AI Chatbot for Nigerian Businesses: 2026 Build vs Buy Guide

Key Facts

• A custom AI chatbot for a Nigerian business typically costs ₦300,000 to ₦1,500,000 to build.
• Running cost lands at ₦40,000 to ₦300,000 a month, depending on volume.
• WhatsApp moved to a per-template message pricing model on 1 January 2026.
• Customer-initiated service messages within the 24-hour window are free at the Meta layer.
• Utility templates sent inside an open service window are free; outside it, they are charged per message.
• A 72-hour free entry window opens when a customer reaches you through a Click-to-WhatsApp ad or a Facebook Page CTA.
• A trained AI chatbot typically deflects 40 to 70 percent of repetitive customer questions in the first 90 days.
• The Nigeria Data Protection Act 2023 sets a 72-hour breach notification rule (Section 40) and applies to all chat data containing personal information.

AI Chatbot for Nigerian Businesses: 2026 Build vs Buy Guide

If your team is drowning in repetitive WhatsApp messages, missed leads after hours, or support tickets that all ask the same five questions, a chatbot is the obvious next step. The harder question is whether to buy a ready-made one, build a custom one, or do something in between, and how the new Meta pricing changes the maths.

This guide gives you the real cost, the real trade-offs, and a clear way to decide based on the size of your business, the channels you serve, and the rules you have to follow.

The figures and patterns here come from chatbot projects Nexoris Technologies has built for clients between 2025 and 2026, the Meta WhatsApp Business Platform documentation, the Nigeria Data Protection Act 2023, the Nigerian Communications Commission, DataReportal Digital 2026 Nigeria, and current public pricing pages from Paystack, Flutterwave, and the major large language model providers.

Quick Answer: What an AI Chatbot Costs and Does in Nigeria in 2026

A working AI chatbot for a Nigerian business sits in three price tiers, depending on how much it has to do.

A bot trained on your own content, connected to WhatsApp and your website, with a clean handover to a human agent, sits comfortably in the hybrid tier for most Nigerian SMEs and mid-market businesses.

Why Nigerian Businesses Are Putting Bots on WhatsApp in 2026

The short answer is that customers expect instant replies, and human teams cannot be online at 11 p.m. on a Sunday. Three things have made 2026 the year chatbots stopped being a "nice to have" in Nigeria.

The first is reach. According to NCC industry data published in early 2026, Nigeria had 151.6 million internet subscribers in January 2026, the vast majority connecting through mobile. DataReportal's Digital 2026 Nigeria report puts WhatsApp among the top messaging platforms in the country by usage, with most Nigerian internet users on it daily. If your customers will not download your branded app, they will still send you a WhatsApp message.

The second is the new Meta pricing. On 1 January 2026, Meta moved the WhatsApp Business Platform from per-conversation to per-template message pricing. Service messages from customers stay free within the 24-hour window. Utility templates sent inside that window are free.

Outside the window, marketing, utility, and authentication templates are billed per message, with volume discounts now available on utility and authentication categories. The headline effect for Nigerian businesses is that customer service deployments are structurally cheaper than they were in 2024.

The third is the maturity of large language models. The 2026 generation of models handles African English, Nigerian Pidgin, and major Nigerian languages reasonably well, with locally trained options closing gaps on accent and idiom. Moniepoint launched M, an AI chatbot for Nigeria's informal economy, in late 2025, joining Zenith Bank's ZiVA and Airtel's Airtel Assist on the list of serious deployments by major Nigerian institutions.

Rule-Based, AI Chatbot, or AI Agent: Which One Are You Buying?

These three labels are not the same thing, and the price difference is large. Buyers ask for chatbots and end up paying for agents, or pay for agents and only get rule-based bots. Here is the difference in plain English.

A rule-based chatbot follows a fixed script. It shows menus, waits for a button click, and routes the user down a defined path. It cannot handle anything off-script. Build cost is low, running cost is low, and the experience feels like a phone IVR.

An AI chatbot uses a large language model to understand free-form questions and reply in natural language. It is grounded in your own knowledge base (FAQs, policies, product catalogue) so the answers match your business. It can hand over to a human when it does not know.

An AI agent does everything an AI chatbot does and also takes actions. It books an appointment, opens a ticket, generates a payment link, looks up an order, or updates a CRM record on the user's behalf. This is where 2026 chatbot work is heading.

For most Nigerian businesses in 2026, the right answer is an AI chatbot now and an AI agent in the next phase, once you see where the volume is.

How an AI Chatbot Actually Works

A 2026 AI chatbot is not one piece of software. It is six pieces working in sequence, and knowing the sequence helps you understand what you are paying for and where things break.

A customer sends a message on WhatsApp, your website, or another channel. The message hits the channel layer first. For WhatsApp that is your BSP. For your website it is a chat widget. For Instagram and Facebook it is the Meta Messenger API. The channel layer passes the message to your backend over a webhook.

The backend pulls the customer's history if it has any, then runs the retrieval step. This is where RAG searches your knowledge base for the passages most likely to answer the question. The top three to five passages are picked and sent on with the original message.
The LLM step comes next. The model receives the question, the retrieved passages, and a system prompt that tells it how to behave (tone, escalation rules, what to refuse). It generates a draft reply.

If the question needs an action, like checking an order, generating a payment link, or booking an appointment, the backend calls the right integration (Paystack, your CRM, your booking system) and feeds the result into the reply.

The reply goes back through the channel layer to the customer. The whole round trip takes one to three seconds on a healthy setup. If the confidence score is too low, or the user asks for a human, the conversation is handed off with the full transcript attached.

Build, Buy, or Go Hybrid: Which Fits Your Business

The right path depends on three questions: how many conversations you handle, how regulated your data is, and how deeply the bot needs to plug into your other systems.

• Buy a SaaS plan if you handle under 1,000 conversations a month, mostly answer the same questions, and do not need the bot to read or write to your other systems. Most Nigerian retailers, salons, small clinics, and service businesses sit here.
• Go hybrid if you handle 1,000 to 10,000 conversations a month, want the bot to use your own knowledge base, and need it to talk to your CRM, ticketing, or booking system. This is the most common deployment for Nigerian SMEs and mid-market businesses.
• Build custom if you are in a regulated industry (banking, lending, health, insurance), need full ownership of the data and the model behaviour, or want the bot to drive deep workflows across your ERP, payments, and case management. Banks, HMOs, and licensed lenders sit here.
  
  Learn more about custom ERP systems in Nigeria in 2026.

What an AI Chatbot Costs in Nigeria in 2026

The real cost has four parts: the build, the platform fees, the message fees, and the model fees. Here is what each one looks like.

• One-time build cost: A SaaS plan often has zero or low setup. A hybrid deployment built on a SaaS platform with a custom knowledge base, integrations, and proper testing costs ₦300,000 to ₦800,000. A custom build with its own backend, retrieval system, and CRM ties costs ₦600,000 to ₦1,500,000. Add ₦200,000 to ₦600,000 if you want voice and IVR coverage on top of text channels. The build effort follows the same shape you see when scoping mobile app development costs in Nigeria, with a smaller surface area but tighter integration work.
• BSP fees: A Business Solution Provider gives you access to the WhatsApp Business Platform, message templates, an inbox, and analytics. Plans typically run ₦25,000 to ₦150,000 a month, depending on volume and features. Some BSPs charge a one-time onboarding fee of $200 to $1,000.

The BSPs Nigerian businesses can realistically use in 2026 fall into a few bands:

Pricing and standing change. Verify current rates and template approval times with each provider before locking a contract.

• Meta message fees: Under the per-template pricing model effective 1 January 2026, you pay for marketing, utility, and authentication templates. Service messages and utility templates inside the 24-hour customer service window remain free. Per-template costs for Nigeria vary by category and update periodically. Verify against current Meta documentation before locking a campaign budget.
• Large language model token fees: If your bot uses an LLM, it pays per token of input and output. For a moderate-volume Nigerian business handling 3,000 to 5,000 chats a month, this lands at ₦40,000 to ₦150,000 a month, billed in dollars.
• Hosting and maintenance: Backend hosting, monitoring, logs, and small fixes run ₦30,000 to ₦100,000 a month for hybrid and custom builds.

12-Month Total Cost of Ownership

The build is the smallest line at every volume. Token and BSP fees scale faster than people expect.

How Meta's New 2026 Pricing Model Works

As of 1 January 2026, Meta charges per delivered template message rather than per conversation. There are four message categories, each priced differently.

• Service messages: Sent inside the 24-hour service window after a customer messages you first. Free at the Meta layer.
• Utility templates: Order confirmations, shipping updates, appointment reminders, account alerts. Free if delivered inside the 24-hour service window. Charged per message outside it. Volume discounts apply on utility templates.
• Marketing templates: Promotions, offers, re-engagement campaigns, broadcasts. Charged per delivered message. The most expensive category.
• Authentication templates: OTPs, two-factor codes, login verifications. Charged per delivered message, with volume discounts.

Click-to-WhatsApp Ads and the 72-Hour Free Window

A Click-to-WhatsApp (CTWA) ad is a Facebook or Instagram ad that opens a WhatsApp conversation when the user taps it. Under Meta's 2026 rules, every CTWA tap opens a 72-hour free entry window during which the business can reply to that customer at no Meta message cost.

The practical effect is that paid acquisition turns into free service traffic. Every Naira spent on a CTWA campaign opens a three-day window where a chatbot can answer questions, qualify the lead, take a payment, or book an appointment with no template fees on top of the ad spend.

The 2026 Nigerian playbook looks like this. Run small, well-targeted CTWA campaigns daily rather than large weekly bursts. Let the chatbot handle the volume inside the 72-hour window. Keep marketing templates (the most expensive category) for re-engagement of customers whose window has closed.

Most well-run deployments end up with 70 to 85 percent of message volume in free windows, paying Meta only for the small slice of marketing pushes that genuinely need a template.

The practical playbook for Nigerian businesses: treat WhatsApp as a reactive channel with a chatbot doing the heavy lifting on free service messages, use CTWA ads to open free 72-hour windows, and keep marketing template usage targeted rather than blasted.

Channels Worth Covering in Nigeria

Most Nigerian businesses get the best return from one or two channels, not five. Here is the order most projects ship in.

• WhatsApp Business Platform comes first. It is where the customers already are. A WhatsApp bot connected to your knowledge base and a human handover delivers the largest deflection in the shortest time.
• Website chat comes next. It captures leads from people researching you, and the same knowledge base powers it.
• Instagram and Facebook Messenger come third, useful for retail, hospitality, and consumer brands. They share the same underlying platform as WhatsApp through Meta.
• Voice and IVR come fourth, mainly for call centres, banks, telcos, and HMOs handling high call volume. Voice AI is covered in its own section below.
• SMS and USSD still matter for parts of the customer base without smartphones. Most projects do not need them at launch.

For a typical Nigerian SME or mid-market business in 2026, WhatsApp plus website chat covers 80 to 90 percent of useful interactions.

How to Design the Conversation and Knowledge Base

A chatbot is only as good as the content it stands on. The most common reason a Nigerian deployment underperforms is a thin knowledge base, not weak technology.

Three steps make the difference.

1. Pull together your knowledge base before building. FAQs, product or service descriptions, policies (returns, refunds, cancellation, shipping), step-by-step guides, escalation rules, and a tone document. Put each item in plain text and label it cleanly. The richer this set, the better the bot performs.
2. Map the top 20 questions your team gets. These are the conversations the bot must handle perfectly on day one. Get them from your support inbox, WhatsApp logs, or email tickets. Write the ideal answer for each in your brand voice.
3. Define the handover rule. When the bot is unsure, when the user asks for a human, when the conversation has gone three rounds without resolution. Write the rule down, then test it by deliberately confusing the bot.

A 200-question knowledge base built once, kept fresh quarterly, is the difference between a bot that deflects 60 percent and one that deflects 20 percent.

How to Stop the Bot from Inventing Answers

The single biggest reason AI chatbots embarrass Nigerian businesses is hallucination, where the model invents a confident answer that is not in your content. The fix is a technique called Retrieval-Augmented Generation, or RAG.

In a RAG setup, the bot is not free to answer from the model's general knowledge. Every time a user asks a question, the system pulls the most relevant passages from your knowledge base, then asks the LLM to generate an answer using only those passages. If nothing relevant is found, the bot says it does not know and offers to hand the conversation to a human.

A serious 2026 chatbot for a Nigerian business should be RAG-grounded, with a confidence score on every answer, an audit log of which passages were used, and a clear "I do not know, let me hand you over" path when the score is too low. Without these, the bot will hallucinate sooner or later, and the first time it does it in a customer support conversation about money, the cost is more than any year of SaaS fees.

Bots That Speak Pidgin, Yoruba, Hausa, and Igbo

A chatbot that only speaks formal English misses how Nigerian customers actually write. Plenty of messages come in Pidgin, with Yoruba, Hausa, or Igbo phrases mixed in. Modern large language models handle this better than they did two years ago, but they still need help.

Three things make a multilingual Nigerian bot work. The first is a clear language scope, decided at the start, so you train and test for the languages the bot will actually see. The second is a knowledge base in the same languages, so the bot has phrases to match against, not English ones translated badly. The third is a tested handover rule, so when the bot is unsure, it routes the customer to a human agent who speaks the same language.

For voice, locally trained automatic speech recognition and text-to-speech models handle Nigerian accents better than generic foreign tools. If you serve a customer base that prefers voice notes over typing (very common in Nigeria), this is worth budgeting for.

Voice AI for Nigerian Businesses

Voice AI moved from experimental to deployable in 2025 and is now a real option for Nigerian businesses with high call volume. The category covers AI that picks up phone calls, holds a natural conversation, books appointments, takes orders, and hands over to a human when needed.

A voice AI tier on top of a Nigerian WhatsApp and website bot typically adds ₦300,000 to ₦900,000 to the build and ₦80,000 to ₦300,000 a month in running costs. The economics make sense for call centres, banks, telcos, healthcare networks, and any business spending more than ₦500,000 a month on inbound phone support.

The trade-offs to know. Latency matters more on voice than on text. Nigerian accent recognition still benefits from locally tuned models. Compliance under NDPA still applies because phone transcripts are personal data. Phone numbers and outbound calling are subject to NCC rules.

Voice AI works best as a phase-two project, after the text chatbot has shown deflection in the data.

What NDPA 2023 Requires for Chat Data

A chatbot processes personal information. Names, phone numbers, account references, complaints, sometimes ID numbers and health details. All of that is regulated under the Nigeria Data Protection Act 2023, enforced by the Nigeria Data Protection Commission.

Five obligations matter most for chatbot deployments.

1. Lawful basis and consent: You need a clear legal reason to process the messages. For most customer service bots, that reason is performance of a contract or legitimate interest. Marketing bots usually need explicit consent.
2. Purpose and storage limits: Keep transcripts only for as long as you need them, and tell the customer what that period is in your privacy notice.
3. Data subject rights: Customers can ask what you hold on them, ask you to correct it, or ask you to delete it. The bot's design should make those requests easy to handle.
4. Cross-border transfer: If your LLM or hosting sits outside Nigeria, the NDPA expects either an adequacy decision or contractual safeguards. Most major cloud providers offer the right contracts. You have to sign them.
5. Breach notification: If chat data is exposed, you have 72 hours from awareness to notify the NDPC and, depending on severity, the affected customers (NDPA 2023, Section 40).

Where a chatbot makes decisions that significantly affect a customer (approving a loan, blocking an account, declining a claim), NDPA 2023 Section 37 requires that the customer can request human review and contest the outcome. Build the human escalation path in from day one rather than retrofitting it after a complaint.

A small compliance package baked into the build (consent flow, retention policy, audit logs, role-based access, breach response runbook) costs less in design than it does after a complaint or a regulator visit.

Integrations That Move the Bot from Useful to Valuable

A chatbot that only answers FAQs is useful. A chatbot that books appointments, looks up orders, processes payments, and updates your CRM is the one that pays for itself. Five integrations show up in most Nigerian deployments.

• CRM (HubSpot, Salesforce, Zoho): The bot creates and updates leads automatically, so sales sees the conversation history in one place.
• Helpdesk and ticketing: The bot opens tickets, attaches transcripts, and hands them over to humans without losing context.
• Payments. Paystack, Flutterwave, Monnify, and Remita: The bot generates payment links, confirms transactions, and triggers receipts.
• Bookings and calendars: The bot offers slots, books, and sends reminders. Hospitality, beauty, and clinic businesses live here.
• Internal databases and ERPs: The bot looks up order status, inventory, account balances, and account history without humans pulling reports.

For a deeper look at how these integrations are built and managed, see our custom software development services .

From Decision to Live Bot: A Six-Week Plan

A hybrid AI chatbot in Nigeria goes from signed scope to live deployment in roughly six weeks. The schedule below is the one Nexoris Technologies uses on most mid-market builds. SaaS-only projects compress this to one to two weeks. Custom builds in regulated industries stretch it to eight to twelve.

1. Week 1: Scoping and knowledge base curation. Confirm the channels, the languages, the integrations, and the handover rules. Pull the top 100 to 200 questions your team gets and write the ideal answer for each. Sign the data processing agreements with your LLM provider and your BSP.
2. Week 2: BSP setup and template approval. Register your WhatsApp Business number with the chosen BSP. Draft and submit your utility, marketing, and authentication templates for Meta approval. Set up your test website widget. Approvals usually take 24 to 72 hours per template, so submit early.
3. Week 3: RAG pipeline and integrations. Build the retrieval layer over your knowledge base, wire in the LLM, and connect the first two integrations (usually CRM and payments). Define the confidence threshold and the handover rule in code.
4. Week 4: Internal UAT. Your team uses the bot like a customer for a full week. Hard questions, edge cases, deliberate confusion, off-topic chatter. Every failure goes into a tracked log, and the knowledge base or prompt gets updated each evening.
5. Week 5: Soft launch. Open the bot to 10 to 20 percent of real customer traffic. Watch the handover rate, the confidence scores, and the conversation logs daily. Fix the top five issues every morning.
6. Week 6: Go-live and analytics setup. Open the bot to all customer traffic. Stand up the dashboard tracking deflection rate, average handle time, first-response time, and customer satisfaction. Schedule a weekly review of the conversations the bot got wrong.

How to Measure the Return on a Chatbot

Five numbers tell you whether a chatbot is paying for itself.

• Deflection rate: The percentage of conversations handled fully by the bot, with no human involvement. A well-trained bot on a Nigerian customer service line lands at 40 to 70 percent within 90 days.
• Cost per resolution: What it costs you, on average, to close a customer query. Bot resolutions are typically 5 to 15 times cheaper than human resolutions.
• Average handle time: How long a conversation takes from start to finish. A bot answers in seconds; humans in minutes.
• First-response time: The most visible metric to customers. Bots take it from minutes or hours to under 10 seconds.
• Conversion lift on lead bots: For sales-focused bots, the percentage of conversations that turn into qualified leads. A good lead-qualification bot lifts conversion 15 to 40 percent over a static contact form.

Worked ROI example

A 200-staff Nigerian fintech with three customer support agents handling 6,000 WhatsApp conversations a month. The average human handle time is 8 minutes. Fully loaded cost per agent-hour is ₦4,500.

Pre-bot monthly cost: 6,000 × 8 ÷ 60 = 800 hours × ₦4,500 = ₦3,600,000 a month.
After deploying a hybrid chatbot deflecting 55 percent of conversations:
Bot handles 3,300 conversations.

Humans handle 2,700 × 8 ÷ 60 = 360 hours × ₦4,500 = ₦1,620,000.
Bot running costs: ₦400,000 a month.

Post-bot monthly cost: ₦2,020,000. Monthly saving: ₦1,580,000. Annual saving: ₦18,960,000. The build paid back in roughly six weeks.
These numbers vary by industry and conversation type, but the shape holds.

Worked Use Cases by Industry

Short examples to make the cost ranges and ROI shape concrete.

1. Bank or fintech. Customers check balances, raise disputes, request statements, and report card issues over WhatsApp. A custom bot with KYC checks and core banking integration. Build: ₦1,200,000 to ₦2,500,000. Monthly: ₦300,000 to ₦800,000. Deflection: 50 to 65 percent. Fintech-specific scope and compliance considerations are covered in detail in our fintech app development in Nigeria guide.
2. Hospital or clinic. Patients book appointments, ask about test results, request prescription refills, and triage symptoms. A hybrid bot with EHR integration and a clinician handover. Build: ₦600,000 to ₦1,200,000. Monthly: ₦150,000 to ₦300,000. Deflection: 40 to 60 percent. Clinics often deploy this alongside a hospital management system so the bot has live data to work with.
3. School. Parents ask about fees, calendars, results, and events. A SaaS-tier bot with school management system integration. Build: ₦200,000 to ₦500,000. Monthly: ₦60,000 to ₦150,000. Deflection: 60 to 75 percent.
4. Hotel or short-let. Guests check rates, book rooms, request services during their stay, and leave reviews. A hybrid bot with property management system integration. Build: ₦500,000 to ₦900,000. Monthly: ₦120,000 to ₦250,000. Deflection: 50 to 70 percent.
5. E-commerce store. Shoppers ask about products, track orders, and request returns. A SaaS-to-hybrid bot with Shopify or WooCommerce integration. Build: ₦300,000 to ₦700,000. Monthly: ₦80,000 to ₦200,000. Deflection: 55 to 70 percent.
6. Real estate agency. Buyers ask about listings, request viewings, and submit documents. A hybrid bot with CRM integration and document upload. Build: ₦400,000 to ₦800,000. Monthly: ₦100,000 to ₦200,000. Deflection: 45 to 65 percent.
7. NGO or development organisation. Beneficiaries ask about programmes, eligibility, and field operations. A multilingual bot with offline-friendly handover. Build: ₦500,000 to ₦1,000,000. Monthly: ₦100,000 to ₦250,000. Deflection: 50 to 65 percent.

Why Chatbot Projects Fail in Nigeria, and How to Avoid It

Most failed chatbot projects fail for the same six reasons. Knowing them in advance is the cheapest way to protect the investment.

1. Thin knowledge base: The single most common cause. A bot trained on 20 FAQs and a homepage cannot handle real customer questions. Build the 200-question base before you write any code, then keep it fresh quarterly.
2. No handover rule: Without a clear rule for when to pass a conversation to a human, the bot either escalates everything (defeats the point) or escalates nothing (frustrates customers). Define the rule in writing: low confidence, user asks for a human, three rounds without resolution, or any conversation about money beyond a defined threshold.
3. Wrong channel choice: Picking website chat when 90 percent of your traffic is on WhatsApp, or building a voice bot for a customer base that types. Match the channel to where the conversations already happen, not where you wish they did.
4. No Nigerian language testing: A bot trained and tested only on formal English breaks the first time a customer writes "Abeg, I no fit login." Test the bot with real Nigerian messages from real Nigerian users before launch.
5. BSP picked on price alone: A cheaper BSP with slow template approval, weak inbox tools, or thin Nigerian support costs more in delays than the savings. Approval speed and support quality matter more than the headline monthly fee.
6. NDPA treated as an afterthought: Building first and bolting compliance on later costs three to five times more than designing the consent flow, retention policy, and audit logs into the build from week one. The NDPC has issued real fines. Plan for compliance from week one rather than after a complaint.

The vendor checklist in the next section is structured to surface most of these risks before you sign.

How to Choose a Chatbot Vendor in Nigeria

Send any provider these ten questions before you sign. The quality of the answers tells you everything.

1. Is the bot trained on our own content using a RAG-style approach, or only on generic data?
2. Where is the conversation data stored, and under whose contract?
3. How does the handover to a human agent work, with what information attached?
4. Which channels does the price include, and which are extra?
5. What does the BSP layer cost, and which BSP do you use?
6. What does the LLM cost at our expected volume, and in which currency?
7. How do you handle Pidgin and Nigerian languages?
8. What does the NDPA compliance package look like in writing?
9. What is the exit path if we want to switch vendors in two years?
10. Can we see three Nigerian deployments you have built in the last 12 months?

A serious provider answers all ten in writing without hesitation. The wider criteria for picking a software partner, beyond chatbots, are covered in how to choose a software development company in Nigeria, with a fuller list of vetted teams in our best software development companies in Nigeria round-up.

Chatbot Terms Used in This Guide

• BSP (Business Solution Provider). The authorised company that connects your backend to the WhatsApp Business Platform. You cannot use WhatsApp programmatically without one.
• CTWA (Click-to-WhatsApp). A Facebook or Instagram ad that opens a WhatsApp conversation when tapped, with a 72-hour free reply window for the business.
• RAG (Retrieval-Augmented Generation). An architecture where the bot retrieves relevant passages from your knowledge base, then asks the LLM to answer using only those passages.
• LLM (Large Language Model). The AI model that generates the bot's replies. GPT, Claude, Gemini, and Llama are the common ones.
• NLU (Natural Language Understanding). The bot's ability to read what the user means, not just what they typed.
• Intent. The reason behind a user message. "I want to cancel" and "How do I cancel" share the same intent.
• Fallback. What the bot does when it does not know the answer. A good fallback hands the user over to a human.
• Handover. The point at which the bot passes the conversation to a human agent, with the full transcript attached.
• Template. A pre-approved message format used to start a conversation on WhatsApp outside the 24-hour service window.
• Service window. The 24-hour period that opens when a customer sends you a message, during which you can reply freely without using a template.
• Embedding. A mathematical representation of text that lets the bot find similar passages in your knowledge base.
• Vector store. The database where embeddings are kept and searched.

How We Wrote This Article

The Naira figures and deployment patterns here come from chatbot projects Nexoris Technologies has built for Nigerian and pan-African clients between 2025 and 2026. Meta pricing references reflect the per-template message model effective 1 January 2026. Nigerian market context comes from the Nigerian Communications Commission's industry data published in early 2026 and DataReportal's Digital 2026 Nigeria report.

Compliance references draw on the Nigeria Data Protection Act 2023 and the Nigeria Data Protection Commission's General Application and Implementation Directive. Currency-sensitive items use Central Bank of Nigeria official rates. All public sources were checked at the time of writing.

Source References

1. Nexoris Technologies internal delivery data: AI chatbot projects completed 2025 to 2026
2. Meta WhatsApp Business Platform pricing and policy documentation, 2026 per-template model effective 1 January 2026 (developers.facebook.com/docs/whatsapp/pricing)
3. Nigeria Data Protection Commission: NDPA 2023, Section 37 (automated decisions) and Section 40 (breach notification), and the General Application and Implementation Directive (ndpc.gov.ng)
4. Nigerian Communications Commission subscriber statistics, January and February 2026 (ncc.gov.ng)
5. DataReportal: Digital 2026 Nigeria Report
6. Punch newspaper: Moniepoint unveils AI chatbot to support informal businesses, 31 October 2025
7. Federal Inland Revenue Service: VAT on services (firs.gov.ng)
8. Central Bank of Nigeria: official exchange rates (cbn.gov.ng)
9. Paystack pricing (paystack.com/pricing)
10. Flutterwave pricing for receiving payments
11. GSMA Mobile Economy Sub-Saharan Africa 2025 / 2026
12. OWASP LLM Application Top 10 (security guidance for AI chat systems)
13. Gartner: customer service automation research