A WhatsApp AI chatbot should be built by a team with direct experience in the WhatsApp Business API's approval process, message template restrictions, and the 24-hour customer service window, since these platform rules shape the bot's design more than the AI itself does. Qualified partners typically combine three capabilities: registered WhatsApp Business Solution Provider access or a direct Meta API connection, a large language model integration grounded in your business knowledge so answers stay accurate, and backend connections to your order, CRM or booking systems so the bot can do more than answer FAQs. Agencies that only offer template-based flow builders without LLM grounding tend to produce rigid bots that frustrate customers with anything outside a fixed script. When evaluating a vendor, ask for a working reference deployment, a clear explanation of data handling given WhatsApp's end-to-end encryption model, and a maintenance plan for keeping the bot's knowledge current after launch. Regional providers with local language and payment integration experience matter more for WhatsApp than for web chat, since usage skews heavily toward specific markets. Nanobase AI builds WhatsApp Business API chatbots with full LLM grounding and backend integration rather than static template flows.
Most vendor pitches sound identical until you ask the right questions
A vendor proposal for a WhatsApp AI chatbot tends to read well regardless of whether the underlying product is a genuinely grounded LLM system or a rigid template-flow builder with an AI label attached, because both can demo a working conversation. The differences that actually matter, how the bot handles a question outside its script and whether its knowledge stays accurate as your policies change, only show up when you ask specific, pointed questions rather than watching a scripted demo.
A vendor evaluation scorecard
| Criterion | Weak signal | Strong signal |
|---|---|---|
| WhatsApp API access | Reseller with unclear approval status | Registered Business Solution Provider or direct Meta API access with a documented approval history |
| Answer grounding | Fixed decision-tree flows only | LLM grounded in your knowledge base with retrieval, able to handle phrasing outside the scripted flow |
| Backend integration | FAQ answers only, no system connections | Direct connections to order, CRM or booking systems for real actions |
| Reference deployment | Cannot show a live, comparable production bot | Working reference in a similar industry or use case |
| Data handling | Vague answer about WhatsApp's encryption model | Clear explanation of what data is stored, where, and for how long |
| Maintenance plan | No plan beyond initial launch | Defined process for updating the bot's knowledge as your business changes |
A vendor that cannot clearly answer the grounding and maintenance questions is likely to deliver something that works well at launch and degrades quietly afterward, which is the most common failure pattern in WhatsApp bot projects.
Why WhatsApp-specific experience matters more than general chatbot experience
WhatsApp Business API rules shape bot design more than the underlying AI does: the 24-hour customer service window restricts when free-form replies are allowed, and anything outside that window requires a pre-approved message template, a constraint that does not exist for a website chat widget. A vendor without direct experience navigating Meta's approval process and template restrictions often underestimates the design and testing time needed, leading to a bot that violates messaging rules or gets flagged during review. Regional experience matters too, since WhatsApp usage skews heavily toward specific markets, and a vendor with local language and payment integration experience in those markets typically ships a more usable bot than one with only generic chatbot experience.
Questions to ask before signing
- Can you show a live production WhatsApp bot you built, not just a demo environment?
- How do you handle a customer message that falls outside the bot's expected flow?
- What happens to the conversation when the 24-hour messaging window expires mid-issue?
- How do you keep the bot's answers current when our pricing or policies change?
- What data do you store, where, and who has access to it?
A vendor's answers to questions two and four, handling off-script messages and keeping knowledge current, reveal more about long-term fit than the initial demo ever will.
Frequently asked questions
Is a WhatsApp chatbot different enough from a website chatbot to need a specialist vendor?
The underlying AI grounding approach is similar, but the messaging window rules, template approval process, and media handling for images customers send are WhatsApp-specific, so direct experience with those constraints reduces the risk of a bot that technically works but violates platform rules.
Should we require a registered Business Solution Provider?
It is a reasonable requirement for larger deployments, since it typically signals a more established, compliant relationship with Meta, though some legitimate smaller vendors operate through approved resellers, so treat it as one signal among several rather than an absolute filter.
How do we verify a vendor's grounding claims before committing?
Ask to test the bot's reference deployment directly with a question phrased differently from an obvious scripted example, and see whether it retrieves an accurate answer or falls back to a generic or incorrect response.
What is a reasonable timeline for a WhatsApp chatbot vendor to deliver a working pilot?
For a scoped use case with existing knowledge base content, four to eight weeks to a working pilot is a reasonable benchmark, with longer timelines expected for deep backend integrations or multiple languages.
How Nanobase AI helps
Nanobase AI builds WhatsApp Business API chatbots with full LLM grounding and direct backend integration rather than static template flows, and can walk through this scorecard against your specific requirements before you commit to a vendor. This evaluation connects directly to the technical build process for a WhatsApp Business API chatbot powered by an LLM, and to broader AI agents and process automation work once the bot needs backend actions rather than answers alone.
Ready to discuss your project? Contact Nanobase AI or email hello@bumu.tech.