In short: FYAdvisers builds website chat, email triage and AI receptionist duties that answer only from your approved sources and pass prices, commitments and out-of-scope questions to a person. If your answers already live on your website, Intercom, Zendesk or Tidio may be the faster start.
An AI chatbot for a Singapore business is worth building when the same questions arrive every day, the answers already exist somewhere in your business, and your team would rather spend its time on the conversations that need judgement. FYAdvisers builds AI chatbots and customer service automation for small and mid-sized B2B companies in Singapore: website chat, email triage and AI receptionist duties, grounded in your own information and designed to hand over to a person cleanly.
This sits within our wider AI integration work. If WhatsApp is your main channel, see our page on the WhatsApp AI chatbot.
What we build
Website chat
A chat window on your website that answers routine questions from your approved information, asks follow-up questions to understand the enquiry, and records the details in your CRM or inbox. When a visitor asks about price, availability or anything outside its scope, it collects their details and tells them who will follow up.
Email triage
Many B2B enquiries still arrive by email. AI can read each incoming message, work out what it is (new enquiry, existing order, complaint, supplier, spam), pull out the key details, and route it to the right person with a summary and a suggested reply. A person reviews and sends.
AI receptionist duties
The receptionist role covers greeting, routine answers, taking details and passing the conversation on. We build this for chat and email channels. Voice AI for phone answering is a separate product category with its own vendors; if that is what you need, we can help you compare options or find a provider.
Knowledge-base grounding
The chatbot answers from a set of sources you control, such as service descriptions, policies, FAQs, product sheets and past approved replies. This approach is often called retrieval-augmented generation. The model looks up the relevant passages before replying, so its answers follow your information rather than its general training. We agree who owns those sources and how they are kept current.
Handover to humans
Every chatbot we build has clear handover rules: which topics always go to a person, how the person is notified, and what the customer is told. The person receives the conversation so far and a short summary, so the customer does not have to repeat themselves.
Typical uses for Singapore B2B companies
- Enquiry intake. Collect company name, contact, requirement, timeline and location, then route to the right salesperson. Pairs well with our speed to lead work.
- Service and product questions. Coverage areas, lead times, minimum orders, documents required.
- Order and job status. Look up a reference number in your system and report the current status.
- Matching an enquiry to an offer. At Jeff Learning Center, a tuition centre, an enquiry form collects student details and AI suggests the best-matching course. The suggestion goes only to the owner, with an instant alert, and his team contacts the lead.
- Internal help desk. Staff ask questions about internal procedures and get answers with links to the source document.
How a project runs
- Discovery. We look at a sample of real conversations and emails, agree which questions the chatbot should handle, which it must hand over, and what a good answer looks like.
- Prototype with your real data. We load your approved sources and test against past questions, with personal details removed where we can. You see the answers side by side with what your team actually said.
- Build. We connect the chatbot to your website, inbox, CRM or other systems, using workflow automation or custom software where needed.
- Human review. Before launch, and for an agreed period after, your team reviews answers. Prices, commitments and advice always stay with a person.
- Monitoring. Conversations and handovers are logged, so gaps in the knowledge base and wrong answers can be spotted and fixed.
What it costs
Costs fall into three groups. Prices below are from each provider’s own pricing page in October 2026 and change often.
- Off-the-shelf subscriptions. Intercom’s pricing page lists its Fin AI agent at $0.99 per outcome, on top of seat plans from $29 per seat per month. Zendesk lists pay-as-you-go automated resolutions at US$2.00 each, with Suite Team at US$55 per agent per month billed annually. Tidio lists a free plan and paid plans from US$24.17 a month.
- Custom build. A one-off project to design, connect and test the chatbot. One software development company’s May 2026 guide to Singapore chatbot costs puts AI-powered FAQ bots at S$15,000 to S$35,000 and customer support bots at S$45,000 to S$90,000. Scope drives the figure more than anything else.
- Usage. AI model providers charge per million tokens of text processed. Anthropic lists Claude Haiku 5.5 at US$0.10 input and US$0.50 output per million tokens, and Google lists Gemini 3.1 Flash-Lite at US$0.25 and US$1.50. Messaging channels such as WhatsApp add their own charges.
The main cost drivers are the number of systems the chatbot connects to, how many channels it runs on, how much source material needs cleaning, and how much review your team wants. Our AI chatbot cost guide goes into more detail. We outline the cost of our own work after a conversation about scope.
Data and governance
A chatbot that collects names, contact details or enquiry content handles personal data under the PDPA. PDPC’s overview of the data protection obligations covers consent, notification, protection and retention. PDPC’s Advisory Guidelines on Use of Personal Data in AI Recommendation and Decision Systems, issued on 1 March 2024, explain how consent, notification and accountability apply when AI systems use personal data. IMDA and the AI Verify Foundation’s Model AI Governance Framework for Generative AI, released on 30 May 2024, sets out practical points on accountability, testing and transparency.
In practice, we agree with you:
- What the chatbot tells people about who they are talking to and how their data is used.
- Which AI provider processes the data, under its business terms, and where.
- What is stored, who can see it and how long it is kept.
- Which topics the chatbot must never answer on its own.
Build or buy
Buy first if your questions are general, your answers live on your website, and you are happy for the tool’s own interface to handle handover. Intercom, Zendesk, Tidio and similar products are mature and quick to switch on.
Build when the chatbot needs to look up or update records in your own systems, follow handover rules specific to your business, work across website, email and messaging with one set of rules, or keep data within an arrangement you control. Many projects are a mix: an off-the-shelf channel with custom connections behind it.
Who this suits
- B2B companies that receive a steady flow of similar enquiries by website or email.
- Teams where enquiries wait because the person who knows the answer is on site or in meetings.
- Businesses with good source material, such as service sheets and policies, that customers rarely read.
It suits less well if each enquiry needs a site visit or bespoke quotation before anything useful can be said. In that case, faster routing and a good intake form may do more than a chatbot.
What to prepare
- Twenty to fifty real customer questions, with personal details removed.
- The documents and pages that hold the correct answers.
- A list of topics that must always go to a person.
- Who on your team will review answers and own the knowledge base.
- The systems the chatbot would need to read from or write to.
Get in touch and we will tell you whether a chatbot, a simpler automation or an off-the-shelf tool fits best.
Frequently asked questions
What is the difference between an AI chatbot and an AI receptionist?
They overlap. An AI chatbot usually means a chat window on a website or messaging channel. An AI receptionist is a role: greeting enquiries, answering routine questions, taking details and routing the conversation to the right person. That role can run on chat, email or phone, and the same rules about grounding and handover apply to each.
Will the chatbot make up answers?
Language models can state wrong information confidently, so we limit the chatbot to your approved information, tell it to say when it does not know, and route anything outside that scope to a person. We test it against your real past questions before it goes live and keep a log of what it said.
Should we buy an off-the-shelf chatbot or have one built?
If your questions are mostly answered by your website and help articles, an off-the-shelf tool such as Intercom Fin, Zendesk or Tidio is often the faster start. A custom build makes sense when the chatbot has to read from or write to your own systems, follow your own handover rules or work across several channels.
Can the chatbot reply on WhatsApp?
Yes, through Meta's WhatsApp Business Platform, which has its own rules on opt-in, template messages and per-message pricing. Our WhatsApp AI chatbot page covers this in detail.
What does an AI chatbot cost in Singapore?
It depends on whether you subscribe to a tool, have one built, or both, plus usage charges for AI models and messaging. Our cost guide sets out the published pricing we checked in October 2026. We outline the cost of our own work after a conversation about scope.