SalesMate AI

The AI sales assistant built for every small business owner selling on WhatsApp and Instagram across Africa.

Comment 1

Inspiration

Walking through any market in Nigeria, you see the same story. everywhere. Hardworking business owners selling quality products on WhatsApp and Instagram — but losing customers simply because They don't know what to say. A customer asks about price, the The seller panics and drops it. A customer goes silent, the seller gives up. A competitor offers ₦5,000 less; the sale is lost.

These aren't bad business owners. They just don't have a sale. team. No training. No script. No strategy.

That's what inspired SalesMate AI. I wanted to build something. that puts a professional sales advisor in every small business owner's pocket — for free.

What it does

SalesMate AI is an intelligent sales assistant built specifically for small businesses selling on WhatsApp and Instagram.

Core features:

  • Intent Detection — reads every customer message and automatically classifies it: Price Objection, Buying Intent, Delivery Inquiry, Product Question, Complaint, or General Inquiry

  • 3 Reply Variations — generates three tailored reply options based on detected intent, customer type, and chosen tone

  • Multi-language Support — works in English, Nigerian Pidgin, Hausa, Yoruba, Igbo, and French

  • Objection Handling — converts hesitation like "It's too expensive" or "let me think about it" into persuasive, value-focused responses

  • Competitor Price Analyzer — helps sellers defend their price confidently without panicking or dropping it unnecessarily

  • Sales Insight Card — after every conversation, generates a full intelligence report including Buy Probability, Customer Sentiment, Urgency Score, Seller Mistake Warning, and a Single Best Action

  • Follow-up Generator — creates a 3-stage follow-up sequence (Day 1, Day 3, Day 7) for when customers go silent

  • Full Conversation Context—paste an entire chat history. and the AI understands the full story before replying

  • Customer Memory — personalizes replies for returning and VIP customers based on past interaction history

  • Sales Script Generator — produces a complete sales playbook for any product: Opening, Interest Builder, Objection Handler, Closing Line, and Exit Message

  • Quick Content Generation — instantly creates Instagram captions, product descriptions, WhatsApp messages, and limited-time offers

How we built it

SalesMate AI was built entirely using MeDo with a progressive, layered architecture:

Layer 1 — Context gathering The user inputs product details, price, customer type (New, Returning, VIP), tone (Polite, Persuasive, Urgent), and preferred language.

Layer 2 — Intent detection Before any reply is generated, a dedicated AI step reads the customer message and returns a single intent label. This label feeds directly into every subsequent step.

Layer 3 — Reply generation The reply generator receives product context, customer type, tone, language, and detected intent simultaneously — producing three tailored, context-aware reply options.

Layer 4 — Sales intelligence A separate AI step analyzes the full conversation and produces the Sales Insight Card — including the Seller Mistake Warning, which flags errors the seller may be making in real time.

Layer 5 — Follow-up and scripts Independent prompt steps handle follow-up sequences and sales script generation, each receiving the full product and conversation context.

Every feature was built progressively, tested with real Nigerian market scenarios, and refined before the next feature was added.

Challenges we ran into

  1. Prompt chaining consistency Getting each AI layer to receive the right variables and produce Consistently structured outputs took significant iteration. Early versions would sometimes ignore the detected intent or produce replies that contradicted the chosen tone.

  2. Multi-language quality Generating natural Nigerian Pidgin that sounds human — not like translated English — required careful prompt engineering. Generic translation instructions produced robotic results. Specifying "natural Nigerian Pidgin as spoken by a market seller" dramatically improved output quality.

  3. Balancing features with credits Building progressively with limited credits meant every prompt had to work correctly on the first or second attempt. There was no room for waste. This forced clarity and precision in how every instruction was written.

  4. Design consistency Ensuring a clean, minimal design across all components — without hardcoded colors or inconsistent spacing — required a full design audit and systematic refactoring of every component.

Accomplishments that we're proud of

  • Built a fully functional, multi-layered AI sales tool using MeDo alone—no external code, no backend

  • The Seller Mistake Warning feature — no other sales tool we know of tells the seller what they are doing wrong. in real time during a live negotiation

  • Multi-language support covering 6 languages, including Nigerian Pidgin—making the app genuinely usable across West Africa

  • A clean, minimal design that feels professional and is easy to use for sellers with no technical background

  • Building the entire app progressively and finishing ahead of the submission deadline

What we learned

  • Intent detection changes everything. The moment the app started labeling customer messages before replying, it stopped feeling like a text generator and started feeling like a genuine sales advisor.

  • Prompt specificity is everything in MeDo. Vague Instructions produce vague outputs. The more specific and structured the prompt, the more reliable and impressive the result.

  • Less is more in UI design. Removing decorative colors, Reducing font sizes, and standardizing spacing made the The app look significantly more professional—without adding a single new feature.

  • African context matters. Building for Nigerian Pidgin, pricing in Naira, and testing with relatable products like Ankara bags and handmade sneakers made the app feel built. for its users—not just adapted for them.

What's next for SalesMate AI

  • WhatsApp Business API Integration — automatically suggest replies directly inside WhatsApp without copy-pasting

  • Voice Input — sellers speak the customer message. AI transcribes and responds

  • Sales Analytics Dashboard — track conversion rates, common objections, and top-performing reply types over time

  • Team Mode — allow small businesses with multiple staff to use the same product library and reply templates

  • Expanded language support — Twi, Swahili, Amharic, and Zulu to cover East and Southern Africa

SalesMate AI was built for the millions of small business owners. across Africa selling on WhatsApp and Instagram every day — with no training, no team, and no time.

We give them the words. They close the deals.

Built With

  • medo
Share this project:

Updates