How to Build an AI Sales Assistant for a Nigerian Business
Sales in Nigeria happens in conversations. A customer sees an Instagram post, sends a WhatsApp message, asks the price, asks whether you deliver to Ibadan, asks for a discount, disappears for two days, and comes back at 11pm. Sales staff cannot answer every message instantly, follow up every lead consistently or remember every conversation, and the leads that get slow replies buy elsewhere.
An AI sales assistant handles the first part of that journey reliably: instant answers, consistent qualification, correct prices, automatic follow-up and a clean handover to a person when the customer is ready to buy or wants to negotiate. This guide explains how to build one for a Nigerian business, from the sales playbook to the CRM connection, with guardrails that stop it from promising what you cannot deliver. For the broader picture of sales automation, see AI Sales Automation for Nigerian Businesses; this article is about building the assistant itself.
What an AI sales assistant does
An AI sales assistant is a language-model-powered agent that runs the early and repetitive stages of a sales conversation. The difference between it and a support chatbot is intent: a sales assistant is designed to move an enquiry toward a purchase or a qualified handover, not just to answer questions.
Core functions:
| Function | What it does | Human involvement |
|---|---|---|
| Answer product questions | From your catalogue and specifications | None, unless unknown |
| Qualify | Asks budget, location, timing, use case, decision-maker | None |
| Price and quote | Gives list prices or builds a quote within rules | Approval for exceptions |
| Capture | Saves name, phone, interest and stage to the CRM | None |
| Follow up | Sends reminders and nudges on a schedule | None, unless reply needs judgement |
| Book | Schedules calls, demos, site visits or showroom visits | Confirms availability via calendar |
| Hand over | Passes warm leads with a summary to a salesperson | Salesperson takes over |
| Report | Shows enquiries, conversion by stage and lost reasons | Management reviews |
AI Agents vs Chatbots for Businessesnt that only answers and one that can act (create CRM records, book, send follow-ups). A sales assistant needs the second kind.
Step 1: write the sales playbook
Answer-ready summary: The playbook is the document that defines how the assistant sells: who your customers are, what they ask, what qualifies a lead, what you can promise, how you price, when to hand over and in what tone. It should be written by the sales lead, not the developer, and it becomes the assistant's instructions. Without it, the assistant improvises, and improvisation in sales means wrong prices and broken promises.
Contents of a good playbook:
- Customer segments and how they differ (retail buyer versus corporate, Lagos versus other states).
- Top 30 questions from WhatsApp, Instagram and calls, with approved answers.
- Qualification questions in order of priority, and what makes a lead hot, warm or cold.
- Pricing rules: list prices, what discounts (if any) the assistant may mention, when to say "a sales representative will confirm".
- Objection handling: delivery time, payment security, warranty, "is it original?", "can I pay on delivery?".
- Handover triggers: customer ready to pay, wants negotiation, corporate or bulk enquiry, complaint, anything the assistant cannot answer.
- Tone: professional, warm, concise Nigerian English; how to respond to Pidgin; never pushy.
- What never to say: fake scarcity, unverified claims, medical or financial advice, promises about delivery dates you cannot guarantee.
Step 2: prepare product, pricing and policy content
The assistant is only as accurate as its content. Prepare:
- Product or service catalogue with names, variants, specifications, what is included, images where the channel supports them.
- Prices with an "as of" date and the rule for updates. If prices change weekly with exchange rates, connect the assistant to a live price source or instruct it to give ranges and confirm on handover.
- Delivery and fulfilment: areas, fees, timelines, pickup options, installation where relevant.
- Payment methods: bank transfer details (shared only through verified flows), Paystack or Flutterwave links, POS on delivery, part-payment terms.
- Policies: returns, warranties, cancellations, refunds.
- Proof of legitimacy: CAC registration, physical address, verified social handles, which the assistant can share to reassure cautious buyers.
How to Build an AI Knowledge Baseeliable.
Step 3: choose channels and platform
For most Nigerian businesses the primary channel is WhatsApp, followed by Instagram DMs and the website.
- [WhatsApp Business Platform](https://developers.facebook.com/docs/whatsapp) (API): required for an AI assistant on WhatsApp. It allows automated conversations, multiple agents, logging and templates for follow-ups. Conversation charges are USD-based; verify current pricing with Meta or your provider. How to Build an AI WhatsApp Chatbot in Nigeria.
- Instagram and Facebook messaging: connected through Meta's messaging APIs so the same assistant handles DMs.
- Website chat: a light widget using the same engine, with a WhatsApp handover button. How to Add ChatGPT to a Business Website.
Platform options:
| Option | Fits | Trade-off |
|---|---|---|
| No-code sales chatbot builder | Simple catalogues, small teams | USD subscriptions, limited CRM and pricing logic |
| Assistant built on a model API with your own backend | Custom pricing rules, CRM integration, full control | Needs a developer and maintenance |
| Sales assistant inside an existing CRM or commerce platform | Businesses already on a capable CRM | Dependent on that platform's AI features |
Step 4: connect the CRM and lead capture
A sales assistant without a CRM is a chat log nobody reads. Every conversation should create or update a contact with:
- name, phone (WhatsApp), email if given,
- product interest and quantity,
- location and delivery needs,
- budget or price sensitivity signals,
- stage (new, qualified, quoted, follow-up, handed over, won, lost) and lost reason,
- the conversation summary and next action.
If you do not have a CRM, a structured Google Sheet or a simple pipeline tool is a workable start. How to Connect AI to Your CRMthe integration patterns. How to Automate Lead Qualification With AI.
Step 5: set pricing, discount and escalation rules
Answer-ready summary: The assistant should give prices only from an approved source, mention discounts only within limits you configure, build quotes from rules (unit price, quantity, delivery zone, installation), and escalate to a person for negotiation, bulk or corporate enquiries and anything outside the rules. Pricing behaviour must be deterministic code or configuration, never a free-form model decision.
Rules that work in the Nigerian market:
- List prices from a source of truth. A price table or live API, not the model's memory.
- Quote builder: quantity breaks, delivery fees by zone, installation fees, VAT handling where applicable. Output a written quote the customer can forward.
- Discount policy: either none ("our prices are fixed; here is what is included") or a small configured allowance, with anything beyond routed to a salesperson.
- Validity: quotes carry an expiry date because of exchange-rate movements.
- Payment instructions: only through verified flows (payment links, or bank details displayed from configuration, never typed by the model), with a clear message about how to confirm payment.
- Escalation: negotiation, bulk, corporate, complaints, custom specifications, and any request where the assistant's confidence is low.
Step 6: build follow-up sequences
Follow-up is where sales are won and where staff are least consistent. Configure sequences the assistant runs automatically:
- Enquiry without decision: a gentle check-in after 24 hours with a useful addition (a photo, a specification, delivery detail), then one more after three days, then stop unless the customer engages.
- Quote sent: reminder before expiry, offer to answer questions or book a call.
- Payment pending: confirmation request and help if the transfer failed.
- Post-purchase: delivery updates, satisfaction check, cross-sell only where appropriate.
On WhatsApp, follow-ups outside the 24-hour customer-service window must use approved message templates; plan the templates as part of the build. How to Automate Sales Follow-Up With AI. Keep an opt-out and respect it.
Step 7: test, launch and measure
- Build a test set of fifty real conversations from your WhatsApp history, including price haggling, Pidgin, typos, and off-topic messages.
- Score the assistant on accuracy of answers, correctness of prices, appropriate handovers and tone.
- Run in shadow mode for a week: the assistant drafts, a salesperson approves or edits before sending.
- Launch on one channel with disclosure ("You are chatting with our sales assistant; a team member can join at any time").
- Measure weekly: enquiries handled, response time, qualified leads, quotes issued, handovers, conversions by source, lost reasons, and USD usage.
- Review conversations for wrong answers and missed handovers; update the playbook and content.
- Expand to more channels and products once the numbers hold.
What changes for Nigerian sales conversations
Price is the first question. Customers ask "how much?" before anything else. The assistant should answer directly (or give a clear range with the reason) rather than deflecting; deflection loses trust.
Negotiation is expected. Many customers will ask for a discount. Decide your policy and let the assistant state it gracefully, then hand over if the customer persists and the deal is worth it.
Trust signals matter. Scam fears are real. The assistant should share CAC registration, physical address, verified payment links and clear policies, and never pressure.
Bank transfer confirmation. Many purchases end with a transfer and a screenshot. Connect payment confirmation to your gateway or bank alerts where possible; otherwise, hand over for manual confirmation and never mark an order paid on the strength of a screenshot alone.
Delivery realities. Lagos traffic, interstate logistics and pickup preferences shape expectations. Give honest timelines and let customers choose delivery partners or pickup where you offer them.
Instagram-led discovery. Many enquiries begin on Instagram; connect DMs so the assistant can continue and move the customer to WhatsApp for checkout.
Data protection. Names, phone numbers and purchase histories are personal data under the Nigeria Data Protection Act 2023. Keep a privacy notice, minimise collection, secure the CRM, and use AI vendors with no-training terms. Confirm obligations with the NDPC.
USD costs. Model usage and WhatsApp conversation charges are USD-based. Use efficient models for routine turns and monitor spend in naira.
Example: a solar installation company in Lagos
Example (hypothetical): A solar and inverter installation company in Lagos receives dozens of WhatsApp and Instagram enquiries daily, mostly "how much for a 5kVA system?", "do you install in Ajah?", "can I pay in instalments?", and "how long will the battery last?". Two sales staff reply when they can, forget to follow up, and keep leads in their personal phones.
The company builds an AI sales assistant on the WhatsApp Business Platform and Instagram. It answers product questions from a catalogue of packages, asks qualification questions (location, daily power needs, appliances, budget range, timing), builds an indicative quote from a configured price table with installation and delivery by zone, and captures every lead into the CRM with a stage. It books site-survey appointments into the team's calendar, sends follow-ups on quotes, and hands over to a salesperson when a customer wants to negotiate, needs a custom design, or is ready to pay.
The sales staff now spend their time on site surveys, custom designs and closing, and management sees every lead and where it stalls. The assistant is instructed never to promise battery lifespans beyond the manufacturer's stated figures. This is a hypothetical scenario, not a Linestech client result.
How much does an AI sales assistant cost in Nigeria?
Answer-ready summary: Indicative 2026 costs: a no-code sales chatbot on WhatsApp costs a USD subscription plus ₦100,000–₦500,000 setup; a custom AI sales assistant with catalogue grounding, quote rules and CRM integration runs roughly ₦800,000–₦3,000,000; a multi-channel assistant with quote builder, calendar booking, payment confirmation and reporting costs ₦3,000,000–₦6,000,000+. Add monthly USD model usage, WhatsApp charges, hosting and maintenance. Actual quotes vary with scope, vendor and exchange rate.
| Scope | Indicative one-off (₦) | Recurring | Notes |
|---|---|---|---|
| No-code WhatsApp sales bot | 100,000–500,000 setup | USD subscription, WhatsApp charges | Simple catalogues, fixed prices |
| Custom assistant with CRM and quote rules | 800,000–3,000,000 | USD usage, hosting, maintenance | Most SMEs |
| Multi-channel with booking, payments, reporting | 3,000,000–6,000,000+ | USD usage, WhatsApp charges, hosting, maintenance | Higher volumes, field sales |
| Playbook and content preparation | 100,000–500,000 or internal time | Quarterly refresh | Often underestimated |
Indicative 2026 ranges; actual quotes vary with scope, vendor and exchange rate. Separate build from running costs and compare two or three written quotations on identical scope. AI Chatbot Development Cost in Nigeriaia break down the components.
Cost drivers: catalogue complexity, pricing logic, number of channels, CRM and calendar integrations, payment confirmation, languages and reporting depth.
Mistakes to avoid
- Letting the model set prices. Prices come from a table or API. Models misremember and hallucinate numbers.
- No handover for negotiation. Customers who want to haggle and get a scripted refusal leave. Route them to a person when the deal matters.
- Pushy scripts. Nigerian buyers respond to clarity and respect, not pressure. Fake urgency damages the brand.
- Personal phone numbers. Leads on staff phones leave with staff. Use the WhatsApp Business Platform and a CRM.
- Ignoring the 24-hour window and template rules on WhatsApp; follow-ups fail silently.
- Marking orders paid on screenshots. Confirm through the gateway or bank.
- Stale content. Prices and stock change; assign an owner and a refresh schedule.
- No conversation review. The transcripts are the best sales training material you have.
Conclusion
Building an AI sales assistant is mostly a sales design exercise wrapped in technology: a clear playbook, accurate product and price content, deterministic pricing and escalation rules, a CRM that records every lead, follow-up sequences that respect WhatsApp's rules, and honest measurement. Get those right and the assistant answers instantly, qualifies consistently, quotes correctly and hands warm leads to people who close. Get them wrong and you have an expensive way to give customers the wrong price.
If your business is losing leads to slow replies and forgotten follow-ups, Linestech can help you design the playbook, build the assistant on WhatsApp, Instagram and your website, and connect it to your CRM and payment flows.
Frequently asked questions
Can an AI sales assistant close sales on its own?
For simple, fixed-price products with online payment, yes: it can answer, quote, send a payment link and confirm through the gateway. For negotiated, high-value or custom sales, it should qualify, quote and hand over to a person who closes.
Will customers know they are talking to an AI?
They should. Disclose it at the start and make it easy to reach a person. In practice, customers care more about fast, accurate answers than about who gives them, and honesty prevents frustration later.
How does the assistant handle "how much?" when prices vary?
It gives the list price or a clear range from your price source with the reasons it varies (size, quantity, delivery zone), then builds a quote after two or three qualifying questions. Deflecting the price question loses trust; giving a wrong price loses money. Both are avoided by grounding prices in configuration.
Can it work with our existing CRM?
Most modern CRMs have APIs the assistant can use to create and update contacts and deals. If your CRM is a spreadsheet, the assistant can update that too, though a proper CRM makes reporting and follow-up much easier.
What happens when the customer replies in Pidgin or a local language?
Language models handle Pidgin and common phrases in major Nigerian languages reasonably well. Test with real messages and set the assistant's reply style in the playbook. For languages your team does not support, route to a person.
How do we stop the assistant from promising delivery dates?
Instruct it to give the standard timeline from your policies ("Lagos 1–2 working days, other states 3–5") and to say that exact dates are confirmed at dispatch. Never let it invent a date.
Does it replace our sales team?
It replaces the repetitive part of their work: first replies, basic questions, qualification and reminders. Sales staff spend more time on demos, negotiations, relationships and closing. How Nigerian Businesses Can Use AI Without Replacing Staff.
Sources and further reading
Figures, platform rules and regulations change. These are the primary references behind this article and the places to check before you act on it.
