Examples

Example 1: Turning a Client-Call Summary Into Candidates

A SaaS client wrapped a quarterly review call, and the account manager wrote a short summary: the client's inbound leads tripled over ninety days, the client's head of growth said the new content engine "finally made our pipeline predictable," and a benchmark from the call put cost per lead at roughly half the category average.

Step 1: Send the source. In the team channel she types the command with the brand first and the summary after it:

/proof-extract acme-saas The client's inbound leads tripled over ninety days. Their head of growth said the new content engine "finally made our pipeline predictable." Cost per lead is roughly half the category average.

A private message says the proof points are being extracted.

Step 2: Read the summary. A summary posts in the channel: three new candidates were saved, marked unverified, and the source yielded three in total. It suggests running the scrutiny command.

Step 3: Look at the candidates. On the Proof Points screen in the app, the Needs Review count has gone up by three. Each candidate shows its type, a category, its attribution, and a confidence percentage:

• Outcome: inbound leads tripled over 90 days, in the lead generation category
• Quote: the head of growth on the predictable pipeline, with the person's role
• Stat: cost per lead at roughly half the category average

Because the source is her own work and names a client, the first two also carry a Public Version that restates the point without the client's name. Nothing is usable yet. They are candidates.

Example 2: Running Scrutiny, Then Approving

A content lead has collected a dozen candidates over a few weeks from several client calls and a case study. Before drafting season, he wants to clean up the set and decide what the team can cite.

Step 1: Run Scrutiny. On the Proof Points screen he clicks Run Scrutiny. Two candidates describe the same retention result from two different calls. Scrutiny keeps the longer one and demotes the other as a duplicate. The screen reports how many points were verified, culled, and demoted, and the counts update.

Step 2: Read the survivors. Points that passed show a Scrutiny Passed badge. A vague claim about "great engagement" with no number attached was culled, so it still shows Not Yet Scrutinized. A couple of the strongest points are starred automatically.

Step 3: Remember the gate. Passing Scrutiny ranks a point. It does not release it. None of the points can be cited yet.

Step 4: Approve and reject. In Needs Review he works through the list:

• The retention outcome: Approve. It joins the Approved Library.
• The "great engagement" claim: Reject, with the reason "no metric, not citable." The reason is saved, and later extractions are told to avoid that kind of claim.
• A strong client quote that names the client: Approve. Its Public Version sits beside it.

The Needs Review count drops, and the approved points are now available to the drafting agents.

Example 3: A Draft Cites an Approved Point

A week later, Marketing Content drafts a blog post for the same client.

Step 1: Open the draft. In the app, the draft shows a highlight listing the proof points it drew on. Each is a chip.

Step 2: Follow a chip. The content lead selects one. It opens the Proof Points screen with that point pinned at the top under "Cited In Your Draft," so the original wording, attribution, and source are right there.

Step 3: Approve with confidence. The claim in the draft matches the evidence. Only points that were approved in Needs Review can be cited, so nothing unreleased reaches a draft. She approves the draft.

Example 4: Looking Up Evidence in Slack

An account manager is on a call and needs a retention stat.

Step 1: Run the lookup.

/proof-points acme-saas category=retention

Step 2: Read the card. A card lists up to 15 approved points for the brand that passed Scrutiny, with each point's wording, category, confidence, and attribution.

Step 3: Widen it. She wants to see approved points that have not passed Scrutiny too, so she adds verified=false. Candidates still waiting in Needs Review do not appear in either list.