Ten agents at go-live. Twenty-five at ceiling.
Not one system with ten labels. Each agent holds a defined scope, its own tool permissions and an autonomy level that decides whether it acts or waits for a person.
| Agent | What it does | Autonomy | Where you see it |
|---|---|---|---|
| Inventory AI Stock sync, reorder points, dead-stock flags | Reconciles stock across every connected channel, prevents oversell, raises reorder points against demand signals and reports dead stock with ageing. | Supervised | Inventory ledger, channel sync log |
| Listing AI Copy, attributes, category mapping, images | Builds and repairs listings to each marketplace's field rules, completes attributes from the catalogue master, maps categories and checks image compliance. | Draft only | Listing pipeline, stage 3 human QC |
| Pricing AI Competitor scrape, repricing, margin guard | Tracks competitor prices, reprices inside the client's margin floor, loads promotions and checks parity across channels. Below-floor attempts are blocked and logged. | Supervised | Pricing intelligence, floor-breach log |
| Support AI Drafts replies across five channels | Reads buyer messages across marketplace inboxes, email, chat and the own store, retrieves order and policy context, and drafts a reply for release by your reviewer. | Draft only | Customer support, review queue |
| Refund AI Eligibility checks, claim filing, reconciliation | Checks refund eligibility against policy and order state, prepares marketplace claims, reconciles against settlement reports. Anything above the ceiling needs a named approval. | Draft only | Refunds, settlement reconciliation |
| Returns AI Pickup chasing, inspection triage | Authorises returns within policy, chases pickup, triages inspection outcomes and routes disputes to the queue. | Supervised | Returns workflow |
| Compliance AI Policy and claim screening before publish | Screens every outbound listing and message for prohibited claims, policy breaches and category restrictions before anything is published. | Draft only | Compliance screen, blocked-item log |
| Ads AI Campaign pacing, bid and placement moves | Paces campaign spend, adjusts bids and placements within budget limits, reports on spend efficiency by account. | Supervised | Advertising ops |
| Forecast AI Demand projection, replenishment planning | Projects demand by SKU and channel, produces replenishment plans and flags stock-out risk ahead of it happening. | Supervised | Inventory intelligence |
| Fraud AI Order and return abuse detection | Detects order and return abuse patterns, serial-returner behaviour and suspicious address clustering, and raises cases for human decision. | Draft only | Case management |
Specialists, added when volume justifies them.
Bought in blocks of five for ₹1,75,000 when allocated volume grows. A block adds about ₹80,000 a month after its own running cost, repaying itself in a little over two months. You never carry capacity you are not being paid to run.
Channel specialists
Per-marketplace agents tuned to a single platform's policy set, field rules and case system — used when one channel carries enough volume to justify dedicated handling.
Category specialists
Agents configured for a vertical's attribute schema and restriction list — apparel sizing, electronics compliance, beauty and personal care claim rules.
Settlement & recon
Deeper reconciliation across marketplace settlement reports, fee disputes, short-payment identification and claim recovery.
Content & enhancement
A+ content, storefront modules, enhanced brand copy and seasonal refresh cycles across a catalogue.
Quality assurance
An internal audit agent that samples the fleet's own output ahead of the weekly QC pass, catching drift before it reaches a scorecard.
Reporting & insight
Client-facing performance reporting, anomaly narration and month-end pack preparation for the accounts you hold.
What a ten-agent fleet covers
- Around 8 managed seller accounts at standard band
- 6,000 orders and 1,800 support contacts a month
- 800 catalogue actions and 500 returns a month
- Continuous running, including nights and weekends
- One part-time reviewer clearing the queue each working day
At the twenty-five agent ceiling
- Roughly 3.2× the standard-band volume
- Gross payout around ₹4,24,000 a month
- Running cost around ₹1,23,000 a month
- Income after costs around ₹3,01,000 a month
- One full-time reviewer required from agent 16 onward
- Upgraded workspace VPS and a larger AI consumption balance
Every agent, running, on a live workspace.
The fleet page tells you what each agent does. The demo shows you it doing it, including what happens when one hits a hard limit and refuses.