AI receptionists andautomations made simple.

A talking website is a modern website with a browser-based AI voice assistant. Visitors can ask questions naturally, hear approved answers, explain what they need and take an agreed next step such as booking an appointment. It adds conversation to the website without replacing its normal pages, navigation or contact options.
See the Talking Website packages →

A talking website gives visitors immediate help without making them search through every page. It can answer common questions, qualify enquiries, capture useful details and offer appointments while interest is high. This can be particularly useful for mobile visitors and people who would rather speak than complete a long form, while human contact remains available when needed.
Compare the packages →

Speaking can be more convenient when someone is using a phone or needs to explain a problem in their own words. A Stanford-led mobile text-entry study found that English speech input was three times faster than touchscreen typing in that particular test. This does not mean every voice-assistant conversation will be three times faster, but it helps explain why speaking can reduce input effort. A talking website can also ask an immediate follow-up question instead of making the visitor compose a long message or search through several pages.
Read the Stanford mobile input study →

Yes. Voice should be an additional choice, not the only way to use the website. Some visitors may be in a noisy or private setting, prefer reading and typing, or have a hearing, speech or cognitive disability that makes a voice interface difficult. A well-designed talking website keeps its normal pages, navigation, forms and human contact options available alongside the voice assistant.
Read the W3C accessibility guidance →

Web Workshop’s Talking Website package is $990 including GST. It includes a choice of one long page or up to five SEO-ready pages, an AI website assistant, one booking calendar, copywriting, first-year hosting, SSL, backups, two revision rounds and three months of launch support. LLM and platform usage is separate and depends on use.
See current package details →

The assistant uses an LLM—the system’s language “brain”—to understand a visitor’s spoken question and respond conversationally. It uses approved information from the website plus its own customisable knowledge base, which can add or refine answers for the business. It can also connect to selected tools such as a booking calendar. Clear instructions tell it what it may answer, what action it may take and when it should refer the visitor to a person.
Explore how Talking Websites work →

Not quite. A chatbot usually exchanges typed messages, while a talking website lets a visitor speak and hear a spoken response. Both may use an LLM, approved business information and connected tools, but the voice experience also needs microphone permission, audio design and concise conversational answers. The normal website should still provide readable pages, navigation and contact choices.

A talking website speaks with someone who is already visiting the website through their browser. An AI phone receptionist answers calls made to a business phone number. They can use related knowledge and booking tools, but each channel needs its own conversation flow, testing and usage setup. Web Workshop offers the browser assistant in the Talking Website package and adds phone answering in AI Receptionist Launch.
Compare website and phone packages →

Often, yes. An existing website may be able to add a secure voice assistant if its platform supports the required code and the site uses HTTPS. We first check compatibility, design, accessibility and likely effects on page performance. If the existing site is difficult to adapt or needs replacing, the assistant can instead be included in a new Talking Website.
Ask about your existing website →

Yes, when the required tools are configured. It can collect agreed contact and enquiry details, check a connected calendar, offer genuinely available times and create a booking. Critical details such as names, email addresses and appointment times should be confirmed, and the assistant should only say a booking succeeded after the calendar confirms it.

Visitors do not normally need an app. A talking website can run in a supported mobile browser over a secure HTTPS connection, and the visitor must allow microphone access before the conversation starts. Audio quality and connection speed can vary by device, browser and network, so normal navigation, enquiry forms and phone contact should remain available as alternatives.

They can be designed with privacy and security safeguards, but the business must decide what information is collected, why it is needed, who can access it and how long it is retained. Spoken details supplied to an AI assistant may be personal information. Australian businesses should use appropriate collection notices, minimise unnecessary data and consider their obligations under the Privacy Act and Australian Privacy Principles. This is general information, not legal advice.

The voice assistant itself does not guarantee better rankings, traffic or AI citations. Search engines and AI systems still need useful, accurate and crawlable page content with descriptive headings and internal links. That is why important answers should remain visible on the website rather than existing only inside the assistant. Clear FAQ content and accurate structured data can help machines understand the topic, but they do not guarantee a special search result or inclusion in an AI answer.
See the Talking Website approach →

An AI receptionist is a voice assistant that answers business phone calls, speaks with callers, answers approved questions, collects useful details and completes agreed next steps. Those steps can include booking an appointment, transferring a call, creating a customer record or alerting a team member.

AI means artificial intelligence: software that can recognise speech, understand the likely meaning of a caller’s words and produce a relevant response. It does not think like a person. It works from language patterns, approved business information, connected tools and rules for when a human should take over.

When a call arrives, speech recognition converts the caller’s words into text. An AI model uses your approved business information and call rules to decide how to respond, and a synthetic voice speaks the answer. The system can then save details, book an appointment or hand the call to a person.

It can answer calls, explain opening hours and service areas, qualify enquiries, capture contact details, book appointments, send confirmations, update connected systems and escalate urgent or sensitive calls according to rules set by the business.

Yes. An AI receptionist can answer calls after hours, during busy periods or when staff are unavailable. Different greetings and escalation rules can be used for business hours, evenings, weekends and emergencies.

In most setups, yes. Calls to an existing business number can usually be forwarded to the AI receptionist for all calls, overflow calls, after-hours calls or calls that your team cannot answer.

Modern voice systems can speak naturally and handle conversational language, but performance varies with background noise, accents, names and industry terminology. Testing with real Australian call scenarios is important before launch, and the assistant should confirm critical details such as names, numbers and addresses.

Yes. The call flow can transfer selected calls, notify an on-call person or create a priority callback. Your business defines which situations need immediate human attention and what information should be collected before handover.

Yes. We can connect it with widely used calendar and scheduling platforms, including Microsoft Outlook, Google Calendar and Cal.com. Depending on the workflow, it can check availability, create appointments, send confirmations and support rescheduling.

Yes. We can connect with most CRM systems so caller details, notes and follow-up actions reach the correct customer record. If there is no native integration, we can usually create a connection through the CRM’s API, an automation platform or an MCP server.

It should not guess. It can explain that the information needs to be confirmed, collect the caller’s question and arrange a transfer, alert or callback according to your business rules. Unknown questions can also be reviewed later to improve the approved knowledge base.

It does not need to. Many businesses use AI to handle routine calls, overflow and after-hours enquiries while people continue to manage relationships, complex requests, judgement and sensitive situations. It can also support an existing receptionist by reducing repetitive phone work.

Cost depends on call volume, conversation complexity, integrations, support and usage. Web Workshop scopes the call flow first so the proposal reflects the business rather than a generic package. The useful comparison is total cost against missed enquiries, administration time and the value of faster response.

Setup time depends on the number of call types, business information, integrations and testing required. A focused first version is usually faster to design and validate than an assistant expected to handle every possible call from day one.

It can be designed with privacy and security safeguards, including collecting only necessary information, controlling access, selecting appropriate suppliers and setting retention rules. Australian businesses must consider their Privacy Act and Australian Privacy Principles obligations, particularly when calls include health or other sensitive information. This is general information, not legal advice.

It is most useful where calls are regularly missed, staff are interrupted by routine questions or enquiries arrive outside staffed hours. Common examples include trades, real estate, restaurants, clinics and other service businesses where a timely response can influence a booking or job.

LLM stands for large language model. It is a type of AI that has learned patterns in language from a very large amount of text. This allows it to understand everyday questions, follow written instructions and produce a useful reply. You can think of it as the language brain inside an AI receptionist or agent, although it still needs business information, clear boundaries and other software to do a real job.

An AI agent is a practical system built around an LLM. The LLM supplies language understanding and reasoning; the agent adds a goal, instructions, approved business knowledge, memory and controlled access to tools. Together, these let it work through a multi-step job, ask for missing information, choose an approved next action and report the result instead of only generating an answer. This is why agents are seen as the latest practical development of LLM technology, although they still operate within limits set by people.

Recent LLMs have become better at following detailed instructions, maintaining the context of a conversation, handling varied language and selecting a structured tool when one is needed. At the same time, more business software can be connected securely through APIs. Together, these advances allow an LLM to sit inside a controlled workflow and help complete useful work, although every agent still needs careful design, testing and supervision.

A multi-agent system uses two or more specialised AI agents that cooperate on different parts of a job. For example, one agent might classify an enquiry, another check a calendar or customer system, and another prepare a follow-up, with a workflow controlling what each may do and how information is passed between them. This can help with genuinely complex processes, but one well-designed agent is often simpler and more dependable for a focused business task.

The phone system provides the ears and voice, while the LLM provides much of the language understanding. It works out what the caller is asking, remembers the conversation, uses approved business information and decides what to say or do next. For example, it can understand several different ways of asking for an urgent plumber, collect the address and job details, and follow the business’s instructions for escalation.

A traditional chatbot usually follows a fixed script, menu or list of keywords. If the customer phrases a question differently or changes direction, it may get stuck. An LLM-based agent can understand more natural language, keep track of the conversation, ask a useful follow-up question and choose the next appropriate step. It still works within instructions and limits set by the business.

A chatbot mainly provides answers. An agent combines an LLM with instructions and carefully chosen tools so it can also complete approved tasks. For example, instead of only saying that appointments are available, an agent can check the calendar, offer suitable times, make the booking, update the customer record and send a confirmation. The tools provide the ability to act; the LLM helps choose and organise the steps.

Older automation works best when every input and step is predictable. Newer LLM-based agents can also work with everyday language and less tidy information, such as phone calls, emails, notes and documents. This can make it practical to sort varied enquiries, extract details from differently formatted invoices, prepare a draft reply, gather missing information or choose an approved next step without building a separate rule for every possible wording.

No. If a task is simple and always follows the same steps, ordinary automation is often faster, cheaper and more dependable. An LLM is most useful when the job involves conversation, documents, varied wording or choosing between several possible next steps. A good solution may use AI to understand the situation, then use straightforward automation to complete the predictable actions.

Only within carefully chosen limits. LLMs are capable but can misunderstand information or produce an incorrect answer. A business agent should have approved information, restricted access, clear instructions, records of what it does and a point where a person reviews or takes over. Low-risk tasks can run automatically, while sensitive, unusual or high-impact decisions should require human approval.

Business automation means using software to take care of routine admin automatically, so someone does not have to repeat the same steps by hand. For example, when an invoice arrives by email, an automation can save it, record the important details, send it to the right person for approval and file it for later. It handles the predictable parts of a job while people remain in control.

Think of automation as a checklist that runs itself. You decide what should start the job and what should happen next. For example, when a customer requests a quote online, the system can reply immediately, save their details, notify the right staff member and create a reminder to follow up. Nobody has to copy the same information from one place to another.

Many repetitive office tasks can be automated. Examples include sorting incoming invoices, saving attachments, replying to new enquiries, sending appointment reminders, following up unpaid invoices or quotes, updating customer records, requesting missing documents and preparing a daily summary for the owner. The aim is to reduce copying, checking and chasing while keeping important decisions with people.

Automation carries out a series of steps you have chosen, while AI can help understand information that is not always presented in exactly the same way, such as a phone call, email or invoice. They often work together. AI might read an emailed invoice and identify the supplier, amount and due date; automation can then save those details, request approval and schedule payment.

Start with one job that staff do often, takes time and is easy to explain. Good first choices include copying web enquiries into a customer system, sending appointment reminders, filing invoices or reminding staff to follow up quotes. A useful first automation should solve a clear problem, be easy to check and give people time back without changing the whole business at once.

Usually, yes. Email, calendars, online forms, accounting software, spreadsheets, customer databases and job-management systems can often share information. We first check the programs you already use and what each one allows. The goal is to make them work together so staff do not have to keep copying the same names, dates, messages or documents between systems.

Usually, the goal is to remove repetitive admin, not people. Automation can do the copying, filing, reminders and routine updates, while staff handle customers, make decisions and deal with anything unusual or sensitive. Used well, it gives people more time for work that needs judgement, experience and a human conversation.

The cost depends on what you want to automate and how many programs need to work together. A small job, such as sending a confirmation and creating a follow-up task from a web enquiry, costs less than rebuilding a process used across several teams. We first identify the time or money the task currently consumes, then recommend a solution that is sensible for the likely saving.

It depends on the job. A simple automation involving one or two programs is quicker than a process with many steps, approvals and unusual cases. Before it goes live, we agree on what should happen, connect the systems and test it with realistic examples. Starting with one focused task makes it easier to launch safely and improve over time.

Start by asking how often the task happens, how long it takes and what delays or mistakes cost the business. If five staff members each spend an hour a week copying information or chasing updates, that is time the automation may be able to return. A worthwhile automation should save more time, money or missed opportunities than it costs to build, run and review.

It should stop safely and tell the right person what needs attention. For example, if an invoice is missing a purchase order number or a customer record cannot be found, the automation can place the item in a review list and alert a staff member. Important work should never disappear silently; there should always be a clear record and a human way to finish the job.