it” |
CRM update
| Field | Suggested value | Status |
|---|---|---|
| Stage | Discovery | Needs confirmation |
| Next step | Call Dana after the 15th | Stated |
| Amount | $2,500/month indicated, not committed | Needs confirmation |
| Close date | — | Not stated |
Draft email
Subject: Spring campaign — what I heard, and next steps
Check before sending
- “After the 15th” has no exact date — pin one down before it slips.
- Rob approves spending. Decide whether the proposal review should include him.
- Don’t tell Dana the invoice is handled until billing actually has it.
The instruction file — it produced the output above
Save it as follow-through.md where the team can reach it, set it once as standing instructions in ChatGPT, Claude, Gemini or Copilot — then just add the work.
# Follow-up packet — workflow instructions ## The job Turn the call notes or transcript provided with these instructions into a follow-up packet a person will review. Nothing is sent, saved or promised — every section is a draft for the seller. ## Return exactly these sections ### 1. Recap Five to eight short bullets: who met, what the advertiser is trying to accomplish, concerns raised, money discussed and anything promised by either side. Neutral language, no spin. ### 2. Deal stage One suggested stage (New / Discovery / Proposal / Verbal / Closed / At risk) with a one-line reason quoting the note that supports it. If the notes support two readings, show both. ### 3. Actions A table with columns: Action | Owner | Due | Source. Fill Owner and Due only when the notes state them; otherwise write "—". In Source, quote the words in the notes the action comes from. Never invent a deadline to make the table look complete. ### 4. CRM update Suggested field values (stage, next step, amount, close date), each marked "stated" or "needs confirmation". Leave amount and close date as "—" unless a number or date appears in the notes. ### 5. Draft email A short follow-up email (under 150 words) that confirms what was heard, states the next step and asks about anything unclear. No pricing, no performance promises, no discounts unless the notes state them. ### 6. Check before sending Two or three questions the seller must answer before this goes out: unclear owners, ambiguous numbers, anything left blank above. ## Rules that override everything above - Use only what is in the notes. Do not add names, dates, prices, inventory or claims from anywhere else. - If two parts of the notes conflict, show both and flag the conflict. Do not pick one. - If the notes are too thin to fill a section, say so rather than padding it. - Plain language, no hype.
You still decide
The seller checks every fact, adds the real owners and dates, and decides what should be saved or sent.
Measure this
Share of completed calls followed up within 24 hours
Use the same prompt on five completed calls and record the time from call to ready-to-send follow-up.
02 · Retention
Before the renewal
See which renewals need attention this week.
A cancellation is often the last signal, not the first. Earlier clues are already in account dates, late invoices, lapsed campaigns and long gaps between conversations.
What goes in
- Renewal and campaign end dates
- Invoice status
- Last meaningful contact and current account value
What AI prepares
- A prioritized list for the team to inspect
- The specific signal behind each priority
- A suggested opening question for the seller
- Revenue represented on the list
The demonstration
All businesses, people and numbers are fictional
What went in
Synthetic account export — ten rows, gaps included
Today's date: 2026-09-22 Account list — every business and number below is fictional. Account,Monthly value,Renewal date,Invoice status,Last campaign end,Last meaningful contact Valley Toyota,$6800,2026-10-20,21 days past due ($6800),running,2026-09-18 Hartley's Home & Garden,$2500,2026-10-04,current,running,2026-08-30 Bluebird Bakery,$450,,current,2026-07-10,2026-07-08 Summit Powersports,$1200,2027-03-01,current,2026-09-01,2026-09-15 Ridgeline Credit Union,$3900,2026-10-08,current,running,2026-06-02 Lakeview Dental,$800,2026-11-15,14 days past due ($800),running,2026-09-10 Carter & Sons Roofing,$2100,2026-10-25,current,2026-08-15,2026-08-15 The Corner Bookstore,$350,2026-12-01,current,running,2026-09-19 Meadowbrook Assisted Living,$1600,2026-10-02,current,running, Prairie Equipment Co,$5200,2027-01-15,45 days past due ($10400),2026-09-05,2026-07-28
What came out
Output from an AI assistant given the instruction file and the rows above, trimmed for length. Ten rows fit on a slide; the same prompt handles a full account list.
This week
| Rank | Account | Value at stake | Signals | First question |
|---|---|---|---|---|
| 1 | Prairie Equipment Co | $10,400 past due · $5,200/mo | Invoice 45 days past due ($10,400); campaign ended 2026-09-05 with nothing following; no meaningful contact in 56 days (since 2026-07-28) | What changed for you after the September campaign wrapped? |
| 2 | Ridgeline Credit Union | $3,900/mo | Renews 2026-10-08, 16 days out; no meaningful contact in 112 days (since 2026-06-02) | What’s moved up or down your priority list since we talked in June? |
| 3 | Meadowbrook Assisted Living | $1,600/mo | Renews 2026-10-02, 10 days out; last meaningful contact not recorded | How is the current campaign working from your side? |
| 4 | Valley Toyota | $6,800/mo | Invoice 21 days past due ($6,800); renews 2026-10-20, 28 days out | Is there anything on the invoice we should straighten out before we talk about next year? |
| 5 | Hartley's Home & Garden | $2,500/mo | Renews 2026-10-04, 12 days out; invoices current; last contact 23 days ago (2026-08-30) | What would you need to see from this fall to feel good about renewing? |
Ranks 2 and 3 are close: Meadowbrook renews sooner, but Ridgeline pairs its renewal with the longest silence on the list.
Watch
- Carter & Sons Roofing ($2,100/mo): campaign ended 2026-08-15 with nothing following; renews 2026-10-25 — likely moves up next week.
- Bluebird Bakery ($450/mo): campaign lapsed 74 days ago (2026-07-10); renewal date missing.
- Lakeview Dental ($800/mo): invoice 14 days past due ($800); renews 2026-11-15.
Totals
- Monthly value in the this-week group: $5,200 + $3,900 + $1,600 + $6,800 + $2,500 = $20,000 per month.
- Past-due invoices in the group: $10,400 (Prairie) + $6,800 (Valley Toyota) = $17,200.
Data gaps
- Meadowbrook Assisted Living: last meaningful contact is blank — treated as unknown, which raised its rank. Confirm before calling.
- Bluebird Bakery: renewal date missing — urgency can’t be judged.
- Campaigns marked “running” carry no end dates in this export, so the lapse signal only applies to ended campaigns.
The instruction file — it produced the output above
Save it as retention.md where the team can reach it, set it once as standing instructions in ChatGPT, Claude, Gemini or Copilot — then just add the work.
# Renewal watch list — workflow instructions ## The job Turn the account rows provided into a ranked "look at these this week" list a sales manager will review. Surface signals; never predict, diagnose or write off an account. ## What arrives with these instructions Account rows pasted as CSV or a table, or attached as a file, along with today's date. Expected fields: account name, monthly value, renewal date, invoice status, last campaign end and last meaningful contact. Work with what is present and list what is missing. ## Rank using only these signals - Renewal inside 30 days - Invoice past due - Campaign ended with nothing following it - No meaningful contact in 45+ days - Account value — size the dollars at stake; a small account with every signal can still rank below a large account near renewal ## Return exactly these sections ### 1. This week Accounts to look at, ranked, as a table: Rank | Account | Value at stake | Signals | First question. Signals must name the specific fields and dates driving the rank. First questions must be open questions about the account's situation, not scripts. ### 2. Watch Accounts with one weaker signal, one line each. ### 3. Totals Dollars represented in the this-week group, with the math shown. ### 4. Data gaps Missing fields that made the ranking less reliable, and which accounts they affect. ## Rules that override everything above - Never state or imply why an advertiser is drifting. "No contact in 62 days" is a fact; "they're unhappy" is a guess. Facts only. - Never label an account lost or churned. - Show every date calculation against today's date. If today's date is missing, ask for it before ranking. - If two accounts rank close together, say so rather than forcing a false order.
You still decide
The manager confirms the account data and the seller decides how to approach the customer. A signal is a reason to look—not a prediction that the account is lost.
Measure this
At-risk revenue contacted before the renewal window closes
Choose three signals, run one team’s active-account list and review the top ten together.
03 · Inbound
When an inquiry arrives
Get from inquiry to a proposal starting point faster.
A promising email arrives with a business goal but not enough detail to price or propose responsibly. The reply stalls while someone gathers the missing pieces.
What goes in
- The original inquiry
- Inventory, packages and pricing rules your team can use
- Known account context and any seller-supplied price
What AI prepares
- A plain-language read of what the buyer wants
- The questions that still need answers
- Two or three package directions grounded in your inventory
- A draft reply and suggested CRM updates
The demonstration
All businesses, people and numbers are fictional
What went in
Synthetic inquiry, plus what the seller supplied
Every business, person and number below is fictional. --- The inquiry, as it arrived --- From: Mike T. <mike@summitpowersports.example> Subject: advertising?? Hey — not sure who this goes to. We're Summit Powersports out on Route 9. We do the RV show every March and honestly that's most of our year right there, we mostly want people showing up at the show. My cousin said you guys have some kind of digital option now on top of the paper? What does something like that cost? We did some radio last year, hard to say if it did anything. Anyway let me know. — Mike --- What the seller supplied --- Approved packages I can talk about: 1. Show Countdown: print + daily site placements the 3 weeks before an event 2. Countdown + Reach: same, plus targeted email and social extension No approved rates supplied yet — pricing goes through the sales manager. Known: Summit is not a current advertiser; no account history in the CRM.
What came out
Output from an AI assistant given the instruction file and the material above, trimmed for length. Note what it refuses to do: no price appears anywhere.
What they want
Mike’s goal is attendance at the March RV show — a time-bound foot-traffic goal. “Digital option” is his cousin’s phrase, not his requirement; treat channels as means, not the ask. He is also signaling low trust in advertising he can’t measure (“radio last year, hard to say if it did anything”).
What we still need
- Budget — no figure or range was mentioned.
- Timing — the show date, and when he’d want the lead-up to start.
- Decision — whether anyone besides Mike signs off.
- Success — what “people showing up” looks like in numbers to him.
- History — what the radio effort was, so the plan doesn’t repeat what he already distrusts.
Directions
- Show Countdown — print plus daily site placements for the three weeks before the show. Fits because his year concentrates on one date, and this concentrates spend on the weeks that decide it. Price pending — seller to supply.
- Countdown + Reach — the same, plus targeted email and social extension. Fits if budget allows reaching beyond current readers in the final weeks. Price pending — seller to supply.
- Only two approved packages were supplied, so only two directions are offered.
Draft reply
Subject: RE: advertising??
CRM entry
| Field | Value |
|---|---|
| Source | Inbound email |
| Contact | Mike T. |
| Company | Summit Powersports, Route 9 |
| Stated goal | Attendance at the March RV show |
| Indicated timing | Lead-up to March; show date not yet known |
| Amount | — |
| Next step | Reply sent; propose 15-minute call |
The instruction file — it produced the output above
Save it as inbound.md where the team can reach it, set it once as standing instructions in ChatGPT, Claude, Gemini or Copilot — then just add the work.
# Inbound inquiry — workflow instructions ## The job Turn an inbound advertising inquiry into a proposal starting point a seller will finish. Never quote a price the seller did not supply; never invent inventory. ## What arrives with these instructions (a) The inquiry as it arrived, and (b) whatever the seller can share: approved packages or inventory, rate information if any, and anything known about the account. ## Return exactly these sections ### 1. What they want The buyer's business goal in plain language, separated from the product words they used. A buyer who asks about "some digital thing" may really be asking for foot traffic. ### 2. What we still need The questions that must be answered before proposing: budget, timing, decision-makers, past efforts, how they'd judge success. List only the ones the inquiry leaves open. ### 3. Directions Two or three package directions built only from the supplied inventory, each with what it is, why it fits the stated goal, and a price — a number only if a usable rate was supplied, otherwise "price pending — seller to supply". If fewer packages were supplied than that, say so instead of inventing more. ### 4. Draft reply A short response (under 120 words) that shows the inquiry was understood, offers a call and asks the two most important open questions. No price unless one was supplied. ### 5. CRM entry Suggested fields: source, contact, company, stated goal, indicated timing, next step. ## Rules that override everything above - No invented inventory, availability, audience sizes, performance claims or discounts. If no audience numbers were supplied, the directions contain none. - The reply makes no promise about results. - If the inquiry is too vague to propose anything, say so and make "What we still need" the centerpiece.
You still decide
The seller chooses the direction and supplies or approves the price. Keep inventory, audience claims, availability and expected results tied to information your team supplied.
Measure this
Time from inbound inquiry to a seller-approved first draft
Use five recent inquiries and compare the draft with what the team ultimately proposed.
04 · Preparation
Before the meeting
Walk into the call with a brief you can use.
Sellers either walk in underprepared or spend too long pulling together account history, public information and a useful opening question.
What goes in
- The seller’s objective for the call
- Account history and prior campaign results
- The offer plus relevant public information with links to its sources
What AI prepares
- A one-page call brief
- What to lead with—and what not to assume
- Five questions worth asking
- Sources and missing information kept visible
The demonstration
All businesses, people and numbers are fictional
What went in
Synthetic objective, account history and sourced context
Every business, person and number below is fictional. Objective: Dana at Hartley's Home & Garden agreed to a call after the 15th. Goal for this call: agree on a spring campaign shape and budget, and set a proposal date. Account history (our CRM): - Fall co-op campaign ran Sept–Nov through a previous rep; no reporting was ever sent to the client. Dana raised this twice on the last call. - Dana indicated "probably $2,500 a month, maybe more if I can show my brother it works." Her brother Rob co-owns the store and handles finances. - Last call notes: store is rebuilding its outdoor kitchen display area in March; garden season opens mid-April; Dana wants monthly numbers. Public information (allowed sources): - "Hartley's Home & Garden marks 40 years on Elm Street" — Ridgeline Courier, 2026-06-14 - Hartley's announced a Saturday workshop series (container gardening, raised beds) on its public Facebook page, posts dated 2026-08-02 and 2026-08-20 - A regional chain, GreenBarn, opened a location 6 miles away — Ridgeline Courier business brief, 2026-07-30
What came out
Output from an AI assistant given the instruction file and the material above, trimmed for length. Every fact stays attached to where it came from.
Lead with
Hartley’s is having a milestone year — 40 years on Elm Street — just as a regional chain opened six miles away. The spring campaign conversation is really a differentiation conversation: the anniversary, the rebuilt outdoor kitchen area and the workshop series are a story a chain can’t tell. And because the last campaign burned trust on reporting, the shape of this one has to make the monthly numbers a feature, not a promise.
Five questions
- How was the buying trip — anything you’re excited to push this spring?
- What will the outdoor kitchen area look like when it’s done, and when should people first see it?
- The Saturday workshops — who’s showing up, and who do you wish were?
- What would monthly reporting need to show for Rob to call this a win?
- Has GreenBarn’s opening changed anything for you since July?
What we know
- The fall co-op campaign produced no client reporting; Dana raised it twice (our CRM, last call notes).
- Indicated budget “probably $2,500 a month,” more if results can be shown to Rob, co-owner and finance decision-maker (our CRM, last call notes).
- Outdoor kitchen display rebuild lands in March; garden season opens mid-April (our CRM, last call notes).
- Hartley’s marks 40 years on Elm Street this year (Ridgeline Courier, 2026-06-14).
- A Saturday workshop series is running: container gardening, raised beds (Hartley’s public Facebook, 2026-08-02 and 2026-08-20).
- GreenBarn opened a location six miles away (Ridgeline Courier business brief, 2026-07-30).
Worth testing
- Hypothesis: GreenBarn’s opening is adding urgency to spring. Test with question five.
- Hypothesis: Rob is the real approval step, and a proposal stalls without him. Test: ask whether Rob should be in the room when the proposal comes.
- Hypothesis: the workshops are the bridge to the younger homeowners Dana asked about. Test with question three.
Don’t assume
- That the $2,500 is committed — it was “probably,” and contingent on proof.
- That GreenBarn has actually hurt business. Nothing supplied says so.
- That the reporting failure is forgiven just because a new campaign is being discussed.
If your tool can search the web
Most now can. Let it gather the public context itself if your organization allows that—but hold the line in the instructions: a public fact without a working link beside it doesn’t go in the brief.
The instruction file — it produced the output above
Save it as preparation.md where the team can reach it, set it once as standing instructions in ChatGPT, Claude, Gemini or Copilot — then just add the work.
# Call brief — workflow instructions
## The job
Turn the seller's objective, account history and supplied public information into a one-page brief the seller can read in the two minutes before the call.
## What arrives with these instructions
The objective for the call, whatever account history exists, and public information the organization allows — each public item with its source (publication and date, or a link).
If you can search the web and the seller says the organization allows it, you may add public facts you find on your own — each must carry a working link, and anything you cannot link goes in "Worth testing".
## Return exactly these sections, in under one page
### 1. Lead with
Two or three sentences: the single most relevant thing happening in this advertiser's business, and how the call's objective connects to it. Business first, product second.
### 2. Five questions
Questions worth asking, ordered from easiest to hardest. Good questions come from this account's situation, not a generic checklist.
### 3. What we know
Facts, one line each, with the source in parentheses. Account history counts as a source ("our CRM, last call notes").
### 4. Worth testing
Two or three hypotheses that would change the conversation if true, each labeled as a hypothesis and paired with the question that would test it.
### 5. Don't assume
What the materials do not establish, so the seller doesn't walk in claiming it.
## Rules that override everything above
- Every fact in "What we know" carries its source. A claim supplied without a source goes in "Worth testing" instead — and so does any web fact without a working link.
- No inferences about anyone's personal life and no use of personal traits. This brief is about a business.
- If the objective and the evidence point in different directions, say so at the top.You still decide
The seller opens the sources, checks that each fact matters and removes any assumption that is not supported.
Measure this
Prep time per priority call and whether the questions improved the conversation
Prepare briefs for three priority calls, then ask sellers which parts helped the conversation.
05 · Practice
Before the real objection
Rehearse the hard part before the buyer hears it.
The first real objection should not be the first time a seller has tried to answer it.
What goes in
- Buyer and business context your team can use
- The offer and the desired next step
- Likely objections raised in prior conversations
What AI prepares
- One realistic objection at a time
- Coaching after the seller responds
- A stronger question or reframe to try again
- A short practice note for the manager
The demonstration
All businesses, people and numbers are fictional
What went in
Synthetic scenario, drawn from stated business priorities
Every business, person and number below is fictional. Offer: launch sponsorship of our new high-school sports newsletter — weekly, covering six schools; the sponsor is present in every edition plus a monthly presence in print. Buyer: marketing director at Ridgeline Credit Union. Business priorities the buyer has stated publicly: growing new checking accounts and a youth financial-literacy program. Likely objections, heard in prior conversations: 1. "We already sponsor the county fair, and I can't prove that does anything." 2. "High-school sports feels like a banner at the field. How is this different?" 3. "Our board asks what we got for every sponsorship dollar."
What came out
One loop of the rehearsal, from an AI assistant given the instruction file and scenario above. In live practice the seller’s lines are yours—spoken aloud in your app’s voice mode—and they’re scripted here so the coaching loop is visible.
Buyer
I’ll be straight with you — we already sponsor the county fair, and I couldn’t tell my board what that does for us. Why would this be any different?
Seller
That’s fair. But this newsletter goes out every week to local families — it’s a whole season of visibility instead of one weekend. Your name would be in front of the exact community you serve.
Coach
What worked: you acknowledged the concern instead of talking past it. What was missed: the objection was about proof, and the answer was about exposure — more visibility is the thing the buyer already can’t measure. Neither stated priority (new checking accounts, youth financial literacy) appeared. Try: “What does the fair sponsorship report back to you the Monday after it ends?” — then contrast it with what this reports monthly. Try again, or next objection?
Seller
Let me try again. What does the fair report back to you the Monday after it ends? … Right — that’s the difference here. This reports monthly: editions, placements and what they drew. And the audience is six schools’ worth of the same families your youth financial-literacy program was built for. If it isn’t showing you anything by month three, you’ll know — which is more than the fair can say.
Buyer
Monthly reporting would honestly help with the board. What I’d still need to understand is what it costs against what the fair costs us — that’s the comparison they’ll ask for.
Practice note
- Rehearsed: the measurement objection on the newsletter launch sponsorship, two attempts.
- Improved: the second attempt turned the objection into a question and tied the answer to the buyer’s stated priorities.
- Work on next: the cost-comparison question the buyer ended on — bring the fair-versus-newsletter comparison to the next rehearsal.
Do this one out loud
Every major AI app now has a voice mode, and a rehearsal you speak is worth far more than one you type. Set the scenario up in text, switch to voice on your phone and run it aloud—five minutes in the car before the call. Say “wrap up” and the practice note arrives in text.
The instruction file — it produced the output above
Save it as practice.md where the team can reach it, set it once as standing instructions in ChatGPT, Claude, Gemini or Copilot — then just add the work.
# Sales rehearsal — workflow instructions ## The job Run a two-mode practice session — Buyer and Coach — for a seller rehearsing a real conversation. Alternate: one objection, then coaching, then let the seller retry. ## What arrives with these instructions The offer, the buyer's role and business context, and the objections that are likely, drawn from real prior conversations. The rehearsal may be spoken or typed — the rules are the same. In voice, keep every turn short and conversational, and treat a spoken "wrap up" the same as a typed one. ## Buyer mode - Play the buyer realistically: busy, fair, unconvinced. Not hostile, not a pushover. - Raise ONE objection at a time, in the buyer's natural voice, then stop and wait for the seller's response. - Stay strictly inside the business context supplied. If the buyer wasn't given a fact, the buyer doesn't invent one. - If the response genuinely addresses the objection, respond the way a fair buyer would — movement is allowed. If it doesn't, push back once, specifically. ## Coach mode — after each response, before the next objection 1. What worked — one line, specific. 2. What was missed — one line, specific. 3. One stronger question or reframe to try, worded exactly. 4. Then ask: "Try again, or next objection?" ## On "wrap up" Give a practice note: three lines a sales manager could read — what was rehearsed, what improved during the session, what to work on next. Never a score or a grade. ## Rules that override everything above - No invented personal motives, no personality profiling, no sensitive traits. The buyer is a business role, not a psychological portrait. - Coaching addresses the response, not the person. - This is rehearsal, not evaluation. Never produce anything positioned as a performance review.
You still decide
Use business context, not sensitive traits or invented motives. The manager—not the AI—judges the seller’s performance.
Measure this
Share of priority calls rehearsed beforehand
Choose one live offer and run a ten-minute spoken rehearsal before three priority conversations.
— Set it up once
These run where your team already works.
Each of the five instruction files works as standing instructions in the tools most organizations already pay for. Set it up once; after that the team just pastes the work and reviews what comes back—typed, or out loud in each app’s voice mode when the work is a rehearsal.
ChatGPT
Create a Project (or a custom GPT), add the instruction file as its instructions and name it after the workflow. The team opens it and adds the work.
Claude
Create a Project and drop the file into its instructions. Same habit: open, add the notes or rows, review what comes back.
Gemini
Create a Gem with the file as its instructions. It behaves the same way: standing instructions, fresh work added each time.
Microsoft Copilot
Build an agent from the instruction file if your plan allows it; otherwise keep the file where the team can reach it and paste it at the top of a new chat.
Before rollout
Use work accounts on a business tier
Not personal free accounts. Confirm with the vendor that your organization’s data is not used to train models, and that retention matches your policy.
Before rollout
Decide what may be pasted in—before anyone pastes it
Real customer names and CRM exports, or not. Nothing regulated, nothing personally sensitive. Write the rule down; it’s the boundary the owning manager enforces.
— Manager scorecard
Know whether the pilot is actually working.
Write down where the team starts. Check a few real examples each week. On day 30, compare the result and choose: keep it, change it or stop.
| Workflow | Owner | Baseline | Measure | Decision |
|---|---|---|---|---|
| Follow-through | Share of completed calls followed up within 24 hours | Keep / Change / Stop | ||
| Retention | At-risk revenue contacted before the renewal window closes | Keep / Change / Stop | ||
| Inbound | Time from inbound inquiry to a seller-approved first draft | Keep / Change / Stop | ||
| Preparation | Prep time per priority call and whether the questions improved the conversation | Keep / Change / Stop | ||
| Practice | Share of priority calls rehearsed beforehand | Keep / Change / Stop |
Quality
Check five examples. What did a person have to correct or reject?
Safety
Did the process use information it should not—or move ahead without review?
Value
Did the result improve enough to keep using it?
— 30-day pilot guide
One workflow. One owner. One measure that matters.
- Before day 101
Choose one place work gets stuck
Pick a recurring sales task with a visible cost. Test one workflow, not all five at once.
- Days 1–302
Choose an owner and set the rules
One manager owns the test. Write down what information the team may use, what AI may prepare and who must review the result.
- Days 1–703
Measure today’s process
Look at five to ten recent examples before changing anything. Record the current time, missed steps, revenue involved or practice completed.
- Days 8–2904
Use it on real work
Use the same prompt and process each time. Review a few examples every week and record what people changed, rejected or could not use.
- Day 3005
Keep it, change it or stop
Compare the result with where you started. Continue only if the workflow improved the work without creating new risk.
Pilot card
Fill this out before day one.
Start:Decide:
- Workflow
- Blank field
- Accountable owner
- Blank field
- Where work gets stuck
- Blank field
- Information we can use
- Blank field
- Who reviews the result
- Blank field
- Where we are today
- Blank field
- What should improve
- Blank field
- Decision
- Keep / Change / Stop
— For the executive
Your whole job here fits in one email.
Nothing here requires you to learn a tool. Pick an owner, ask for a baseline and expect a result in 30 days—this email covers it. Edit the brackets and send it Monday morning.
Subject: One AI workflow to test for 30 days
Coming soon for local publishers
What I’m building next for publishers
Seiche helps local publishers turn fragmented public information and reporting into a persistent, source-linked civic record—so communities can understand what is changing and what happens next.
Learn more and share feedbackIf one of these workflows matters but the handoffs still live in email, spreadsheets and memory, Built Revenue can help turn it into a workflow the team can actually run.