Structured work in seconds
Extract requested items, match against the merchant's catalogue, identify missing details and prepare a response or quotation draft using business facts.
ZidiSales · public direction
Make every customer conversation useful sales work.
ZidiSales is built for Kenyan catalogue-led merchants handling real customer enquiries on WhatsApp. Its role is not to hand business authority to AI; it is to turn unstructured messages into reviewable, traceable operational work that staff can approve.
Status: The product direction keeps a human approval boundary for consequential actions. AI may interpret messages and prepare drafts; deterministic business rules and people remain responsible for stock, payment, pricing and authority decisions.
Conversation → catalogue truth → review → action. The long-term value is not a generic chatbot. It is a safer operations layer where every suggested reply can be checked against the merchant's real business data.
The current problem
WhatsApp enquiries are valuable, but manual handling is fragile.
Merchants often have staff reading informal messages, checking stock and prices, preparing quotations, replying, recording orders and trying not to lose follow-ups. ZidiSales is designed to make that work structured and reviewable.
Human approval stays in control
During the pilot and beyond, outgoing replies and consequential decisions can stay reviewable. The system should route uncertainty to people rather than inventing confidence.
Not just conversation
Catalogue, price and inventory data are the source of truth. ZidiSales is intended to produce operational records and a traceable queue, not merely persuasive text.
Three-stage path
Earn the right to become a business operations layer.
Configure a few businesses deeply
Set up catalogue-aware workflows for real merchants, charge for implementation and support, and learn which parts of the workflow repeat. Measure response time, follow-up, conversion and staff correction—not vanity engagement.
Turn repeatable patterns into product
Where evidence supports it, make catalogue import, business rules, approval queues, reporting and connectors easier to operate. Each capability still needs tenant isolation, auditability and graceful failure behaviour.
Connect trusted business systems
Explore controlled integrations with WhatsApp, inventory, M-Pesa, accounting and dispatch. The target is a vertical operations layer, not autonomous commerce: payment, discount, stock mutation and authorization never become uncontrolled AI decisions.
Portfolio connection, not premature merging
Stok may later supply inventory intelligence—but only if it helps merchants.
There is a credible future where Stok's inventory information makes ZidiSales' sales workflow more useful. The products will remain independently valuable until customer evidence proves that an integration is worth the additional complexity.
Merchant-first
Merchants are the businesses using ZidiSales; customers are the people who message, enquire with or buy from those merchants. The vocabulary and data model keep that relationship clear.
Evidence beats attachment
If businesses receive too few enquiries, catalogues are unreliable, staff reject review or the product does not affect cost or revenue, the direction should change.
Privacy and control
Business data, customer conversations, roles and approvals require explicit tenant boundaries and careful retention rules before expansion.