ChatGPT for Sales and Marketing: B2B Battle Cards and One-Pagers

Build a battle card by giving ChatGPT your own offer, the competitor's public pages and the objections you hear, then asking for a grid, honest weak points and short replies, and check every claim against its source before anyone uses it. The one-pager comes from the same facts, trimmed to one side of paper.

The step most owners skip is the check. ChatGPT writes confident comparisons whether or not the competitor really offers what it says, and a rep who repeats a wrong claim in front of a buyer loses more than the card saves. The method below separates drafting, which the tool does quickly, from verifying, which stays with you.

Follow me on Instagram@sagnikteaches

What each document is for, and who reads it

A battle card is an internal sheet. A salesperson, or the owner doing the selling, reads it in the five minutes before a call where a named competitor is in the running. It says where you win, where you are weaker, what to ask, and how to answer the objections that always appear. It is allowed to be blunt because the buyer never sees it.

Connect on LinkedInSagnik Bhattacharya

A one-page service sheet is external. A prospect reads it, forwards it to a colleague and decides whether to book a conversation. It describes what you do, who it is for, what is included, how long it takes, what it costs or how pricing works, and what happens next. It must contain nothing you could not defend in writing.

Subscribe on YouTube@sagnikteaches

Both draw on one source of truth: a plain "facts file" that you build first. If you skip the file and ask ChatGPT to write from a vague prompt, it fills gaps with plausible filler, and plausible filler is exactly what gets caught out later.

The running example: a four-person operations consultancy

Everything below uses an illustrative firm. It is a four-person consultancy that helps mid-sized manufacturers tidy up their purchasing and stock processes. It sells a fixed-scope "process review" for a set fee and a monthly retainer for follow-on work. In most deals it meets one of three alternatives: a large consultancy's proposal, an independent freelancer, or the prospect doing nothing and hiring an in-house analyst.

Notice that two of those three are not named companies. Many small firms lose to a category, not a brand, so a battle card can cover "the freelancer" and "the do-it-ourselves option" as well as a specific rival. I will build one card for a named large rival, labelled "Competitor A" here, and show how the same prompts adapt to the categories.

Step one: write the facts file before you open ChatGPT

Open a document and fill in five blocks. Spend thirty to forty minutes on this, because it controls the quality of everything after it.

1. Your offer. Services, deliverables, duration, team size, fees or how they are calculated, what is explicitly out of scope.
2. Your proof. Outcomes you can document, such as "cut a client's supplier-onboarding steps from 14 to 9", with the client's permission to mention it in anonymised form. Do not include a figure you cannot trace to a project file.
3. Competitor evidence. Text pasted from their pricing page, service pages, case studies, job adverts and published terms, each with the URL and the date you copied it.
4. Objections you actually hear. Pull them from call notes and lost-deal emails. Use the buyer's words.
5. Unknowns. A list headed "do not know", for example "Competitor A's minimum engagement length". This list is what stops invention.

Here is an illustrative extract of block three for Competitor A:

Competitor A (copied 4 Nov, from their public pages)
- Services page: "Operational excellence programmes for mid-market manufacturers"
- Pricing: not published; the contact form asks for "project budget range"
- Case study page: 3 case studies, all for businesses with 500+ staff
- Careers page: 6 junior analyst roles open, 1 partner role
- Terms page: 12-month minimum for retained work
Do not know: typical team size on a mid-sized engagement; hourly rates.

That is the raw material. ChatGPT never sees a claim that is not in the file, and anything it adds goes onto the check list.

Step two: the competitor grid prompt

Paste the facts file into a new chat, or better into a ChatGPT Project so the same facts stay available for the card, the one-pager and later edits. Projects hold shared instructions and files; OpenAI's help centre says project sharing works on every plan, including Free, and project file limits vary by plan, so a short facts file fits comfortably on any of them.

You are helping a small B2B consultancy prepare a sales battle card.
Use ONLY the facts file below. If a cell in the grid needs information
that is not in the file, write "NOT IN FILE" in that cell. Never estimate.

Build a table comparing us with Competitor A, with these rows:
target customer size, service scope, pricing approach, minimum commitment,
team seniority, proof/case studies, and typical buyer concern.
Add a column called "Source" showing which part of the file supports
each cell. After the table, list every row where the evidence is weak.

FACTS FILE:
[paste]

A realistic first response puts "Mid-market manufacturers" against "500+ staff case studies" and flags that team seniority is inferred from job adverts alone. That flag is useful. It tells you the row is a guess, and a guess does not belong on a card. Illustrative output, trimmed:

The last row shows the discipline. "Junior-heavy" came from a careers page, which is a hint and not a fact, so it leaves the card. What stays is a comparison a buyer could verify for themselves.

Step three: objection handling that sounds like a person

Objections are where a battle card earns its keep, and where AI drafts are most tempting to use unedited. The usual failure is a reply that is polished but evasive. Ask for replies in a fixed shape and supply the buyer's own words.

Here are five objections from lost-deal emails, in the buyer's words:
1. "They have a bigger name."
2. "Your price looks high for a review."
3. "We can get a freelancer for half that."
4. "We worry you will not have capacity."
5. "We would rather build this in-house."

For each, write: (a) what the buyer is really worried about in one
sentence, (b) a two-sentence reply a salesperson can say aloud, using only
facts from the file, (c) one question to ask back, (d) what NOT to say.
Keep each reply under 50 words. No claims about the competitor that are
not in the file.

Illustrative output for objection 2, with what I would change:

(a) They cannot see what the fee buys, so it feels like an arbitrary number. (b) "The review is a fixed fee for a fixed scope: the process map, the findings report and a prioritised plan, delivered in six weeks. You will know the total before we start." (c) "What would you need to see to feel the scope is right?" (d) Do not say "you get what you pay for."

The reply is decent, but I would check the six weeks against what the firm really delivers and swap in the true figure. I would also cut anything that sounds like a slogan. A sales call is a conversation, and a rep who recites a paragraph sounds as though they are reading from a card, because they are.

Step four: the verification pass, claim by claim

Now audit the draft. Ask ChatGPT to extract its own claims so you can check them one at a time.

List every factual claim in the draft battle card as a numbered list.
For each, give: the claim, the sentence from the facts file that
supports it (quote it), and a label: SUPPORTED, PARTLY SUPPORTED, or
NOT SUPPORTED. Do not defend any claim; if unsure, say NOT SUPPORTED.

Then do the human part. For every claim marked SUPPORTED, open the source URL and confirm the wording still matches, since pages change. For PARTLY SUPPORTED, reword to the narrower statement. For NOT SUPPORTED, delete it or move it to the "to investigate" list. Record the check date next to each claim on the final card.

A worked tally for the Competitor A card: of 22 claims in the first draft, 15 were supported, 4 were partly supported and 3 were not supported. The three unsupported lines included a statement that Competitor A "charges by the hour", which the facts file never says. It sounded right because many consultancies do charge that way, which is precisely the kind of thing a model asserts from general patterns. Cutting those three lines cost nothing and protected the rep from a public correction.

Also check these categories whatever the file says: prices and fees (could be stale), named customers (never imply a rival's client is unhappy), anything about a rival's legal or financial situation, and any superlative such as "the only firm that". If you cannot prove an "only", delete the word.

Step five: the one-page service sheet

With the verified facts file, the one-pager is quick. Ask for structure first, then copy.

Using only the verified facts file, draft a one-page service sheet for
"Process Review" aimed at an operations director at a manufacturer with
50 to 300 staff. Sections: the problem in the buyer's words (2 sentences),
what we do (4 bullets), what you receive (deliverables list), how it runs
(week-by-week, 6 weeks), who it suits and who it does not, how we price it,
and what happens next. Maximum 330 words. Plain English, no jargon,
no claims about competitors.

A "who it does not suit" line is the part most one-pagers omit and the part that builds trust: "Not suited to firms looking for a long-term outsourced operations team." It filters out the wrong enquiries, which saves your time.

Here is a before and after of one section. The first draft from the model:

We deliver best-in-class operational transformation to improve efficiency across your organisation.

After editing against the facts file:

In six weeks we map how a purchase order moves from request to payment, find the steps that cause delays, and give you a ranked plan with the effort and saving for each fix.

The second version names a process, a timeframe and a deliverable. A reader can picture what they would receive. If the first draft sounds like a brochure from any firm, it is because the model had nothing specific; the fix is more of your own facts, not a cleverer prompt.

Adapting the card to the freelancer and the do-it-yourself option

For a category rival, there is no web page to copy, so the facts file has to come from your own experience: lost-deal notes, what buyers said about the alternative, and your delivery records. Tell ChatGPT that explicitly.

The "competitor" here is a category: an independent freelancer
doing the same work. There are no public pages. Use only the
buyer comments I paste below. Build the same card, but mark every
statement about freelancers in general as "buyer-reported" and do
not present any as fact about freelancers.

The honest framing for a freelancer matters. Freelancers are often excellent, so the card should say when one is the right choice: a small, well-defined task with a single owner and no need for cover during holidays. Where you win is continuity, a second pair of eyes on the findings and a documented method the client can keep. Say that, and never disparage the person. For the in-house option, the useful questions are "who owns this after the analyst leaves?" and "how long until they are productive?"

A filled-in card, as a rep would see it

To make the output concrete, here is how the finished one-page card for Competitor A might read after the claim check. The wording is illustrative, and every line maps to a row in the facts file.

BATTLE CARD: Competitor A          Reviewed 4 Nov  |  Owner: [your name]
Position: they sell large programmes; we sell a short, fixed-scope review.
Where we win:
 1. Fixed fee, known before we start (their pricing is by enquiry).
 2. No retained minimum (their terms page shows 12 months).
 3. A principal leads the work (our own staffing rule).
Where they are stronger (say so, do not hide it):
 1. Brand recognition with larger buyers.
 2. Breadth: they list services we do not offer.
Ask early: "Is the aim a quick diagnosis or a long programme?"
Do not say: anything about their staffing, rates or clients.
Proof: [anonymised outcome, with project reference]

Keep this to one side of paper. A rep will not scroll through a research dossier before a call, and a card that fits at a glance is the one that gets opened. If a line cannot be traced to the file, it does not appear here.

Two practical notes on using it. First, store the card where reps already work, such as the shared drive folder for the deal or pinned in the team chat, because a card that lives in the marketing folder is never seen. Second, version the file name with the review date, so two people never work from different copies. When you change a row, add one line to the change log saying what moved and why, for example "pricing row updated after the competitor published a fee page on 14 Nov". Six months later that log tells you which of your claims go stale fastest, and those are the rows to check monthly.

Settings and data handling for sales material

Battle cards and service sheets need little confidential data, which makes them a good first use of ChatGPT. Still, keep client names, contract values and anything marked confidential out of the prompts. On personal plans, switch off "Improve the model for everyone", which sits under Settings, then Data controls. ChatGPT Business does not train on your business data by default, according to OpenAI's business pricing page, which lists a Standard seat at $20 a user a month billed annually or $25 billed monthly, for teams of 2 to 200. Prices change, so confirm on that page before you budget.

If you already use a Project, keep the facts file in it and the instructions short: "Use only the supplied files; mark unknowns as NOT IN FILE." Do not build this on a Custom GPT: OpenAI's retirement FAQ says custom GPTs are scheduled to retire on 11 December 2026, with existing ones moving to plugins, so a Project is the sturdier home.

Keeping the card alive

A battle card decays quickly. Give it an owner, a review date and a change log. When a call surfaces something new, such as a competitor's new minimum term, the rep sends one line to the owner, who updates the facts file, regenerates the affected row, and runs the claim check again on that row only.

Track whether the card is used. After each competitive deal, ask the rep two questions: did you open the card, and was anything on it wrong or missing? If nobody opens it, it is too long. If a line is wrong, trace it back to the facts file. Over a quarter, you will find the real fix is nearly always a better source, not a better prompt.

For a wider research habit, see how to do competitor research with AI in one afternoon, and for a card built purely from public sources, the public-research battlecard method. Price rows go stale fastest, so tracking competitors' prices and offers is worth setting up beside the card. If the positioning line at the top is the weak part, the product-marketing positioning tutorial helps. And for the broader picture of where AI fits in a firm, there is the AI guide.

Three more illustrations from other firms

The same routine travels. A recruitment agency selling to hiring managers builds its card against "posting the job ourselves" and "the large agency", and its objection list leans on speed to shortlist, which it must support from its own placement records. A dental practice offering a corporate check-up arrangement to employers uses the one-pager as its main offer document and the card as a briefing for the practice manager, with the competitor row covering a rival practice's published check-up packages.

An estate agency courting a developer for a block of new homes builds the card around fee structure, marketing reach and reporting, then runs the claim check hardest on any statement about a rival's sale times, because that figure is easy to misquote. In all three, the order never changes: facts file, grid, objections, claim audit, one-pager, review date.

A short quality checklist before sending anything out

• Every competitor claim has a source and a check date.
• No "only", "best" or "leading" that you cannot prove.
• The card lists where the competitor is stronger, and says how to respond honestly.
• Replies are under 50 words and sound like speech.
• The one-pager has a "who it does not suit" line and a clear next step.
• Numbers on the one-pager match your delivery records exactly.
• No client names, no confidential figures, no contract values.
• A named owner and a quarterly review date appear on both documents.

One last point on time. Expect around two hours for the first card and one-pager, mostly spent on the facts file and the claim check, and about thirty minutes for each refresh. The drafting itself takes minutes. The value is in the checking, which is why it is worth doing every time.

When you hand the card to a new salesperson, run a ten-minute role-play with them using the objections list. Have them answer the first two aloud without reading, then compare to the card. If they stumble on a line, the line is too long or too clever, so shorten it. Adjust the card after that session, not before, because real speech shows what the page hides.

A final example of a mistake worth guarding against. A rep skimming the card might quote the "12-month minimum" to a buyer as if it applied to every Competitor A engagement, when the terms page only mentions it for retained work. The buyer, who had a one-off project in mind, then discovers the claim is wrong and doubts everything else the rep said. The fix is in the wording on the card itself: write "12-month minimum for retained work (their terms page)", with the qualifier included, so the narrow fact cannot grow into a broad one when it is repeated aloud.

Battle card and one-pager questions

Can ChatGPT research competitors for the battle card?

It can search the web and summarise public pages, but treat each claim as unverified until you have opened the source yourself. Give it the competitor's own pricing and service pages, paste the text in, and ask it to quote only from what you supplied. Anything it adds from memory goes on a check list, not onto the card.

How often should a battle card be refreshed?

Review each card every quarter, and sooner when a competitor changes pricing, launches a service, or a prospect mentions something you have not heard. Put the review date and the source for every claim on the card so a stale line is obvious to the next person who reads it.

Is it safe to paste client details into ChatGPT for these documents?

Leave client names and confidential figures out. A battle card and a service sheet need only your own offer, public competitor information and anonymised customer outcomes. If you do want to use real client material, use a business plan whose settings stop training on your content, and get the client's agreement first.

Should the sales team see the whole battle card?

Give reps the one-page version: a short positioning line, three strengths, three weak points to be honest about, objection replies and the proof points. Keep the research notes and source list in a separate file that the owner or marketing lead maintains.

Further reads

• Why Is AI Bad at Maths? What to Check in Quotes and Invoices — Why figures in a draft need checking line by line.
• How Accountants Can Use ChatGPT Effectively: Review Notes and Memos — Review-note habits that suit any professional firm.
• How to Use ChatGPT for Market Research on Your Local Customers — Find out what buyers say before writing objection replies.
• ChatGPT Prompts for Tax Preparers: Client Emails and Checklists — More prompt templates for a professional-services firm.
• Best AI Tools for B2B Marketing at a Small Manufacturer or Distributor — Ten AI tools matched to industrial marketing jobs: case studies from job notes, datasheet copy, trade emails and follow-ups, with prices and examples.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: OpenAI business pricing page; OpenAI help centre articles on Projects, data controls and the custom GPT retirement and migration FAQ.