Swedana Monolog Watermark
Autonomous Agents/8 Weeks Sprint/Deep AI Build
Uneora v2Autonomous B2B Commerce Operating System inside Telegram

Uneora v2Autonomous B2B Commerce Operating System inside Telegram: "An autonomous procurement agent operating in Telegram that cuts wholesale order processing from 48 hours to 8 minutes."

02 / THE PROBLEM

Multichannel Order Chaos & Manual ERP Translation

Traditional wholesale distributors lose 15–20 hours weekly manually translating handwritten WhatsApp notes, voice messages, photos of paper invoices, and messy PDFs into desktop ERP systems. Standard automation fails because LLMs risk hallucinating monetary purchase orders without strict cryptographic approval gates.

KEY PAIN POINTS BEFORE THIS EXISTED
Unstructured multi-format order noise (handwritten notes, voice pings, messy PDFs)
Severe business risk of unauthorized or hallucinated financial purchase orders
Legacy desktop ERPs disconnected from modern daily communication tools
03 / THE APPROACH

Native Chat Agent with Cryptographic Human Approval Gates

We built an autonomous AI agent that operates natively inside Telegram—where distributors already spend their day—protected by two-tier cryptographic approval gates before any order reaches the database.

DECISION 01

Telegram-First Interface

Eliminated the need for field and warehouse staff to learn complex new software; everything happens inside familiar chat dialogues.

DECISION 02

Cryptographic HMAC Guardrails

Every financial order trigger requires a HMAC-SHA256 signature payload from an authorized human supervisor before database commit.

DECISION 03

Noisy-Data Pre-Filtering

Built a cleaning layer that normalizes supplier product shorthand and currency ambiguities before passing to AI parsing.

04 / WHAT IT DOES

Automates Wholesale Procurement from Voice/Image to ERP Ledger

Turns unstructured multi-modal messages into audited wholesale purchase orders.

01

Multi-Modal Ingestion in Telegram

Accepts voice notes, photos of handwritten pricelists, and informal chat text, extracting exact SKUs and quantities.

Interface Detail: Telegram chat window showing instant parsing of a handwritten invoice image into structured PO draft
02

HMAC One-Tap Approval Card

Sends structured inline approval buttons to procurement managers with cryptographic signatures.

Interface Detail: Interactive Telegram card displaying itemized summary with Approve and Reject buttons
03

Multi-Tenant ERP & Inventory Sync

Row Level Security (RLS) guarantees absolute catalog isolation while updating inventory in real time.

Interface Detail: Web administrative dashboard showing live inventory reconciliation and generated PDF purchase order
05 / OUTCOME & IMPACT

-80% Turnaround Time & Zero Cryptographic Breaches

-80%
Cycle Reduction
Order creation time reduced from 48h to under 8m
99.4%
Parsing Precision
Accuracy in extracting SKUs from noisy supplier catalogs
50+
Active MSMEs
Processing daily wholesale purchase orders in production

Uneora v2 brought modern AI automation to traditional MSME distributors without forcing them out of Telegram or compromising financial safety.

Live in Production — Processing daily active wholesale transactions.
EXPLORE MORE SYSTEM ARCHITECTURES

Related Case Studies & Teardowns

View All (8)
BESPOKE DEVELOPMENT

Ready to build a system like Uneora v2?

Swedana builds anything your business needs — from high-converting marketing sites to full AI-integrated systems in 3 to 8 week sprints with 100% IP ownership.

Calculate Project Scope
06 / TECH STACK & INFRASTRUCTURE FOOTER
Technical Details
Next.js 16PostgreSQL RLSSupabaseTelegram Bot APIHMAC-SHA256 SecurityCloudflare WorkersNext.js 16PostgreSQL RLSSupabaseTelegram Bot APIHMAC-SHA256 SecurityCloudflare WorkersNext.js 16PostgreSQL RLSSupabaseTelegram Bot APIHMAC-SHA256 SecurityCloudflare Workers