Rundocs › Rundoo AI

AI Agents

Last updated 2026-06-12

AI Agents are Rundoo AI workers you set up once — they run on a schedule and do the work. Summaries, draft orders, reports, counter alerts — whatever the agent is built to produce, it produces without you opening the app.

In the POS mode, open Rundoo AI at the bottom of the Analytics group — then switch to the Agents tab at the top of the page.

What an agent is

An agent is Rundoo AI set up to run on its own. You write one natural-language prompt, pick a schedule and the Staff who should get the result, and Rundoo AI runs it every day, week, or month. The output lands however the agent is built to deliver it — a text, an email, a drafted order waiting in Orders, a report sitting in your inbox.

The split is: Chat is ask-now; Agents is schedule-and-forget. Same model, same data, same plain-English prompts — just the difference between you sitting down to ask a question and the answer showing up at 6pm without you opening the app. Start in Rundoo AI to get a feel for how Rundoo AI answers; come back here when the same question is worth asking every day.

Think of an agent as your night manager. While you're locking up, it's reviewing the day and deciding what you need to know in the morning.

The three types

Every agent is one of three Types. The Type is the mental shape of what the agent produces — pick it first, then write the prompt.

Daily Summary

Inspects yesterday's activity and texts you what's worth noticing. Good for the morning-coffee read: sales totals, margin outliers, lapsed customers, inventory that slipped under the line. The flagship is the Daily AI Summary itself — a Rundoo-shaped version of "how did yesterday go?" that real owners tap into before they drive to the store.

Prompt pattern: "At 6pm each day, review yesterday's sales and flag anything unusual — low-margin transactions, customers who went quiet, inventory running low. Text me the top 3–5 issues with transaction IDs and a one-line reason each."

Order

Proposes a purchase order or similar action-oriented output. Good when a restock or reorder question recurs on a predictable cadence: a weekly reorder, monthly cycle count list, seasonal catch-up on slow-moving SKUs. The agent doesn't place the order — it drafts the list so you can review, adjust, and submit.

Prompt pattern: "Every Friday at 7am, review product levels across Menlo Park and Redwood City. Text me a proposed order with SKUs and quantities based on sales the past 4 weeks and current stock."

Report

Answers a recurring business question. Anything you'd otherwise pull a Report for every Monday morning — customers past due, aging balances, top-20 price vs competitors, negative-on-hand. One prompt; one scheduled answer.

Prompt pattern: "Every Monday at 8am, check our top 20 SKUs against nearby competitors. Text me any SKU priced more than 5% above competitors, with the gap and competitor name."

Creating an agent

From the Agents list, tap + Add agent top-right. A modal opens. Every field is hand-authored — no templates, no presets.

In the POS > Analytics > Rundoo AI tab, flip to Agents, then click + Add agent in the top-right. The list shows every agent in your store — name, type, status, scheduled run time, last run — so this is also where you come back to edit or disable one later.

Step 2: Name it and pick a Type

Fill in Name (e.g. "Daily AI Summary") and Identifier (a 2–4 letter shorthand — DAS, CCL, AA — that shows up in the list's ID column). Then pick Type — Daily Summary, Order, or Report. That's the whole menu.

The Type shapes what Rundoo AI expects to produce: a summary, a draft order, or a recurring-question answer. Pick the one closest to what you'd ask for out loud.

Step 3: Assign Staff and set the Schedule

Add Staff in the Staff field — each assigned person gets the agent's output on their phone. Then type the cadence into Schedule in plain English: Every morning at 7am, Fridays at noon, Every day at 8pm Eastern Time.

💡

Whoever's in the Staff field gets the output — and only them. Each assigned person receives it on whichever channels they've turned on under their Notifications tab (App, Email, or SMS); out of the box that's a text. To get an agent's report by email, the assigned staff member checks Agent run → Email on their own Notifications tab (see Staff), with email consent on. Reports can only go to assigned staff members' own email or phone — there's no way to send an agent to an outside address like a customer's inbox.

If two people are assigned, they both get it — useful for owner + manager coverage, or for making sure nothing depends on one phone being on.

💡

Rundoo AI reads your schedule as plain English. Type it how you'd say it — Every morning at 7am, Weekdays at 8pm, Fridays at noon. Specific beats vague.

Step 4: Write the User Prompt and create

The User Prompt is the agent's whole instruction. Tell it what to look at, what to flag, and how to format the output. Keep it concrete — date ranges, thresholds, what "unusual" means to you. End with "…text it to my phone" so the delivery is explicit.

When every required field is filled, the Create button in the bottom-right turns blue. Click it. The new agent lands at the bottom of the Agents list with Status = Active and the next Scheduled Run Time already computed from your plain-English schedule.

Example agents to try

Three worked examples. Copy the prompt into a new agent, tweak the thresholds to match your store, and you're running.

💡

The text arrives short on purpose. Tap the link in the message to jump into the app and see the full context — the transaction, the customer, the report the number came from.

Daily AI Summary

The flagship — a daily text at 6pm with yesterday's sales plus anything worth looking at. Best shown with a real story: Jason, a paint-store owner, got his Daily AI Summary one evening. It flagged sale CUN6A-1213LPJE-S, ULTRA SPEC 500 EGG -BASE 3 QT, priced at or near cost. He tapped in, pulled the customer's history, and found a 10-week pattern — the same SKU sold at cost over and over. A clerk had entered a one-time competitive price months back; the old POS saved it as permanent. Undetected margin loss: $4,200+. Jason called the customer, repriced the line, retained the account.

~10 minutes a day with the summary caught what 5–8 hours a week of report-diving hadn't.

Type: Daily Summary
Schedule: Every day at 6pm
Prompt: "At 6pm each day, review yesterday's sales across all locations and flag anything unusual — low-margin or below-cost transactions, customers whose order cadence broke, SKUs running low on hand. Text me the top 3–5 items with the transaction ID and a one-line reason each."

Weekly reorder

Classic counter use case. Instead of pulling a Buyer's List by hand every Friday, the agent proposes the order and delivers it — you review, adjust quantities, and submit in Orders.

Type: Order
Schedule: Every Friday at 7am
Prompt: "Every Friday at 7am, review product levels across Menlo Park and Redwood City. Text me a proposed order with SKUs and quantities based on sales the past 4 weeks and current on-hand stock. Prioritize top-selling SKUs."

Competitor pricing watch

A Report-type agent for price discipline. Runs every Monday and pings you only when something's out of line — no noise on weeks where nothing moved.

Type: Report
Schedule: Every Monday at 8am
Prompt: "Every Monday at 8am, check our top 20 SKUs by revenue against nearby competitor prices. Text me any SKU priced more than 5% above competitors, with the price gap and competitor name. Skip any SKU where we're within 5% or where we're already lower."

💡

An agent's output is AI-generated from your live data. Before acting on a dollar figure — sending a PO to a vendor, crediting a customer, repricing a line — open the matching Report or detail page the text references and verify. Rundoo AI can still miscount, misread a date range, or round the wrong way.

Running, editing, and deactivating

Click any row in the Agents list to open its detail page. Same fields as the modal, with two labels relabeled: Type becomes Category, Schedule becomes Scheduled run time. The form is an Agent settings block — edit any field, then Save changes at the bottom.

Two buttons live in the top-right:

Run now — fires the agent immediately, regardless of schedule. Good for testing a prompt rewrite or grabbing a mid-week pulse without waiting for the next run.
Deactivate — flips Status to Inactive. The agent stays in the list (so you don't lose the prompt) but stops running on its schedule. Toggle back to Activate when you want it running again.

The list's Last Run column is the only run history Rundoo surfaces — the timestamp of the most recent execution. For an agent you want to audit deeper, the delivered text on the assigned Staff's phone is the record.

Recommended Rundocs

Rundoo AI — the ask-a-question Chat mode.
Common AI workflows — a menu of prompts and agent setups by business area.
Running Reports — the report library agents reference.
Staff — add phone numbers so assigned Staff actually get agent texts.