12 AI Prompts to Improve Sales Strategy (Copy-Ready)

The best AI prompts for sales strategy hand the model your real evidence (won and lost deals, buyer quotes from calls, pipeline exports) and ask for a decision you can check, not a brainstorm. The 12 copy-ready prompts below cover the ideal customer profile, segment focus, win/loss, competitors, pricing, messaging, buying committees, territory, capacity, launches and a quarterly pre-mortem.

Key takeaways

• A strategy prompt without data returns a textbook. Paste or connect closed deals, call excerpts and pipeline numbers.

• Ask for the evidence behind every conclusion (deal names, counts, buyer quotes) so you can audit the reasoning.

• Ask the model to show its math and flag thin samples. Strategy errors usually hide in small numbers.

• Run the same prompts every quarter. The change between two runs is the strategic signal.

• In Proshort, Ask Aria runs these questions across every call, deal and CRM record, with sources, and can send the answer on a schedule.

How do you write an AI prompt for sales strategy?

A sales strategy prompt has five parts: a role, the evidence, the question, the output format and a rule for uncertainty. Anthropic's guidance is to treat the model like a brilliant but new employee who lacks context, to put long documents above the question, and to ask it to quote the relevant passages before it answers. OpenAI's prompt engineering guide adds reference text and a few examples of the output you want.

Evidence matters more than ever because buyers have moved. In Salesforce's State of Sales report (7th edition), sellers say customers increasingly want ROI, personalization and extensive education before buying, and take longer to decide. A strategy built on last year's assumptions misses that shift; a strategy built on this quarter's calls does not.

The 12 sales strategy prompts at a glance

Prompts for targeting: who to sell to

1. Define your ideal customer profile from evidence

When to use it: Annual planning, or when win rate falls in a segment you thought was core.

You are a revenue strategy analyst. Below are our last [50] closed-won and [50] closed-lost deals with industry, employee count, region, deal size, cycle length, lead source and loss reason: [paste export]. Find the three traits that best separate won from lost deals. For each, give win rate with and without the trait, the number of deals behind it and two example deals. Then write a one-paragraph ICP an SDR could use to disqualify an account in 30 seconds. Mark any trait based on fewer than 10 deals as low confidence.

What good output looks like: Three traits with counts and win rates, honest low-confidence flags, and an ICP specific enough to say no to accounts.

Common mistake: Pasting only won deals. Without losses, the model describes your customer base instead of what predicts a win.

2. Choose where to focus next quarter

When to use it: Quarterly planning, when leaders disagree about which segment deserves headcount.

Using the pipeline and bookings data below, compare our segments [list segments] on win rate, average deal size, cycle length and pipeline created per rep: [paste data]. Rank them by expected bookings per rep per quarter. Recommend where to add or cut focus, and name the one assumption that would reverse your recommendation if it were wrong.

What good output looks like: A ranked table, one clear recommendation and a named assumption you can watch (for example, mid-market cycles staying under 60 days).

Common mistake: Ranking by deal size alone. Large deals with long cycles can starve a quarter of bookings.

3. Find the real reasons you lose

When to use it: Monthly win/loss review, or after two losses to the same objection.

Here are CRM loss reasons and call excerpts from [N] lost deals in [period]: [paste]. First, quote the buyer words that explain each loss. Then group the losses into no more than five root causes, ranked by revenue lost. For each cause, say whether sales process, rep skill, product, pricing or targeting owns the fix. List every deal where the CRM loss reason contradicts what the buyer said.

What good output looks like: Root causes backed by buyer quotes, plus a contradiction list (the CRM says price, the calls say no champion).

Common mistake: Trusting the CRM dropdown. Price and timing are where reps file losses they never fully understood.

4. Spot expansion-ready customers

When to use it: Monthly account review for account managers and customer success.

From the account notes and recent call summaries below, list customers showing expansion signals: new teams mentioned, new use cases asked about, a new executive sponsor, or growth plans. For each account, give the signal with its source and date, the likely expansion play, and the risk if we push too early. [paste notes]

What good output looks like: A short list where every signal has a date and a source, and a risk line that stops reps from pitching a customer still mid-rollout.

Common mistake: Treating product usage as the only signal. A new stakeholder joining a call often matters more.

Prompts for winning: how to compete and price

5. Position against a named competitor

When to use it: When a competitor starts appearing in more deals, or after a loss you did not see coming.

Act as a competitive intelligence lead. Using the calls and deals below where [Competitor] was mentioned, list the moments buyers bring them up, the claims buyers repeat about them, our win rate when they are in the deal, and three rep answers that came before a win. Then write five lines: when to compete, when to walk away, and the one question that reframes the evaluation. [paste calls and deals]

What good output looks like: Positioning grounded in what buyers said this quarter, including when to walk away.

Common mistake: Asking the model what it already knows about the competitor. Its training data is old and generic; your calls are current.

6. Read what discounting is telling you

When to use it: Quarter-close retro, or when average discount creeps up.

Here are [N] closed deals with list price, final price, discount, the stage when a discount was first offered and who asked for it: [paste]. Find which segments, reps or stages carry the deepest discounts, and whether deeper discounts came with faster closes or higher win rates. Recommend one discount guardrail and estimate its effect on these deals. Show your math.

What good output looks like: A pattern you can act on (discounts offered before the business case was agreed) and a guardrail with worked numbers.

Common mistake: Assuming discounting is a rep problem. It usually starts at a stage where value was never proven.

7. Rewrite messaging in buyer language

When to use it: Before a messaging refresh, a new website or a new pitch deck.

Below are excerpts where buyers describe, in their own words, the problem we solve: [paste excerpts]. Extract the ten phrases buyers use most, with how many calls used each. Then rewrite our value proposition [paste current version] in under 40 words using their language, and list every phrase in our version that no buyer ever said.

What good output looks like: Ten buyer phrases with counts, a plainer value proposition, and a list of internal jargon to drop.

Common mistake: Letting the model polish the copy. You want buyer words, not better marketing words.

8. Map the buying committee by segment

When to use it: Segment planning, and whenever late-stage deals stall on an unknown stakeholder.

For our won deals above [deal size] in [segment], list every role that joined a call, the stage when each first appeared, and who signed. Describe the typical buying committee, the role that most often appears late and stalls deals, and the stage by which we should have met them. [paste call attendee and deal data]

What good output looks like: A committee map per segment and a rule reps can follow, such as meeting security before the proposal.

Common mistake: Using CRM contact titles. Use who actually attended calls; contact lists are full of people who never engaged.

Prompts for planning: where to spend the quarter

9. Prioritize accounts for a territory

When to use it: Start of quarter, territory changes, or a new rep taking over a book.

Score the accounts below for [rep or team] on fit with our ICP [paste ICP], timing signals [list signals] and existing relationships. Return the top 25 with a one-line reason each, the 25 we should deprioritize, and the accounts where the data is too thin to judge. [paste account list]

What good output looks like: Three buckets with reasons, including an explicit too-thin-to-judge bucket.

Common mistake: Leaving out that third bucket. Without it, the model fills gaps with confident guesses.

10. Check that the plan adds up

When to use it: Before you commit a target to the board or hand quotas to managers.

Our target for [period] is [amount]. Here are rep headcount, ramp status, win rate by segment, average deal size and current pipeline by stage: [paste]. Work backwards: how much pipeline we need by stage, how much we have, the gap, and the three cheapest ways to close it. Show every calculation so I can check it.

What good output looks like: Step-by-step math you can verify in a spreadsheet, and gap-closing options ranked by cost.

Common mistake: Not asking for the working. Language models still make arithmetic slips, so check every number before it reaches a slide.

11. Test launch readiness

When to use it: Four weeks before a product or pricing launch.

We are launching [product or feature] for [persona]. Using the launch brief and the objections from similar launches below, write the five questions buyers will ask in the first month, the best current answer to each, the proof we still lack, and a ten-minute role-play scenario to test reps before launch. [paste brief and objections]

What good output looks like: The proof-we-still-lack list is the valuable part: it tells marketing what to build before reps get caught out.

Common mistake: Generating answers without real objections. Launch questions repeat what buyers asked last time.

12. Run a quarterly pre-mortem

When to use it: Week one of the quarter, after the plan is set.

Here is our sales strategy for [quarter]: [paste]. Assume it is the end of the quarter and we missed the number by 20%. Write the three most likely reasons, each tied to evidence in the data below, the early-warning signal we would see by week four, and the metric that would show it. [paste pipeline and call data]

What good output looks like: Three failure stories, each with a week-four signal you can put on the dashboard today.

Common mistake: Asking whether the plan is good. The model tends to agree with you; a pre-mortem makes it argue the other side.

How do these prompts work inside Proshort?

Every prompt above works better on connected evidence than on a pasted spreadsheet. Ask Aria sits on top of every deal, call and CRM record. Ask prompt 3 or prompt 5 in plain English, tag an account or deal with @, and the answer comes back with its sources as text, a chart, a table or a ready-to-send doc. Questions you repeat can be scheduled, for example as a daily 8:00 email.

If your team already works in ChatGPT or Claude, Proshort's app in ChatGPT and its Claude connector (MCP) let you run the same prompts on your own calls, deals and accounts. See the integrations page for the CRM, calendar and Slack connections.

Strategy only pays off when it reaches the call. The AI playbook is built from the questions and objections buyers actually raise, refreshed every 24 hours, and Aria Live puts the best answer, with its source, on the rep's screen during the call. Only the rep sees it. Recorders like Gong score the call afterwards; Aria helps the rep get it right on the call.

"Our win/loss reviews have also become significantly more valuable." Deepak PG, Senior Director Sales Operations, MetricStream (read the MetricStream case study)

For the rep-level and manager-level versions of these prompts, see 9 AI prompts to improve sales productivity and 7 AI prompts for pipeline reviews.

Frequently asked questions

What is the best AI prompt for sales strategy?

Start with the win/loss prompt (prompt 3). It forces the model to quote what buyers said, group losses into root causes and show where your CRM loss reasons are wrong. Most strategy changes that matter come out of that list.

Can ChatGPT or Claude build a sales strategy?

They can analyze evidence and draft options quickly, but they cannot decide for you, and they only know what you give them. Feed them real deal and call data, ask them to show their math and quote sources, and keep the final call with the people who own the number.

How often should we run sales strategy prompts?

Run targeting and capacity prompts quarterly, win/loss and competitive prompts monthly, and the pre-mortem in week one of every quarter. Comparing two runs of the same prompt is often more useful than any single answer.

Is it safe to paste CRM and call data into an AI tool?

Check your company policy first and use an enterprise workspace that does not train on your data. Remove personal details you do not need. Proshort is SOC 2 Type 2 and ISO 27001 certified; see the security page for details.

How are sales strategy prompts different from sales productivity prompts?

Strategy prompts decide where and how to compete: segments, positioning, pricing and capacity. Productivity prompts save a rep time on daily work such as call prep, follow-ups and CRM updates. Most teams need both, run by different people on different cadences.

Want to run these prompts on your own calls and deals? Book a Proshort demo and bring prompt 3.