What it handles
- Questions it can't answer
- Requests for a person
- Serious or sensitive issues
- Users who switch language
- Personal details behind a login
System · Chat
AI Chat Agents
An AI chat agent answers every message on your website, WhatsApp, Instagram and the other channels people use to reach you, with an instant reply at any hour. It qualifies leads, books appointments and hands anything tricky to your team with the full conversation attached. Or it works for your staff, answering from your own SOPs.
- Instant replies on every channel
- 24/7
- Score on every prospect, in a study-abroad agency build
- 1–10
- From kickoff to go-live
- 3–4 weeks
Do you have any slots for a consultation next week?
Yes. Tuesday at 10:00 or Thursday at 14:30. Which suits you?
Thursday, please.
What is an AI chat agent?
An AI chat agent is an AI that holds text conversations on your website, in messaging apps and inside your team's own tools. Client-facing agents answer customers, recommend options, qualify leads and book appointments. Internal agents answer your staff from your SOPs and documents. Both are trained on your own information and hand over to a person when needed.
Every build includes a knowledge base made from your own content, a conversation designed around the agent's job, a branded chat widget, CRM integration, testing on test conversations before go-live, and a documented handover. When a question needs a person, the agent passes it to your team with the whole conversation attached. You own everything we build.
It's an AI chat agent, not a scripted chatbot: it answers in its own words from your knowledge base and uses live tools during the conversation.
Hi, I'm looking for a used hatchback under 12k.
Happy to help. Do you have a preference on fuel type or mileage?
Petrol, under 50,000 miles.
I've found three in stock that match. Here they are.
3 matches · from live stock
Would you like to book a test drive for one of them?
Lead capturedTest drive booked
What's our refund policy for cancelled bookings after 14 days?
Bookings cancelled after 14 days get a 50% refund, per the Customer Policy, section 4.
Can you draft the email to the customer?
Here's a draft for you to check before sending.
Answered from SOPsDraft ready
Sample conversations. Illustrative only, not real customers.
Policy figures in the internal sample are invented for the demo.
Client-facing or internal: which kind do you need?
You need a client-facing AI chat agent if the people asking questions are customers, leads, guests or applicants. You need an internal agent if they're your own staff, looking for answers in SOPs, documents or the CRM. Both are built the same way and can run on the same channels; only the audience and the data they may use change.
| Client-facing agent | Internal agent | |
|---|---|---|
| Who it talks to | Customers, leads, guests and applicants | Your own team |
| What it does | Answers questions, recommends options, qualifies leads, books appointments, escalates to a person | Answers from your SOPs and documents, trains new staff, helps with day-to-day work |
| Where it runs | Website widget, WhatsApp, Messenger, Instagram DMs, Telegram, private chat | Slack, Telegram, WhatsApp, private web chat |
| What it can see | Your knowledge base and the tools you approve | Only the documents, data and tools you approve |
One agent, every channel
| Channel | How it works | Good to know |
|---|---|---|
| Website chat widget | A branded widget on your site, matching your colours and logo, on mobile and desktop | The most common starting point for client-facing agents |
| Telegram | The agent runs as a Telegram bot | The quickest messaging channel to set up, and a simple home for internal agents |
| People message your business number and the agent replies | Needs a verified Meta business account, which takes longer to set up | |
| Facebook Messenger & Instagram DMs | The agent answers messages sent to your page or profile | Comment-to-DM flows are built as AI automations, not as a chat agent |
| Slack | The agent runs as an app in your team's workspace | For internal agents, so your team asks it where they already work |
| Private or logged-in chat | The agent sits behind a login | For guest portals, member areas and anything involving personal details such as bookings or finance applications |
How does an AI chat agent work?
An AI chat agent works in three steps. It gets instructions and a knowledge base built from your own content, it holds the conversation and uses tools such as calendar booking or stock look-up as it goes, and when the conversation ends an automation in Make.com, n8n or Claude routines writes the data to your CRM and notifies your team.
Instructions and knowledge
The agent gets a persona, a conversation design and a knowledge base built from your docs, PDFs, SOPs, spreadsheets, FAQs and website. Live information, such as stock levels or a customer's record, it looks up through a tool instead.The conversation
The agent answers questions, asks what it needs to know one question at a time, recommends options and uses its tools mid-conversation: showing matching products, booking a slot or pulling up a record.Post-conversation automation
When the conversation ends or a goal is reached, an automation formats the data, writes it to your CRM, Google Sheet or Airtable base, notifies your team where needed and sends any follow-up that was promised.
- Voiceflow
- Botpress
- Make.com
- n8n
- Claude routines
- GoHighLevel
- HubSpot
- Airtable
- Supabase
What's captured on every conversation
| Category | Examples |
|---|---|
| Conversation details | Date and time, channel, and the user's handle, number or team login |
| Contact details | Name, email, phone, needs, budget, preferences (client-facing agents) |
| Conversation analysis | Summary, reason for contact, lead score, and whether it was resolved, booked or escalated |
| Records | Full transcript, kept for review, tracking and compliance |
| Feedback | Ratings and comments, where you ask for them |
Tools the agent can use
- Calendar booking: checks real availability, books the slot and sends confirmations and reminders
- Live stock look-up: searches your inventory or catalogue by the customer's criteria and returns only what's available
- Order look-up: finds a customer's orders by order number or email and explains their status
- Discount codes: creates a single-use, time-limited code for the product a shopper wants, within your rules
- Product carousels: recommendations shown as swipeable cards with an image, key details and a next-step button
- CRM look-up and write-back: recognises returning customers, and every update goes through the automation layer
- Human escalation: raises a ticket or alerts your team instantly, with the full conversation attached
- Multimedia: sends images, maps, videos and links
- Follow-ups: summaries, reminders or a booking link after the conversation
- Feedback collection: asks for a rating and invites a review, and alerts your team privately to low scores
- Notifications: instant alerts in Slack, Telegram or email, plus daily digests
Which jobs can a chat agent do?
An AI chat agent can answer support questions, qualify and route leads, book appointments, recommend products from live stock and reply to every enquiry the moment it arrives. Internally, it can answer staff from your SOPs and query your business database in plain English. One agent can cover several of these jobs at once.
| Use case | What the agent does |
|---|---|
| Customer support | Answers common questions from your FAQs and policies, and escalates the rest with the full conversation |
| Lead qualification | Captures new enquiries, asks the right questions in conversation, scores the lead and routes it |
| Appointment setting | Books consultations, viewings and test drives into your calendar, with confirmations and reminders |
| Speed-to-lead | Replies to every web chat and DM enquiry the moment it arrives, day or night |
| Product and service recommendation | Recommends the best matches from your live stock or service range, shown as cards |
| Order tracking | Looks up a customer's orders live and explains where each one is |
| AI coaching assistant | Supports a coach's clients between sessions using the coach's own method |
| Feedback and review collection | Asks for a rating, invites a review and alerts your team privately to low scores |
| Use case | What the agent does |
|---|---|
| AI database assistant | Answers your team's questions across your business database in plain English and writes PDF reports |
| AI sales assistant | Gives your sales team instant answers on products and policies, and a summary of a lead's history before a call |
| Employee training assistant | Trains new staff from your SOPs, answers their questions and checks what they've learned |
| Market research assistant | Researches a niche your team names and saves the findings as a document or database record |
How do chat agents work in your industry?
Chat agents work the same way in every industry, but each industry asks different questions. Buyers ask about stock and test drives, guests about check-in and Wi-Fi, students about courses and visas. Each industry page shows the conversations a chat agent handles there, in that industry's own words.
What won't a chat agent do?
An AI chat agent won't guess. It answers only from your knowledge base and tool results, and when it doesn't know, it says so and passes the question on. It doesn't quote prices, make commitments or give legal, medical or financial advice unless you've agreed it should, and it collects personal data in line with rules such as GDPR.
Guardrails
Sample settings. Agreed with you at kickoff.
Your choices
Some behaviour depends on your preferences, so we agree it at kickoff:
- AI disclosure: whether, and how, the agent says it's an AI assistant when asked
- Prices: whether it discusses prices, and which ones
- Tone and persona: its name, personality and how formal it is
- Escalation: which topics go straight to a person, and who receives them
- Access (internal agents): who can use it, and which documents and systems it can draw on
Known limitations
- Only as good as its knowledge base. Missing or out-of-date information means weaker answers. When your information changes, the knowledge base needs updating.
- Meta channels take longer. WhatsApp, Messenger and Instagram need a verified Meta business account and follow Meta's rules on when a business can message someone first.
- Rich content varies by channel. Carousels, buttons and images display differently on each platform; some show a simpler list.
- Human handover. By default the agent alerts your team or raises a ticket rather than passing the live chat to a person. Live takeover can be added.
- Language accuracy. In multilingual builds, the agent performs best in its primary language.
- Phrasing varies. The agent may word answers slightly differently each time. That's normal, and it can be tuned.
Chat agents we've built
Real builds, anonymised. Each case study shows the brief, the build and how it runs.
Every chat becomes a scored lead or a support ticket
1–10
Score on every prospect
One portal, with an AI assistant that can see the whole database
1 portal
Replaced the team's spreadsheets
What's included in each package?
Every AI chat agent package includes knowledge base set-up from your content, a branded chat widget, conversations logged to your CRM, testing rounds and a documented handover. The packages differ in how many use cases, channels and languages the agent covers. All three take 3–4 weeks, and every build is scoped and priced on the discovery call.
| Core Agent | Advanced Agent | Full Chat AI System | |
|---|---|---|---|
| In one line | One agent. One job. Fully automated. | More use cases. More channels. More reach. | Chat AI for your customers and your team, built and handed over. |
| Ideal for | One clear use case, one channel and an existing CRM | Several use cases, more channels or multilingual support | Chat AI across client-facing and internal work |
| Use cases | One, such as customer support, lead qualification or employee training | Two, such as support with appointment setting | Full coverage, client-facing and internal |
| Channels | 1 | Up to 2 | All supported channels |
| Languages | English | Up to 2 | Up to 3 |
| Automation | Post-conversation CRM update and team alerts | Adds lead scoring and a daily digest | Adds a custom analytics dashboard |
| Knowledge base set-up, branded widget, conversations logged to your CRM | ✓ | ✓ | ✓ |
| Testing and refinement rounds | 1 | 2 | 3 |
| Handover sessions | 1 | 1 | 2 |
Add-ons, scoped before the build so there are no surprise costs: a custom chat interface · extra channels · extra languages · extra use cases or agents · live calendar booking · live stock look-up with product carousels · CRM setup · secure login · a custom analytics dashboard · live human takeover.
Prefer to pay monthly? See the Build Partner RetainerWhich tools do we build chat agents on?
We build AI chat agents on Voiceflow or Botpress, with Make.com, n8n or Claude routines handling the automation after each conversation. Conversations and leads go to the CRM you already use, such as GoHighLevel, HubSpot, Airtable, Google Sheets or Supabase. AI models are chosen per build, based on your needs and the latest versions available.
- Agent platform
- Voiceflow, Botpress
- AI models
- Chosen per build, based on your needs, availability and the latest versions
- Channels
- Website widget, Telegram, WhatsApp, Facebook Messenger, Instagram, Slack (internal agents)
- Automation
- Make.com, n8n, Claude routines
- CRM and data
- GoHighLevel, HubSpot, Airtable, Google Sheets, Supabase
- Calendars
- Google Calendar, Cal.com
- Notifications and follow-up
- Slack, email, Telegram
What do we need from you?
To build an AI chat agent we need CRM access, your content for the knowledge base, access to your website and channels, a persona brief with conversation goals, your escalation rules and one technical contact who can grant access quickly. Slow access and missing content are the most common causes of delay.
- CRM access with read and write permissions, or approval for us to set one up
- Your content for the knowledge base: docs, PDFs, SOPs, spreadsheets, FAQs, website content, policies, opening hours and locations. Prices only if you want the agent to discuss them.
- Live stock or catalogue data, if the agent recommends products
- Website access to install the widget, or a contact on your web team
- Channel access: a Telegram bot, your Meta business account for WhatsApp, Messenger and Instagram, or approval to add the agent to Slack
- For internal agents: who will use it, and which documents, data and tools it may use
- A persona brief and conversation goals, including what it should never discuss
- Escalation rules: what goes to a person, who gets it and how
- One technical point of contact who can grant access quickly
How long does it take?
An AI chat agent takes 3–4 weeks from kickoff to go-live. WhatsApp, Messenger and Instagram can add time, because Meta has to verify your business account first. After launch we watch the first live conversations with you, and bugs found in the first 2 weeks are fixed at no extra cost under the warranty.
| Step | What happens |
|---|---|
| 1. Discovery | We agree whether the agent is client-facing, internal or both, and map what people ask, where they ask it and what a good conversation ends with. |
| 2. Onboarding and kickoff | An onboarding checklist on day one, then a kickoff call to agree persona, conversation design, channels, escalation rules and timeline. |
| 3. Build | CRM integration, knowledge base, prompts, agent tools, post-conversation automation and channel deployment, in that order. |
| 4. Testing and refinement | We test every path and edge case, then you review the agent on a private test link. We refine it from both. |
| 5. Go-live and handover | A recorded walkthrough plus documentation: architecture, conversation design, knowledge base sources, workflow exports and field mapping. |
| 6. Launch check | We watch the first live conversations with you to confirm everything works as tested. |
2-week warranty. Bugs or unexpected behaviour found in the first 2 weeks after go-live are fixed at no extra cost. After that, support is optional: Maintenance, Active Management or Growth Partner.
Frequently asked questions
Questions
Answer
Will an AI chat agent replace our team?
No. A client-facing AI chat agent takes the repetitive questions off your team and hands them conversations that are already qualified, with the context attached. An internal agent gives your staff instant answers from your own material, so they spend less time searching and more time on work that needs a person.
What happens when the chat agent doesn't know the answer?
The chat agent says so, then passes the question to your team. It won't make something up. Every unanswered question is logged, so you can see what's missing and add the answer to the knowledge base. Over time, fewer questions need a person.
Can a chat agent use our live stock or product data?
Yes. We connect the AI chat agent to your inventory, catalogue or database through a tool, so it only recommends what's actually available and can show matches as product cards with images and a next step. This is an add-on, and we need a feed, database or spreadsheet we can read.
Which channel should we start our chat agent on?
For a client-facing AI chat agent, usually your website, because that's where most unanswered visitors are. Once it works well, the same agent can be added to WhatsApp, Messenger, Instagram or Telegram without starting again. Internal agents usually start where your team already chats, such as Slack, Telegram or a private web chat.
Is our data safe with an AI chat agent?
The chat agent collects only the fields your flow needs and stores them in your own systems. Personal details such as bookings or finance information can sit behind a login, internal agents only use the documents and systems you approve, and every build follows the data protection rules that apply to you, such as GDPR.
Will the chat agent tell people it's an AI?
That's your choice. We agree at kickoff whether, and how, the AI chat agent says it's an AI assistant when someone asks, and we'll point out any disclosure rules that apply in your market.
Should we have a chat agent or a voice agent?
It depends on how people reach you. If they call, an AI voice agent answers the phone. If they browse your website or message you, an AI chat agent answers there. Many businesses use both, sharing the same knowledge base.
Can we change the chat agent later?
Yes. Knowledge base updates and refinements from real conversations are covered by the optional monthly support plans after the 2-week warranty. Bigger changes, such as a new channel or a new use case, are scoped as a new piece of work.