Most ecommerce teams run the same ten recurring jobs every week. Most teams also still run them by hand: export, filter, assemble, format, circulate, chase.
An AI colleague — an agentic teammate with its own role, schedule, and deliverables — exists precisely for this layer of work. Here are the ten jobs we see most often, what "done" looks like for each, and how ready-made colleague roles map to them.
1. Catalog health check
Checks: broken variants, missing SEO fields, price mismatches against the feed, products live but out of stock. Deliverable: an Excel sheet, one row per finding, source linked. Runs as: the weekly catalog check — Penny's core job. It is the single highest-value first hire because every other team consumes the file.
2. Competitor and trend watch
Checks: competitor listings and price moves in your category, new-arrival patterns, discount depth. Deliverable: a one-page PDF digest attached to the catalog file. Runs as: Penny's trend digest — same colleague, same run, so merchandising sees findings and context together.
3. Unusual-order review
Checks: odd quantities, mismatched addresses, repeat-failure patterns, orders that look like testing or fraud probes. Deliverable: an escalation email on bad days; a one-line "all quiet" on good ones. Runs as: Sterling, the operations auditor, on a daily schedule — with an on-demand "audit now" for pre-sale deep checks.
4. Promotion and pricing conflict check
Checks: overlapping promotions, stacking mistakes, sale prices that fight each other, discount codes that shouldn't combine. Deliverable: a PDF of conflicts with the exact products and prices involved. Runs as: Bell's weekly promo check — the job that saves a sale from its own discount stack.
5. Campaign calendar digest
Checks: what is scheduled next week, what changed since last week, what looks off against the plan. Deliverable: a short slide deck your Monday marketing meeting opens first. Runs as: Bell's week-ahead digest — the calendar-honesty half of her charter.
6. Support themes digest
Checks: what customers actually asked this week, in their words; which themes grew; which got resolved. Deliverable: a Word/PDF brief with each theme linked to example conversations. Runs as: Iris's weekly support digest — the input your product and CX reviews are missing.
7. Unanswered-questions watch
Checks: where shoppers or the AI couldn't answer — the knowledge-base gaps. Deliverable: the digest section your content team works from directly. Runs as: inside Iris's weekly run. This is the cheapest knowledge-base improvement program that exists: the gaps are literally listed for you.
8. Search visibility check (classic SEO)
Checks: keyword positions for the terms that matter, movement week over week. Deliverable: a ranked table in the digest. Runs as: part of Nova's toolset (GEO specialist), tied to your Semrush keyword set.
9. AI-search visibility check (GEO)
Checks: how AI engines answer "best [your category]" questions — are you cited, are you described accurately, who is cited instead. Deliverable: a monthly PDF with prompts, answers, and citation sources. Runs as: Nova's core job. Five years ago this job didn't exist; now it decides who gets recommended when the shopper never opens a search results page.
10. The weekly roll-up meeting file
Checks: nothing new — it assembles the week's findings into the one file your trade meeting actually opens. Deliverable: slides, or one master spreadsheet. Runs as: any colleague's output can feed it; the point is that nobody assembles it by hand anymore.
The pattern behind the ten
Every job on this list has the same shape: predictable checks, against known sources, ending in a file people already use. That is exactly the job description of an AI colleague — and exactly why they should run on a schedule instead of on someone's Sunday evening.
You don't need all ten on day one. Pick the one whose file your team is most tired of assembling — for most stores that's the catalog health check — hire that colleague first, review the output for a couple of weeks, then add the next. CHATTERgo ships these as ready-made role templates: meet the bench, or read how to hire your first AI teammate before you start.
Frequently asked questions
Do all ten need different AI colleagues?
No. The ten jobs group naturally into five roles: merchandising (1–2), operations (3), marketing (4–5), insights (6–7), and search visibility (8–9), with the roll-up (10) assembled from any of them. Each role runs several jobs in one scheduled pass.
What does it cost to run all five?
Colleagues run on workspace AI included with CHATTERgo plans; the constraint is conversations and AI credits, not per-report fees. Start with one role on your current plan and scale when the files prove themselves.
How do I keep the reports honest?
Findings carry source context on every line — spot-check two or three each week, and tighten the colleague's charter when a check is systematically wrong. The review loop is the management.