The stack
- Tooling
- A Claude plan with Routines + your Slack workspace
- Per weekly run
- a few minutes of agent time, cents of usage
- Setup time
- an hour or two to write and tune the brief
- After setup
- runs weekly, unattended, nothing to babysit
The problem
Every competitive-intel program starts the same way. One person volunteers to keep an eye on the competition, and for three weeks they actually do. Then a real fire lands on their desk, the watching stops, and everybody forgets until a rep loses a deal to a pricing change that shipped six weeks ago and nobody flagged. The intel existed the whole time. It just never reached the person on the call.
The information was public all along. Competitors post changelogs, update pricing pages, announce funding, hire loudly on LinkedIn. Getting to it is easy. Remembering to go get it, every single week, while you are busy, is the part that quietly fails. A chore that slips is a chore that does not happen, and this chore slips first.
I have built this the hard way. n8n on a weekly cron, Perplexity wired in for the research, a Slack node to post, the whole Rube Goldberg machine. It looked great in the Loom. It ran for about a month. Then Perplexity changed something small, the run failed without a sound, and I did not notice for three weeks, because I owned it and I was busy, which is the exact problem I built it to solve. The model did its job. The glue did not, and neither did I. Call it the glue tax: every clever pipeline is a stack of connectors somebody has to keep alive, and that somebody is always slammed.
A Claude Routine collapses the machine into one agent. A scheduled cloud agent researches each competitor with web search, cuts the findings to what actually moves a deal, writes the digest, and posts it to Slack over MCP. No Perplexity plan, no n8n workflow, no me remembering. Cap it hard, force a 'so what' on every line, and reps walk into calls knowing what changed this week.
How it works
- 01 Schedule the run Claude Routineweekly cloud agent, Monday 7am
- 02 Research per competitor Claude web searchlast 7 days only, cited
- 03 Cut and rank Claudematerial moves only, by GTM impact
- 04 Add the so-what Claudeone line per item, for reps
- 05 Cap and format Claudeunder 350 words, scannable
- 06 Post the digest Slack (MCP)intel channel plus rep channel
- Keep a tiered list of the competitors you actually lose to, plus the questions you care about each week
- Stand up a Claude Routine on a weekly schedule, say Monday 7am, with one clear brief
- In the brief, Claude researches each competitor with web search, bounded hard to the last 7 days
- The same agent cuts the noise, ranks by GTM impact, and writes a 'so what' on every line
- Claude posts the capped digest to Slack over MCP, with no Slack node to wire up
- It cross-posts the rep talking points to the channel reps actually live in
- Tune the brief for a couple of weeks, then let it run without you
The playbook
Define competitors and the questions that matter
Write down the competitors you actually lose to, not the dozen names marketing likes to list on a slide. Then write the specific questions you want answered every week: pricing and packaging changes, new features, funding and headcount moves, customer wins and losses, positioning shifts, leadership changes. Specific questions are what keep the digest on GTM-relevant moves instead of every SEO blog post a competitor publishes.
Tier them. The rival you meet in every other deal should not get equal billing with the one you have seen twice all year. Tiering lets the digest lead with what matters, and lets Claude dig harder on the names at the top.
Write down your battlegrounds, the dimensions you win and lose on, because that is the yardstick Claude uses to decide what counts as material. 'We win on multi-region depth, we lose on entry price' tells the model exactly which moves to elevate and which to skip.
TipExplicitly add 'changes to their pricing page or free tier' to the watch list. Pricing moves are the intel reps need most and the easiest to miss, because competitors rarely announce a quiet pricing change but it shows up immediately in your deals as new objections.
Create the Routine and its weekly schedule
Set this up as a Claude Routine, a scheduled cloud agent that runs Claude on a cadence without you sitting there watching. Pick one fixed time, Monday 7am, and never move it. The team learns to expect the digest, and expecting it is half of whether anyone reads it.
This is where the machine goes away. The old build needed Perplexity to research, n8n to schedule and loop the competitors, and a Slack node to post. The Routine is one agent doing all of it. Every box you delete is a box that cannot break on a Tuesday while you are looking the other way.
Connect what the Routine needs up front: web search (built in) for the research, and the Slack MCP connector so Claude can post. After that, the Routine is just a well-written brief on a timer.
TipNot sure a Routine is the right home for this versus running it by hand in the Claude app? The rule of thumb: anything you want to happen on a reliable cadence without you remembering belongs in a Routine. There is a surface guide at /claude/which-claude-should-i-use/.
Write the research and digest brief
The brief is the product. It tells Claude to research each competitor with web search for the last 7 days, cut the findings to what is material, rank by GTM impact, and write a tight digest with a 'so what' on every line. One agent researches and writes, so nothing gets dropped in a hand-off between tools.
The 'last 7 days only' line does more work than it looks. Drop it and web search cheerfully hands you the competitor's biggest news from two years ago, the digest fills with stale items, and the team trains itself to scroll past. Bound the window hard, and make Claude carry every source link through so any claim stays one click from proof.
Cap the length, around 350 words, and mean it. The cap is the single thing that decides whether anyone reads this in month two. A tight digest gets read every week. A two-pager gets skimmed once and muted.
You are our GTM competitive-intelligence analyst. Run this every Monday.
Our product (one line): {{ONE_LINE}}.
Our key battlegrounds (where we win/lose): {{KEY_BATTLEGROUNDS}}.
Competitors, tiered: {{COMPETITOR_LIST}}.
1) RESEARCH. For EACH competitor, use web search to find only what happened in the LAST 7 DAYS, each item with a source link and a date:
- product launches, feature releases, or changelog updates
- pricing or packaging changes (including any free-tier change)
- funding, acquisitions, or major hires / leadership changes
- notable customer announcements, partnerships, or campaigns
- public positioning or messaging shifts
Ignore anything older than 7 days. Do not speculate. If a competitor had nothing material, note it and move on.
2) WRITE the digest:
- Include ONLY material items. Drop minor PR and routine posts. If a competitor had nothing material, OMIT them entirely.
- Rank items by impact on OUR deals: HIGH or MED.
- Per item, one line: Competitor | What happened | Why it matters to us (one-line 'so what') | Source link.
- End with 'What to tell reps this week': 1-3 punchy, specific talking points reps can use on calls.
- Keep the WHOLE digest under 350 words. Be specific. No fluff, no hedging. Banned words: unlock, leverage, supercharge, game-changer.
3) POST the digest to the #competitive-intel Slack channel, and post the 'What to tell reps this week' section to #sales.TipCap the length hard and mean it. A 350-word digest gets read every week and shapes how reps talk; a two-page one gets skimmed once and muted. If you must choose, drop the medium-impact items, not the word cap.
Let the Routine post to Slack
The Slack MCP connector is attached, so Claude posts the digest itself. No webhook, no Slack node, no formatting service in the middle. Point it at #competitive-intel and tell it to make the thing scannable: bold competitor names, the 'so what' on its own line, source links as real hyperlinks and not raw URLs.
Cross-post the 'what to tell reps' section to the sales channel, not just the intel channel. Reps live in their own channel, not in #competitive-intel, and a talking point they never see before a call may as well not exist.
Keeping research, writing, and posting inside one agent is the whole point. The source Claude found at 7am arrives in the Slack message unbroken, so checking a claim is one click, not a scavenger hunt.
Add a battlecard update loop (optional)
Extend the brief so a material move prompts Claude to draft the battlecard update too. When a competitor changes pricing or ships a feature you compete on, have it draft the revised section and post it for whoever owns positioning to approve. This keeps the collateral current instead of letting it rot between quarterly reviews, which is exactly when reps quietly stop trusting it.
Keep a human on the approval. The Routine can flag and draft all day, but a competitive response is a positioning call, and a person should own that before it becomes the official line reps repeat.
If your battlecards live in a connected doc, give the Routine that MCP connector too and have it link the update from the weekly digest, so the intel and the collateral travel together instead of drifting apart.
Tune the signal-to-noise, then run it hands-off
For the first couple of weeks, read it like a critic and tighten the brief. Too much noise, sharpen the 'material only' line and hand it examples of what to drop. Missing things, widen the research or name specific sources to watch. A digest that cries wolf in week one never earns the standing read, so spend the calibration up front.
Click the sources yourself on the high-impact items while you are tuning. Summaries occasionally misread a headline, and you want to catch that pattern before a rep repeats a wrong claim in a live deal.
Once it is tuned, it runs without you and needs only a quarterly look at the competitor list. Markets move: a name you track pivots away, a new rival turns up in three deals. Refresh the list, then leave it alone again.
Inside the prompt
The scoring prompt is short, but every line is there for a reason. Here is what each one is doing and why.
- Material only"Include ONLY material items. Drop minor PR and anything immaterial."
- The filter is the whole product. Without it you ship a news dump and the team mutes the channel by month one.
- Omit quiet competitors"If a competitor had nothing material, OMIT them entirely."
- Stops the digest padding to a fixed shape. A competitor with no news this week should simply not appear.
- Rank by impact"Rank items by impact on OUR deals: HIGH or MED."
- Forces a judgment call instead of a flat list, so the reader's eye goes to what actually changes a deal.
- One-line so-what"one line: What happened | Why it matters to us | Source link."
- The 'why it matters' is the part reps act on. Demanding it per item is what turns news into intel.
- Tell reps section"End with 'What to tell reps this week': 1-3 punchy talking points."
- Translates the analysis into call-ready language. This is the section that gets cross-posted to sales.
- Hard word cap"Keep the WHOLE digest under 350 words."
- The single line that decides whether the digest survives past month one. Drop the MED items before the cap, never the cap.
What you get
A scannable weekly Slack digest of only the material competitor moves, each with a 'so what' line and a source, plus rep talking points, posted by the Routine itself.
*Weekly Competitive Intel, Jun 10*
*HIGH IMPACT*
• *Northstar* | Launched a usage-based pricing tier starting at $0/mo | *So what:* undercuts us on entry price, expect price objections on SMB deals this week; lead with total cost at scale, not sticker price. | <link|source>
• *Northstar* | Hired a VP Sales from a logistics incumbent | *So what:* signals a deliberate push into our core vertical; flag named accounts where we're competing. | <link|source>
*MED IMPACT*
• *Apex* | Shipped an API for bulk data exports | *So what:* closes a gap we used to win on; stop leading with exports, reframe to our multi-region planning depth. | <link|source>
*What to tell reps this week:*
1. Northstar's $0 tier is bait, it caps usage hard. On any Northstar deal, get to total cost at their actual volume fast.
2. We still win vs Apex on multi-region planning depth; their new export API does not touch that.
3. If you're in a logistics deal against Northstar, expect a more aggressive sales motion now that they've hired into the vertical.
_(Competitors with no material news this week omitted.)_- competitorNorthstar
- Bolded and tiered, so the team reads the primary rival first and a fringe name does not get equal weight.
- what happenedlaunched a usage-based tier starting at $0/mo
- One line, this week only. The 'last 7 days' bound is what keeps stale news from filling the digest.
- so what for repsundercuts us on entry price, expect SMB price objections; lead with total cost at scale
- The line that earns the read. A fact with no 'so what' is trivia reps scroll past.
- source link
- Carried through from Claude's web search as a real hyperlink, so any claim is one click to verify before a rep repeats it.
- what to tell reps this weekNorthstar's $0 tier is bait, it caps usage hard; get to total cost at their volume fast
- The cross-posted section. It lands in the sales channel reps live in, not the intel channel they never open.
Pitfalls to avoid
Including everythingAn exhaustive digest is an unread digest. Force 'material only,' drop quiet competitors entirely, and hold the word cap. The cut is the product, not the coverage.
No sources carried throughIntel nobody can check gets ignored, or worse, repeated wrong in a live deal. Make Claude carry the web-search links into the digest so any line is one click from proof.
Not bounding the time windowSkip the 'last 7 days' line and web search pads the digest with big, old news to fill space, and the team learns to tune it out. Bound the window hard in the brief.
Stale competitor listCompetitors pivot and new ones appear. Left untouched, the digest faithfully tracks the wrong companies and misses the one you just started losing to. Revisit quarterly.
Treating the summary as gospelA summary can overstate a move or misread a headline. On the high-impact items, a human clicks the source before a rep acts on it. The digest is a fast pointer, not scripture.
Questions people ask
- Why a Claude Routine instead of stitching together Perplexity and n8n?
- Because every tool in the chain is one more thing that breaks on a Tuesday, and one more bill. My old build needed Perplexity to research, n8n to schedule and loop, and a Slack node to post, and it died without a sound the first time one of them changed. A Claude Routine is one scheduled agent that researches with web search, filters, writes, and posts to Slack over MCP. Fewer boxes, fewer bills, and the digest still lands every Monday without anyone remembering to run it.
- How do I keep the digest from becoming noise nobody reads?
- Two levers, both in the brief: 'material only' so routine PR and blog noise gets cut, and a hard word cap near 350 words. Drop quiet competitors entirely. Forced to choose, cut the medium-impact items, never the cap. A tight digest gets read every week; a sprawling one gets skimmed once and muted.
- Can I trust the AI summary, or do I need to verify every item?
- Have Claude carry its web-search citations into the digest as clickable links, and for the first couple of weeks click the high-impact ones yourself before a rep acts on them. Summaries occasionally misread a headline, and you want to catch that before a wrong claim shows up in a live deal.
- How often should I update the competitor list?
- Quarterly. New rivals appear and old ones fade, and a list left alone tracks the wrong companies while missing the one you just started losing to. Tier it too, so the digest leads with who you actually fight and does not hand a fringe name equal billing to your primary rival.
Related plays
- Competitor Monitoring Skill: Diff a Competitor's Public Surface Every Week →The skill the routine runs: watch a competitor's public surface and diff it weekly.
- Sales Battlecard Template, Built From Live Competitor Intel →Turn the intel into a rep-ready battlecard.
- Battlecard: Build a Rep-Ready One vs Any Competitor →The prompt version of the battlecard.