Routines: SEO, Price Checks, And Competitor Scans

What this part is for

Routines are how a shop stops depending on one long chat at midnight. You set a job to run at a time you choose, it checks the same things every time, and it reports in a format you can actually use. This guide is the "set it to run every week" chapter Steve asked for, built from jobs this team already runs — not from a generic list of ideas.

Each example below has a schedule, what it checks, what it reports, and a sample prompt you can copy. Times are America/Los_Angeles (Pacific). The public guide teaches the pattern. Private numbers stay in the private report.

One honest warning up front: some Monday 9:00 AM jobs currently overlap. A later cleanup can stagger them by 15–30 minutes. That stagger is proposed, not applied, until you approve the exact schedule.

Why routines beat one-off chats

A one-off chat is fine for a new problem. It is a bad way to remember a $0 price, a stale ticket, or a title that stopped earning clicks. Routines exist because those problems are boring and expensive. If a check only happens when you think to ask, it will not happen on the morning you are already behind.

A good routine has four parts: a clock time, a clear owner, a read-only check (unless you later approve an edit), and a short report. If it cannot say who owns it and whether it is allowed to change anything, it is not ready to enable.

A weekday scoreboard

Here is the scoreboard this shop actually uses. Other shops can copy the shape and change the times.

  • About 6:00 AM — Localizer: public website-change scan of nearby shops (saved internally; no shop list in this guide).
  • About 6:30 AM — Designer: morning analytics and SEO check.
  • About 7:30 AM — specialist bots send Cortana their morning notes.
  • 9:00 AM — Designer: opening brief / quote-ready pull. The same morning also currently holds the weekly social plan (Monday) and other Monday work that should eventually be staggered.
  • About 9:00 AM Mondays — Localizer: nearby-competitors weekly check. Monthly, on the 1st, a scorecard of top local competitors, described in general terms.
  • 9:00 AM-ish — Cortana's team brief, built from the earlier notes (a later 9:30 AM customer-facing brief has been discussed; the working brief is the internal one).
  • 10:00 AM Mondays — Designer: Shopify versus counter-system price drift, read-only.
  • During shop hours — Cortana: supplier-order email check, alerts only.
  • End of day — Front Desk: open-ticket aging.
  • Nightly — Designer: $0 price check, alert only.
  • Mondays, Accountant — weekly books close, and a look at ad spend versus new jobs.
  • 1st and 15th — Designer: a biweekly counter-system pricing update, only after your OK.
  • R&D — a morning status, plus overnight parts-map work on its own schedule.
  • Site Reader (helper) — a proposed weekly scan for broken links, public $0 prices, and accidental vendor names in customer copy.

You do not need every row on day one. A starter set is at the end of this guide.

1) Morning analytics and SEO check (Designer), daily about 6:30 AM

Checks: What changed on the site and in search visibility — pages, titles, obvious drops or gains. This shop has already used Search Console work to retitle or tighten snippets on pages that get impressions and few clicks.

Reports: A short daily note, plus a status line in Cortana's brief. Draft titles are drafts. Nothing publishes.

Sample prompt:

Each morning, check our key pages for title/meta issues and traffic or impression drops. List only pages that need a rewrite or a fix. Draft new titles for my approval. Do not publish.

2) 9:00 AM opening brief / quote-ready pull (Designer), daily

Checks: Quote-ready data and anything the counter or the site will need at opening — unfinished website drafts, a failed price check, a preview waiting on you.

Reports: A short opening note so you are not hunting for yesterday's leftovers when the first customer is already at the door.

Sample prompt:

By 9 AM, give me an opening brief: unfinished website drafts, price checks that failed, and anything I need to approve before the shop gets busy.

3) Daily team brief (Cortana), morning

Checks: Calendar, open tickets, unpaid balances, deadlines, and each bot's morning note. The brief is currently built from 7:30 AM reports.

Reports: One brief you can read before the day starts. No customer account numbers. If two bots reported the same issue, it appears once.

Sample prompt:

Collect the bots' morning notes into one brief: appointments, aging tickets, money alerts, and site work waiting on me. No customer account numbers.

4) Nightly $0 price check (Designer), nightly

Checks: Active store products or variants whose price is $0. This is a safety net after updates. Shop QA has already caught test products left public at $0. The check should ignore a cart total that reads $0 when the cart is empty.

Reports: Alert only. The bot does not change a price.

Sample prompt:

Every night, list active products or variants priced at $0. Tell me the name and link. Do not change any price. Ignore cart totals that read $0 when empty.

5) Store vs counter price drift (Designer), weekly, Monday about 10:00 AM

Checks: Whether posted store prices still match the counter system for the same item, and whether a sync error or a large drop appeared. Read-only. Nobody toggles the sync setting unless you say so.

Reports: A drift list (item, store price, counter price). A price-drop alert is suppressed only when it exactly matches a change you already approved.

Sample prompt:

Once a week, compare store prices to our counter system for the same SKU. Report differences only. Do not edit prices and do not turn sync on or off.

6) Weekly social plan (Designer + Social), Monday about 9:00 AM

Checks: Last week's posts, unused real photos, and articles that are live and safe to link.

Reports: A week grid (morning / midday / evening) with draft angles and image suggestions. Designer starts the package. Captions stay with Social until you post. No invented photos, no phone numbers in the post, booking link only.

Sample prompt:

Make next week's social plan: three posts a day, each tied to a real page or a real shop photo. Leave captions as drafts.

7) Competitor scanning (Localizer)

This one already exists, on more than one schedule:

  • Daily, about 6:00 AM: a public website-change scan of competing shops, saved internally.
  • Monday, about 9:00 AM: a nearby-competitors weekly check.
  • Monthly, about 9:00 AM on the 1st: a scorecard of top local competitors.
  • Also, a weekly look at other shops' public social posts for content ideas, not for copying claims.

Reports: A short "what changed" list and idea notes for landing pages or FAQs. The public guide does not name competing shops or vendors. The private report can be more specific if you want it that way; the Localizer's public drafts stay general.

Sample prompt:

Once a week, review public sites and social posts from nearby repair shops. Summarize new services, new offers, or page changes in general terms. Suggest one content idea for our site. Do not copy their wording, and do not mention vendor names.

Other routines worth naming

  • Cortana: a supplier-order email check during shop hours, alerts only. No vendor names in a public article.
  • Accountant: weekly books close on Monday morning, weekly ad spend versus new jobs, a monthly "is this ad account still worth it" check, and a tax-deadline watch.
  • Front Desk: daily open-ticket aging.
  • Designer: a biweekly counter-system pricing update on the 1st and 15th, after your OK.
  • R&D: morning status, plus overnight parts-map work on a schedule.
  • Site Reader, proposed: a weekly scan for broken links, public $0 prices, and accidental vendor names in customer copy.

Each of those follows the same pattern: a time, an owner, a check, a report, and no live edit until you approve one.

Collision handling

If two useful jobs are both set for Monday at 9:00 AM, they will step on each other: slower reports, mixed brief lines, or a missed check that looked like it ran. This team already has that pile-up. The fix is a written stagger — 15 or 30 minutes apart — proposed by Bot Manager, applied only after you approve the exact times. Do not let a bot "just pick a new time" on its own, or the scoreboard will drift again.

Approval and "alert only" rules

Most of these routines are alert-only. That is not timid. It is how you keep a nightly $0 check from inventing a price, a drift report from toggling sync, and a competitor scan from publishing a copied claim. When a routine should be allowed to draft a title, a caption, or a ticket note, the draft still waits. Publishing, sending, paying, and filing stay with you.

If you want a routine to gain the right to make a change, that is a Bot Manager proposal with exact text, not a quiet setting change inside the routine itself.

A starter set you can copy

If you are setting this up for the first time, start with five routines and nothing else:

1. A morning brief (calendar, waiting tickets, money alerts, work waiting on you). 2. A nightly $0-price check on active products, alert only. 3. A weekly store-versus-counter price compare, report only. 4. A weekly social plan with real photos and live links, captions in draft. 5. A weekly competitor scan, written in general terms, one content idea, no copied wording.

Run them for two weeks. Then add SEO title checks, ticket aging, or a books close when you can actually read those reports. A routine you ignore is worse than a routine you have not created yet.

Sample prompts for the starter set

Every morning, build a brief: appointments, aging tickets, money alerts, and drafts waiting on me. No customer account numbers.

Every night, list active products priced at $0, with a link. Do not change a price.

Once a week, compare store prices to the counter system. Report differences only. Do not edit prices and do not change sync.

Once a week, draft a seven-day social plan with three slots a day, real photos, and links to live pages. Leave captions in draft. Do not invent a photo.

Once a week, summarize public changes from nearby repair shops in general terms and suggest one idea for our site. Do not name vendors and do not copy their wording.

Once those five run cleanly, the rest of this chapter is optional — useful, already proven in this shop, and still always a report first.

Back to blog