The problem

An Instagram seller tracks every message in one ordinary inbox where routine questions mix with genuinely risky ones (cancellations, returns, complaints, delivery or sizing issues). Nothing is ranked by severity, and there is no system tracking whether a problem was actually resolved or the order was lost, so recoverable orders quietly disappear in the noise.

The solution

DealSave reads every incoming Instagram message and analyzes it with AI (customer intent, risk level, recoverability, a recommended action), then surfaces it in a Priority Inbox ordered worst-first, automatically linked to the matching order when one exists.

Every risky message becomes a «rescue case» that moves through explicit stages: assign an employee, contact the customer, wait for a reply, stock or delivery, then close as saved or lost, with every transition logged to an audit trail. The AI only drafts a suggested, editable reply; an employee always reviews and sends it. The system never messages a customer on its own.

Key features

• A Priority Inbox ranking every open conversation by risk, with filters (assignee, priority, status, category, date).
• An analysis panel inside each conversation: customer intent, why it is risky, recoverability, a recommended action, and an editable suggested reply.
• A full rescue-case lifecycle (assign, contact, wait, resolve as saved or lost) with an audit trail for every transition.
• Conversations linked to orders and to a customer's history (past orders, returns, cancellations).
• A dashboard and analytics: current critical cases, saved-vs-lost rate, value recovered and response time.
• A team performance page (individual performance only, no competitive ranking).
• Automatic reminders and escalation for aging critical cases.
• A separate «Sales Opportunities» section that surfaces high-purchase-likelihood conversations, with its own lifecycle and assignment.
• A bilingual interface (Iraqi Arabic by default, English toggle) with full right-to-left support.

Roles and permissions

The system is multi-tenant: each business's data is fully isolated from every other's at the database level (Row Level Security), not just in the UI. Every user belongs to exactly one organization and has a role that governs their permissions, with a Team and Roles page in Settings.

What was built

The project has not yet been launched commercially to a paying customer. Every feature runs on real data in a real database rather than static mockups, and one real business Instagram account has been connected and receives real messages.

The first business to see a live demo was a pet-supplies store, whose owner asked for a feature (surfacing conversations with a high likelihood of purchase) that became part of the product.

Technologies

Current status

In active development, originally built as a demo-ready version.

• The app is still in Meta's development mode, so message delivery works for accounts explicitly added as testers until the app review and business verification are completed. This is the next milestone.
• Without an AI service key it shows an explicit notice that analysis is unavailable instead of fabricating a result.

System screenshots

Screenshots of the running system are being prepared.