Run competitive intelligence alone with a closed list, a repeatable note and a named reader

This lesson covers the operating design. The mechanics, from finding candidates to checking a claim, each have their own lesson, linked where they come up. If you have no Project yet, start with the Project-setup lesson.

Write the standing questions before choosing who to watch

The program exists to answer a few questions repeatedly, and the questions decide who belongs on the list. For this example, write three to five in plain language, each tied to a decision someone in your company makes: "Which competitor is winning the deals we lose, and on what?", "Has anyone changed pricing or packaging that our sellers will hear about?", "Which of our approved claims are still true this week?".

A question no decision depends on stays out of the program; answer it in a chat when someone asks. The test is whether you can name the person whose next action changes when the answer changes.

Put the questions in your Project instructions. Every weekly session then starts from the same list, and Claude compares new material against the questions instead of summarizing whatever you paste in.

Start with an illustrative list of three to six competitors

For an initial setup, use three to six competitors and a weekly hour as illustrative bounds, not measured capacity limits. Choose the list you can check in the time available, and reduce it if source reviews are repeatedly missed. The competitor-set lesson covers finding and ranking candidates; here the decision is which of them earn a standing slot.

The questions decide. A competitor that appears in your lost deals belongs on the list even if it publishes little; a loud competitor your buyers never mention does not. If your set includes a company selling in Japan, the Japan-sources lesson explains where its material appears.

Close the list. Between reviews, no one is added. A new name goes into a holding note with the reason someone raised it, and the next scheduled review decides whether it replaces an existing slot. An open list grows every week and stops being a program.

Start with a weekly cadence, then adjust it to the work

The cadence example here is weekly, on a fixed day, for about an hour. A 2022 Product Marketing Alliance survey of product-marketing leaders reports that they analyzed target and adjacent markets weekly (21.5%), monthly (23.7%), quarterly (39.8%), annually (10.8%) or on another cadence (4.2%). That is a broader question than competitor analysis alone and describes survey responses, not a target for this method. A longer interval may create a larger batch for each review.

If repeated reviews produce little worth writing down, lengthen the interval instead of skipping checks without a record. Use a cadence you can keep so the dated notes build history to compare against.

On a paid plan, Claude Cowork's scheduled tasks can run the recurring pull: a prompt runs hourly, daily, weekly or on weekdays. Anthropic lists the feature for Pro, Max, Team and Enterprise. Cloud runs use connectors and files saved to the Claude account; a task that needs local files or apps runs locally and needs the desktop connection available. Free accounts do the pull by hand.

Keep the whole setup in one Project

Anthropic's Projects article confirms Projects exist on every plan, with free accounts limited to five, and that instructions and knowledge files are separate: instructions set how Claude responds, files hold the material.

Add one block to the Project instructions from the setup lesson, so a weekly session cannot drift from the closed list and the standing questions.

Prompt: competitive-intelligence block for Project instructions

Competitive-intelligence program for [product] in [market].
Tracked competitors (closed list, reviewed on [cadence]): [name 1], [name 2], [name 3].
Standing questions, in priority order:
1. [question tied to a decision]
2. [question tied to a decision]
3. [question tied to a decision]
At each [cadence] review I paste what changed. Compare it against the prior dated note in the
knowledge files and the questions above. Report only changes that bear on a
question, with the source URL or screenshot date for each. List anything
that changed but fits no question separately, under "outside scope".
Do not add competitors or questions; flag candidates for the next scheduled review.

Keep the dated notes as Project files so the comparison uses the exact record.

Define the weekly output, then run the hour the same way each time

The example output is one short note, organized by the standing questions. A question with no change that week says "no change" so the reader knows you looked. Each item names the competitor, what changed, the date, and a URL or screenshot the reader could reopen.

1. Gather what changed

   Open your saved sources for each competitor and collect this week's differences. The page-changes lesson covers dated snapshots; the alerts lesson covers filtering so only material items reach this step. Check: every item has a date and a source you saved, not a search snippet.

2. Compare against the questions

   Paste the items into the Project chat and ask for the comparison the instructions describe. Check: every reported change maps to a numbered question, and the "outside scope" list is short. A long one means the questions or the list need revising.

3. Check anything you will repeat

   Open the source for each item that names a price, a claim or a launch before it goes in the note. The verification lesson covers what to check. Check: you could hand the note to a skeptical seller with the sources attached.

4. Write and send the note

   Edit Claude's draft into the note format, save the current dated note alongside the earlier notes, and send it to the named reader on the agreed day. The digest lesson covers writing a note a reader acts on. Check: in a timed test your team chooses, the reader can locate each answer quickly.

Build the review loop

A solo program needs an explicit review loop to remain useful. This illustrative template adds two parts: a weekly reader and a quarterly revisit.

The weekly reader is one named person who makes a decision the note can change, such as a sales lead, a sales-enablement manager or whoever owns positioning. Agree on which day they read it and what they expect to find. Ask which items they used; if repeated reviews show none were useful, revisit the questions before changing the reader.

The template's quarterly revisit reopens the Project instructions themselves. Book it in the calendar when you write the first version so the next review is not left to memory.

• Each standing question still names a decision and the person who makes it; retire any that do not.
• Review whether each competitor appeared in a deal, a buyer conversation or a material change; move the rest to the holding note, and decide which holding-note names, if any, should replace a slot.
• Count the weeks skipped and decide whether the fix is scope, day or effort.
• The reader used at least one item; if not, change the questions or the reader first.
• The competitor files in the Project carry current sources and dates.

Write the revisit's decisions at the top of the Project instructions with the date, so the next review starts from that record.

Where this breaks for one person, and where Steve fits

Steve researches competitor websites, news, reviews and supported social sources, sends competitor news summaries to the channels you choose every morning, and lets you add up to 25 custom news sources per project, refreshed daily. You still choose the final competitor set and the decision the digest should support, and a subscription adds cost and setup. Confirm which monitoring and digest capabilities are enabled for your account when evaluating Steve's competitive-intelligence product.