#37 How to build a competitive intelligence AI system in one afternoon with Claude
12 battle cards, one AI system, prompts included, ready to copy
Hey, welcome back. I’m Laura, just a regular human and product marketer trying to make sense of the AI era without losing my mind.
Positioning, messaging, campaigns, and sales all depend on the same thing: competitive intelligence. In this article, I’ll show you how to build, with AI, the 12 battle cards that make up a competitive intelligence hub. Deep, always up to date, before your coffee goes cold.
The problem isn’t data; it’s time
68% of B2B deals already have a direct competitor on the table. Meanwhile, sales teams rate their own competitive readiness at 3.8 out of 10 (Crayon, 2025).
Read that again. Two out of three deals are already a direct comparison. And the general feeling is still “we’ll figure it out as we go.” Not the best plan when your quarter is on the line.
That improvising is expensive. Product marketing teams spend 30 to 40 hours a quarter updating battle cards that, with luck, stay useful for a few weeks. Then a competitor shows up, changes pricing, launches a new feature, or decides to redo their website on a random Tuesday. And you start over.
For a mid-size company, competing like this can mean 2 to 10 million dollars a year in lost deals. Numbers scarier than Friday’s forecast meeting.
It all comes down to the same question
A PMM does a lot of things. A lot. Some are strategy. Others you do because someone had to and you happened to be online on Slack. Karma for replying fast.
But if you ask me which four things actually move the needle, I’d say these: positioning, messaging, campaigns, and sales enablement. All four depend on the same question: who are we competing against?
Positioning is always relative. You can only explain why you’re different once you understand the alternatives your customer is looking at. Skip that step, and you’re not positioning; you’re guessing. Messaging works the same way: saying “we’re faster” doesn’t mean much, but saying “we’re faster in the exact spot where the market leader forces you into their Enterprise plan” changes the whole conversation. That kind of message comes from solid competitive research, not inspiration.
Campaigns come from the same place, since the best angles show up where a competitor is weak and you’re strong, and that map lives in the battle cards. And then there’s sales, the official recipient of all this work. We build great battle cards, upload them to Notion or Google Drive, and think “done.” Spoiler: no. They go to the folder to rest in peace, next to the 2022 onboarding deck nobody opened again.
2 more not-visible use cases (from my own experience)
The same research you do to win more deals also helps you choose a better platform or find the right partner; only the audience changes. And there’s plain curiosity too: switching industries, prepping for a job interview, or just understanding a market without losing three nights to Chrome tabs.
I think competitive intelligence is one of the highest-impact things a PMM can build. It forces you to stop staring at your own navel and look at the market instead. #ILOVEBATTLECARD
So that’s exactly what we’re going to build: a competitive intelligence hub you can put together in a few focused hours, because AI does the heavy lifting and I’m giving you the exact process and prompts, not just the idea.
Two layers and twelve cards
A competitive intelligence system has two layers, but only one is visible. At the bottom, the foundations: who each player is, what they sell, to whom, at what price, how they position themselves, what their reviews say. This is what holds up everything above it.
On top, the cards. Not one, several, each one built for a different moment.
I use Klue as my reference framework, the competitive intelligence platform that set the industry standard with its series of 12 battle cards.
The 12 cards fall into four moments. Fair warning: this next part is the least fun to read and the most important to get right. Here they are:
They set the scene. Read before you walk in.
How they sell. How they attack deals, not what they are. Which industries they target, whether they drop price fast, whether they push a free plan. You see their move before they make it.
The competitor profile. Who the rival is: size, value proposition, real strengths, real weaknesses. The context snapshot.
They prepare the conversation.
Customer discovery questions. To find out who you’re really competing against, without naming the rival. You diagnose instead of assume.
Their product, in plain terms. What the rival sells, without marketing jargon. You speak with confidence instead of getting stuck in a feature war.
Messages by audience. The technical buyer cares about architecture, the security buyer cares about where the data lives, procurement cares about predictable spend. Same fight, different voice.
Messages by industry. A media company, an e-commerce brand and a public agency buy for very different reasons.
Price versus real cost. For when procurement gets involved. What people actually pay, not what the price sheet says.
Feature comparison. Product by product, with an honest column showing who wins each row. For the technical committee.
They run live, in the moment.
Objections and responses. For when the customer is already looking at the rival and drops their line. Every objection, matched with a response.
Trap questions. You plant them, you don’t state them. The customer concludes on their own.
Fear responses. These defuse the doubt a rival plants about you. Backed by proof, not denial.
They close it out.
The pocket summary. The three key points a sales rep should remember if they only take one card with them.
The rule is simple: the deeper and more accurate your base information, the stronger your battle strategy. And you don’t need internal data to get started. With public information and AI to gather and cross-check it, you can already get far.
One family is missing: why you actually win and why you actually lose. That information can’t come from the internet. It comes from your CRM and from interviews with won and lost customers. Without that data, you don’t make it up; you leave it pending until you have the right info.
This is a complete intelligence system for your market.
The result: CDN + Security Sector
I built the first analysis for a CDN and security service: Transparent Edge competing in Europe against Cloudflare, Akamai, Fastly, Imperva, AWS CloudFront and Google Cloud CDN.
If you didn’t know this sector before, now you do, and it’s worth knowing: the lines between CDN and security are blurring fast; 82% of websites now depend on one, and every major player is competing with a different playbook.
The result has two parts. First, a competitive sector map, to understand the basics of this market at a glance, useful for all players.
Second, and specific to one client, 12 battle cards written from Transparent Edge’s point of view as “we,” ready for their team to use.
Both are available in two formats: a browsable website and a downloadable version.
I know what you’re thinking: this sounds like a lot for one afternoon. And it would be; however, I give you the process step by step and prompts for free. I hope you like this post, at least.
How to build it yourself in one afternoon
I’m not talking about asking a chatbot “make me a battle card” and pasting whatever comes out. That’s the same smoke and mirrors as always. I’m talking about a system with separate pieces and fixed rules, so quality doesn’t depend on remembering everything every single time. So here we go
Create a project in Claude, not a random chat. A space with permanent instructions and documents always available. You don’t rebuild the method every time.
Write a master playbook inside the project instructions (instructions). This is the brain. It doesn’t generate content; it orchestrates: what phase comes first, what to ask when starting, and where to pause for a human to review. Without checkpoints, one small mistake carries through everything after it.
Save everything else as project knowledge (files), each prompt as a separate document in .md format. Markdown, plain text, minimal syntax. File names that say exactly what they contain, like “competitor-profile.md”. That name is what the master playbook uses to reference each piece without confusion. The prompts are:
Competitor profile. How to research a rival: a reliable order of sources, and one rule I don’t negotiate on: anything that isn’t public gets flagged, never invented. One competitor per run, never several at once.
Our own profile. Same method, same format, so you can compare later. Prioritizes your own sales materials, always flagged as “needs validation.”
The 12-card generator. Takes the profiles and turns them into a playbook. Fixed order: set the scene, prepare, run live, close out. And one consistent voice: a rival’s weak point gets treated the same way across every card it appears in.
Win-loss (separate and optional). Only with real internal data. Never turns three isolated cases into a rule.
Two packaging prompts. One builds the single reference document. The other builds the website your team actually uses.
Style guide, used across every prompt. Short sentences, no jargon, and the rule that matters most to me: acknowledging the rival’s strengths is part of the style, not just the method.
🇪🇸 Fair warning: the prompts are in Spanish. I wrote this project in my own language first. Translator or Claude, take your pick.
Start by hand. Scale when it hurts, meaning when you’re still saying “I’ll update it next week” every week.
Phase 1 is a Claude project you set up in an afternoon, no tech skills needed. Artifacts, the Claude feature that turns a prompt into a working webpage, generates the site itself: same format, same colors, every time, no manual cleanup. Its limit is that you’re still driving, but for two or three battle cards a month, that’s plenty.
If volume grows, phase 2 is Claude Code: an agent that works directly with your files, chains the phases together, and can publish the site on GitHub Pages under your own domain. Same review checkpoints, just automated.
Don’t skip to phase 2 on day one. Phase 1 teaches you your real bottleneck first.
Don’t let it collect dust: make it circulate
A 40-page PDF sitting in a shared folder is dead. Nobody downloads a PDF in the middle of a call, and nobody hits Ctrl+F to find a chapter while the customer is already objecting. A document like that gets made once, shared once, and forgotten.
A browsable website fixes this: it opens with one click from a phone, you navigate through an index instead of reading it cover to cover, and when you update a competitor’s profile, you regenerate it, and the same link shows the new version. No old copies floating around Slack.
And publishing it costs nothing: with Claude artifacts, you generate and publish it in one click, for free, with no infrastructure needed.
Same information, two destinations. In PDF, it collects dust. On the web, it flows.
Not just for sales: for the whole company
The classic mistake is treating the battle card as sales material only. It is that, but it’s much more. The same foundation feeds the whole company if you know how to distribute it:
Positioning and strategy. The competitor profile and the gap map are the raw material for your positioning, and they help you decide who to partner with and who to differentiate from. In every quarterly review, the website is the starting point of the conversation, not an appendix. Product uses it to decide what to build where the rival is weak.
Campaigns and content. Every gap in a rival is a campaign angle waiting to happen, one that also becomes ad segments and copy variants for demand gen and paid media. Marketing pulls headlines, landing pages, and email sequences from it. The content team turns “fear responses” into articles and “trap questions” into webinars. One hub, dozens of pieces.
Product and roadmap. The feature comparison shows exactly where you’re really losing. That prioritizes the roadmap based on market reality, not gut feeling. Product management checks the website the way you’d check a map of the terrain.
Customer success and retention. The rival is also targeting your current customers. “Fear responses” and “objections” work just as well for renewals as they do for new sales. CS uses the same cards to protect accounts.
Onboarding and leadership. A new sales rep gets up to speed with the website in one afternoon instead of three weeks of shadowing. And the executive summary gives leadership a real picture of who you win and lose against, and why, turning it from sales-floor anecdote into company data.
The pattern is always the same: one foundation, many people using it. The browsable website is what makes that “many” possible. That’s the point where competitive intelligence stops being a PMM deliverable and becomes shared infrastructure.
Your turn
I’ve told you how I built it. Now it’s your turn. If you see something you’d improve, an unnecessary step, or a missing card, tell me in the comments. I’d rather hear it from you now than find out three months from now, once the system has already been quietly failing for that long.
If this was useful, share it. You don’t need a glowing “must-read” caption, just forward it to the person on your team who’s still defending the 40-page PDF.
Finally, tell me what industry, feature, or type of document you’d like to see turned into battle cards. Cybersecurity, fintech, your own niche that’s impossible to explain in one sentence, whatever it is. I’ll bring the system, you bring the industry, and together we’ll keep one more company from competing at a 3.8 out of 10.
Sources
Primary source, behind most of this section: Crayon, “The State of Competitive Intelligence” (2025 report) https://www.crayon.co/state-of-competitive-intelligence
Secondary source, for the hours and obsolescence figures: Arise GTM, “Competitive Intelligence Automation: The 2026 Playbook” https://arisegtm.com/blog/competitive-intelligence-automation-2026-playbook
Reference source, for the 12 battle card framework: Klue, “Competitive Battlecards 101” https://klue.com/blog/competitive-battlecards-101