Agentic Workflows

The AI teammate that drafts the work.
You stay in the approval seat.

CrunchJunkie watches your AI visibility and marketing data, raises signals with the evidence behind them, and turns them into action plans, reports and content briefs — drafted by AI, sent only when you approve.

No black boxes and no autopilot: every signal shows its evidence and sample size, and nothing goes out to a client without a human clicking Approve.

How it works

From data to done, in four steps

One loop runs through everything: detect what changed, decide what matters, draft the response, deliver it with your sign-off.

Step 1

Detect

Signals fire when something happens — a visibility drop, a competitor overtake, a dead source. A visibility change becomes a signal only at 95% confidence or more, comparing the last 14 days with the 28 before on at least 40 answers a side; a dead source or a broken connection is a counted fact.

Step 2

Decide

No-code When → Then rules decide what happens next: pick a signal type, a minimum severity and a client — or start from a ready-made template and tweak it.

Step 3

Draft

Beyond alerts, an automation can do a first draft of the work itself: an action plan, a white-label client report or an evidence-grounded content brief, written by AI from your own data.

Step 4

Deliver

Drafts wait in your Approvals inbox. You read, edit if needed, and approve — then CrunchJunkie sends to the recipients. Dismiss, and nothing happens. Nothing sends itself.

Signals

An issues inbox, not another dashboard

Instead of you watching charts, CrunchJunkie turns what it already measures into one inbox of things worth your attention — each with a severity, a plain-language explanation and an evidence expander showing the exact numbers behind it.

Visibility drop

A client's AI visibility fell: the last 14 days are at least 15% and 3 points below the 28 before, and we're at least 95% confident it isn't sampling noise.

Competitor overtake

A tracked competitor moved ahead of your client in share of voice across the AI engines you scan.

GEO score drop

A client's GEO Audit score fell on two audits in a row, against the audit before them. It clears itself once an audit scores close to the earlier score again.

Dead or retracted source

An AI answer cites a source about your client that is gone or has been retracted — a reputation risk you'd otherwise never see.

Feed & schema gaps

A product feed or product-page schema problem is holding back a client's AI shopping visibility.

Broken connection

A data connection died, so reports would silently go blank. You hear it from CrunchJunkie before your client does.

Signals run on data you've already scanned — no extra AI calls, no extra cost. Acknowledge, snooze or resolve; a handled signal is never re-raised for the same period.

Cross-channel signals

Signals across AI visibility and your other channels

These need data a visibility-only tool doesn't have: your GA4 conversions beside AI-search visibility, or every client in your book at once. CrunchJunkie has both in one place.

Visibility → revenue

AI visibility fell and the AI-referred conversions in GA4 fell over the same days, shown with both sides' numbers and periods, so you can check whether they are linked.

Portfolio pattern

The same problem hit several clients in the same week — which usually means one shared cause worth fixing once, not client by client.

The visibility side clears the same 95% confidence bar as a visibility drop, and the conversion side needs at least 20 conversions in the 28 days before, so a small account can't trip it.

Automations

When → Then, without code

An automation is a rule: when a signal happens, do something. Email your team, post to Slack, call a webhook — or hand the work to an AI draft action.

Start from a template

Visibility guardrailCompetitor watchAI visibility → revenue guardSource healthFeed & schema healthBroken connection guardCritical-only digest

Automations fire only on signals that have already been raised, and each signal triggers a given rule at most once. Every rule shows how many times it has fired and when it last did.

AI drafts & approvals

A finished draft instead of a to-do

Three AI actions turn a signal into work you can ship. All three land in Approvals first — always.

Action plan

A short, grounded plan for the signal: a one-line summary, the likely cause, and three to five concrete next steps.

Client report

A white-label report built from the client's connected data sources and AI-visibility signals, including an executive summary — open it in the builder and edit before you approve.

Content brief

An evidence-grounded GEO content brief for the project's biggest gap prompt, built from what the winning pages actually do — never invented.

Nothing sends without a human

Every AI draft waits in the Approvals inbox until you read it and click Approve — only then does CrunchJunkie create the share link or email the recipients. Dismiss it and nothing leaves the building. That's the architecture, not a setting.

Quiet is a feature

AI answers are non-deterministic — the same question gets different answers minute to minute. So a change in AI visibility becomes a signal only when it is large enough and we're at least 95% confident it isn't sampling noise (99% for a critical one), and every signal shows the answers behind it and that confidence. A quiet inbox means nothing crossed those bars since the last scan.

Also on the platform

Drive it all from your AI assistant

The MCP connector puts your CrunchJunkie workspace inside Claude, ChatGPT or any MCP client — ask questions of your live data, and let the assistant draft reports, track new prompts or decide what waits in Approvals with the same preview-and-confirm safety.

Explore the MCP connector

On every plan that automates

Signals and the weekly digest come with the platform; Automations scale with your plan.

  • Starter includes 3 active automation rules with email actions.
  • Pro includes 10 active rules and adds Slack and webhook actions (Zapier, Make, n8n or your own endpoint).
  • Agency and Scale include unlimited active rules.
  • A weekly signals digest emails the workspace owner a low-noise roundup — no setup, and no email in weeks with nothing worth sending.
Compare plans
FAQ

Agentic Workflows, answered

No — by design, not by configuration. Notify actions (email, Slack, webhook) alert the people you chose. The AI draft actions always route to your Approvals inbox, and only your explicit approval creates the share link or sends the email. There is no autonomous-publishing mode to switch on.

No. AI answers vary run to run, so CrunchJunkie raises a visibility signal only when the last 14 days differ from the 28 before by a material amount and with at least 95% confidence that it isn't sampling noise. Each signal triggers a given rule at most once, and handled signals aren't re-raised for the same period.

They're metered like every other AI feature in CrunchJunkie: free on your own AI keys (the provider bills you directly), or counted against your monthly allowance on Managed AI. Signals themselves run on data you've already scanned and cost nothing extra.

Starter includes 3 active rules, Pro includes 10, and Agency and Scale are unlimited. Email actions work on all of those; Slack and webhook actions are available on Pro and above. The weekly signals digest is included for every workspace.

Visibility drops, competitor overtakes, GEO Audit score drops, dead or retracted cited sources, product feed and schema gaps, broken data connections and a Meta ad account that has stopped delivering — plus AI visibility falling over the same days as the AI-referred conversions in GA4, and portfolio-wide patterns across many clients.

Yes — the MCP connector exposes your workspace to Claude, ChatGPT and any MCP-compatible client, with reads by default and opt-in, preview-and-confirm write actions. An assistant can list your automations and what waits in Approvals, approve or dismiss one item at a time, and resolve or snooze a signal; every confirmed action is recorded. Automations themselves are set up in the app.

Put your AI-visibility data to work

Start a free trial, connect a client, and let the first signals arrive — with the receipts to prove them and a human in the loop by design.

Bring your own API keys, or choose Managed AI and we'll handle them for you.