You can’t afford a competitive intelligence analyst. The good ones cost $80,000 a year, and the enterprise tools — Crayon, Kompyte, Klue — start at $2,000 a month. So you do what most solo consultants and small-team operators do: check competitor sites every few weeks and find out about a major pricing change three months after it shipped.
The advice you’ve found so far hasn’t helped. It’s all “12 AI tools for competitive analysis” listicles that compare features but never tell you what to actually do on Monday morning. Here’s a workflow for monitoring competitors with Claude and Perplexity that takes 15 minutes a week, costs about $30 a month, and surfaces what changed before your next client call.
Why Most AI Competitive Analysis Advice Falls Apart in Practice
Three failure modes show up over and over.
First, the listicles. They name twelve tools, score them on features, and call it a guide. You finish the article knowing more brand names but no closer to a system. There’s no workflow because the writer never built one.
Second, the Claude-only workflows. These have a real synthesis layer — paste in some competitor pages, ask Claude what changed, get a thoughtful answer. The problem is Claude doesn’t know what your competitor shipped yesterday. It’s working from your manual research, which means the system is only as fresh as your last copy-paste session.
Third, the Perplexity-only workflows. These solve the freshness problem — Perplexity’s real-time search will tell you about a pricing page update that happened on Tuesday. But raw signal isn’t the deliverable. You don’t need a list of what changed; you need a decision about what to do with it.
The missing piece in all three is workflow design: what gets monitored, what triggers analysis, what becomes an action. AI doesn’t replace a competitive intelligence analyst — it removes the grunt work of scanning, summarizing, and pattern-matching, so the only thing left is the human judgment part. The trick is splitting the job between the tool that handles fresh data well and the tool that handles synthesis well. That split is what the rest of this is about.
The Two-Tool Split: Perplexity for Discovery, Claude for Synthesis
Perplexity is the eyes. Its job is to answer “what changed in the last seven days?” across competitor sites, pricing pages, blog posts, changelogs, LinkedIn announcements, Reddit threads, and review sites. It does real-time web search and returns sourced, structured prose you can pipe straight into another tool.
The reason to use Perplexity here instead of Google Alerts or a scraper: Alerts are reactive (something has to be indexed and tagged your way), and scrapers break the moment a competitor redesigns. Perplexity handles open-ended queries — “what new features did X ship this month” — and cites sources you can click to verify. That last part matters more than you’d think; we’ll come back to it. See Perplexity and other research tools for a detailed breakdown of the options.
Claude is the brain. Its job is to take Perplexity’s findings and answer the questions that actually drive decisions. Is this pricing experiment a permanent change or a test? Does this new feature affect our positioning, or is it adjacent? Should we respond, watch, or ignore?
Why Claude over ChatGPT for the synthesis layer: the long context window. You can paste a month of Perplexity findings, your own positioning doc, and your last quarter’s competitor notes, and ask comparative questions in one shot. ChatGPT will do this too, but for prose-heavy strategic work Claude’s advantages tilt the choice.
Cost reality: Perplexity Pro at $20 a month plus Claude Pro at $20 a month covers most solo and small-team needs. If you’d rather pay per use, the API route runs $0.10–$2 per analysis depending on how much context you load — and prompt caching brings that down further if you reuse the same competitor doc each week.
That’s the split. The next question is what you actually type into each tool every Monday.
The Weekly Workflow (With the Prompts You Actually Use)
The setup happens once and takes about ten minutes.
List 3–5 competitors. For each, write down the URLs you want watched: pricing page, changelog, blog, careers page. Then write one paragraph describing your own positioning — what you do, who you do it for, what makes you different. Save the whole thing as competitors.md. Every weekly run will reference this file.
Step 1 — Discovery (5 minutes, Perplexity). One prompt per competitor:
“In the last 7 days, what changes have happened at [competitor] across pricing, product features, hiring, leadership, or public commentary? Cite sources and dates.”
Run it for each competitor in sequence. Save the outputs as plain text — you’ll paste them into Claude next.
Step 2 — Synthesis (5 minutes, Claude). Paste all Perplexity outputs in, then your competitors.md file. Then this prompt:
“Here are this week’s competitor signals and our positioning. Group findings into three buckets: noise (ignore), watch (note for next month), and act (decide this week). For each act item, propose one option and one risk.”
The bucketing matters. Without it, every signal feels equally important and you end up paralyzed. With it, most weeks produce one or two act items and a small list of watch items. That’s the deliverable.
Step 3 — Action capture (5 minutes, human). Write the act items into your weekly review doc with an owner and a date. This is the one step AI doesn’t do, and it shouldn’t. Deciding what to do — and committing to it on a calendar — is where competitive intelligence actually pays off.
Schedule it: 15 minutes every Monday at 9am. Treat it like standup. It works because it’s boring and consistent. If you’ve already built other AI automations into your weekly rhythm, this one slots in next to them.
That’s the whole workflow. Which means now is the time to talk about where it breaks.
Where This Workflow Breaks Down (and What to Do About It)
Hallucinated changes. Perplexity will occasionally summarize a 2023 announcement as if it shipped last week. The cited source link is the safety net. Always click through before classifying anything as “act.” Treat any uncited claim as suspect.
Strategy is invisible. AI can see that a competitor raised prices, hired three engineers, and posted on LinkedIn. It cannot tell you why. The “why” still requires you talking to your customers and reading between the lines of how those moves connect. The workflow gives you the dots. You’re still drawing the line.
Signal vs. noise drift. After a few weeks, Claude’s “noise” bucket starts repeating itself because the same low-stakes items keep recurring. Once a month, update competitors.md with what you’ve decided to permanently ignore. Future runs stop re-surfacing it.
Private moves stay private. A competitor in stealth on a new product line, or in private conversations with your shared customers, won’t show up. AI monitoring covers public surface area only. Your customer conversations are the other half of intelligence — keep doing those.
Set It Up Monday Morning
You can’t afford an analyst. You can’t afford to miss what competitors are doing. You don’t have to choose anymore.
Three steps to start. Tonight: write competitors.md — names, URLs, your one-paragraph positioning. Ten minutes. Monday at 9am: run the discovery prompt in Perplexity, one competitor at a time. Then paste everything into Claude with the synthesis prompt.
After four weeks, you’ll know more about your market than most full-time CI analysts at companies your size — in roughly an hour of work per month. The win isn’t replacing a hire. It’s getting analyst-grade signal as a side effect of a 15-minute weekly habit.