AI chatbot integration with your help desk, CRM and WhatsApp
We build a custom AI chatbot and connect it to the software your customers and team already use. It reads the conversation, answers or acts, and writes the result back where your team works.
Updated
One chatbot, one thread
One memory across WhatsApp, web chat, email and your help desk, so customers don't have to repeat themselves.
You choose where it runs
A hosted model, a cloud tenancy in your country, or air-gapped on your own servers.
Engineers from $45 an hour
Published rates, and a written quote after discovery for a defined project.
What is AI chatbot integration?
AI chatbot integration connects a language model you control to the software your customers and staff already use. The chatbot reads a conversation, answers or acts, and writes the result back where your team works. You choose the model and where it runs, and one chatbot can serve several systems at once.
The connection matters as much as the model. In Zendesk's CX Trends 2026 report, 74% of consumers said they find it frustrating to tell their story over and over to different agents, and 76% would choose a company that keeps one conversation thread (Zendesk). A chatbot with its own separate memory can't deliver that.
Once it is connected, a chatbot can do more than answer questions:
- Read the customer's history from your CRM before it replies.
- Check an order, book a slot or update a case in your own systems.
- Write its reply and a note back to your help desk, so your team sees the whole thread.
- Pass the conversation to a person, with the context attached.
Which platforms can a custom AI chatbot connect to?
We connect custom chatbots to Zendesk, Salesforce, HubSpot, WhatsApp and Genesys Cloud, using the route each platform provides for outside bots. Each route has its own plan requirements, so we confirm your license before you budget. Any other system with an API can usually be connected too.
| Platform | Route for a custom chatbot | What you need |
|---|---|---|
| Zendesk | Sunshine Conversations API and the switchboard | Suite Professional or above, which includes a low-volume version of Sunshine Conversations. |
| Salesforce | REST API for cases, External Services or Apex callouts to your own model, and the Agent API to reach an Agentforce agent | An external client app, and an integration user with only the access it needs. |
| HubSpot | Custom Channels API, and the chatbot webhook action | Sales Hub or Service Hub Enterprise for custom channels. A Data Hub Professional or Enterprise subscription for the webhook action. |
| WhatsApp Business Platform Cloud API | A Meta business account. Meta charges per delivered message, by category. Meta's terms bar general-purpose AI assistants. | |
| Genesys Cloud | Digital Bot Connector and Open Messaging | Confirm with Genesys whether your plan needs its bring-your-own-technology bundle. |
Routes and plan requirements as published by each vendor, checked on 19 October 2026. They change, so confirm on the vendor's page before you sign.
We also connect chatbots to Zoho, Freshdesk, Slack, Shopify, Google Workspace, Microsoft 365 and internal systems that expose an API. For a first version or a long tail of small tools, we use Zapier, Make or n8n rather than custom code.
Using something else? Our guide to which platforms integrate custom AI chatbots covers more help desks, CRMs, contact centers and messaging channels.
Our rates
Published prices in US dollars. Monthly plans are 160 hours, about 15% below the hourly rate.
Mid-level developer
3 to 5 years
$25 / hour
$3,400 a month (160 h)
Senior developer
6 or more years
$35 / hour
$4,750 a month (160 h)
AI / LLM engineer
Agents, RAG, IBM watsonx
$45 / hour
$6,100 a month (160 h)
Senior AI engineer
Leads AI architecture and delivery
$60 / hour
$8,150 a month (160 h)
Two-week trial with the developer you choose. No charge if you stop within the first five working days. US Eastern clients get 09:00 to 12:00 US Eastern time through an evening shift, at 15% extra. Full pricing and comparisons.
How does an AI chatbot integration project work?
First we map your channels and systems and confirm the license each route needs. Then we agree how conversations reach a person, build the chatbot and its connectors, and test them on your real conversations. It goes live on one channel first, and others are added once that one runs well.
- 1
Map your channels and systems
List where customers write to you, which systems hold the answers, and where messages and records will be stored. Note the country for each.
- 2
Confirm the route and the license
Choose the route for each platform and check which plan includes it, because many routes sit above the entry plans.
- 3
Agree the hand-off
Decide who receives the conversation when the chatbot stops, what context they get, and how control comes back.
- 4
Build the chatbot and connectors
Each message reaches the chatbot by webhook. It reads the customer's history, answers or acts, and writes back through the platform's API.
- 5
Test on real conversations
Measure answers and resolutions against a test set of your own past conversations, including the difficult ones.
- 6
Go live on one channel, then add more
Each extra system adds testing and more ways to fail, so start with one or two and add others once they run well.
What happens when the chatbot needs a person?
The chatbot passes control the way the platform expects, such as Zendesk's switchboard or Genesys's connector, with the conversation history attached. Your team picks up in the tool they already use, and the customer doesn't have to start again. The rules for when this happens are agreed before the build.
- When it hands off: the customer asks for a person, the chatbot isn't confident, or the topic is on your list of things it must not handle.
- What your team sees: the whole thread, the customer's record and what the chatbot has already tried.
- How control comes back: the conversation returns to the chatbot only when your rules allow it.
Should you use your platform's own AI agent or a custom chatbot?
Use your platform's own agent if you have one channel, one help desk, modest volume and no data residency rule. It is the fastest route. Choose a custom chatbot when you need one assistant across several systems, when data must stay in your country or building, or when it must act in your own systems.
A custom chatbot is usually worth it when one or more of these is true:
- One assistant across systems. It keeps the same memory on WhatsApp, in your app and in your CRM.
- Data rules. The model can run in your country or on your own servers.
- Volume. Per-resolution fees grow with every conversation, while a custom chatbot costs mostly the same each month. By our arithmetic, at about $1 a resolution, one that answers from your documents can cost less from about 3,400 conversations a month. Below about 1,600, the platform's own agent is cheaper.
- Actions. It can check an order, book a slot or issue a refund in systems the built-in agent may not reach.
- Control. You decide the model, the prompts, the escalation rules and the test set.
A custom chatbot is also a system you own, so someone has to keep it current. If your need is simple, we will tell you on the first call that your platform's own agent is the better buy.
Where does customer data go?
That depends on where the model runs. A custom chatbot can use a hosted model with no-training terms, a cloud tenancy in your country that you control, or air-gapped servers of your own. Messages still pass through the channel you use, so we map where each system stores messages, logs and records.
For data that must stay inside your network, see private AI. Our security page sets out how we handle access, code and data.
How much does AI chatbot integration cost?
There are three costs: the platform's AI or API fees, channel fees such as WhatsApp's per-message rates, and the build. By our planning estimates, a chatbot that answers from your documents takes 240 to 480 hours to build, and one that takes actions takes 800 to 1,600. Engineering is $45 to $60 an hour.
| What it does | Build hours | At Webify.AI rates |
|---|---|---|
| Answers from your documents | 240 to 480 | $8,400 to $16,800 |
| Reads your business systems | 480 to 960 | $22,800 to $45,600 |
| Takes actions in your systems | 800 to 1,600 | $42,000 to $84,000 |
These are planning estimates, not quotes. Each figure blends the rates of the two engineers that build needs, so it can work out below the AI engineer rate alone. A defined project is quoted in writing after discovery, before any build starts.
A first version scoped to one task with one or two integrations typically takes six to twelve weeks. Model usage, hosting and maintenance time come on top.
Our chatbot cost guide and AI agent cost guide show what moves the figure up or down. Every rate we charge is on our pricing page.
What do we need to start?
Tell us where your customers write to you and which systems hold the answers, including your plan level on each. A sample of real conversations and someone who can judge whether an answer is right are enough for a first assessment. Production access can come later.
- Your platforms: the help desk, CRM and channels you use, and the plan you are on.
- Real conversations: a sample of past tickets or chats, with the answers you would expect.
- An owner: someone who knows the content and can review answers each week.
- Your rules: what the chatbot must never do, and any data that can't leave your systems.
Frequently asked questions
Can we use our own AI model inside Zendesk, Salesforce or HubSpot?
Yes. Zendesk takes a custom bot through Sunshine Conversations and the switchboard, and HubSpot's Custom Channels API registers it as a channel in the shared inbox. With Salesforce, your model runs outside and works on cases through the REST API, and Flow can call it through External Services. Each route needs a plan that includes it.
Does WhatsApp allow custom AI chatbots?
Yes, when the chatbot serves your own customers. Meta's terms bar AI providers, including general-purpose AI assistants, when AI is the primary function rather than incidental or ancillary. Meta charges per delivered message by category, so check its current rates for your country before you budget.
Can one chatbot work across WhatsApp, web chat, email and our help desk?
Yes, if it is built behind the APIs rather than inside one product. The chatbot holds the conversation and the customer record, and each channel is a connector.
Do we have to replace our help desk or CRM?
No. The chatbot connects to the tools you already run, and your team keeps working in them. You may need a higher plan on a platform if its API or bot connector sits above your current one.
Can the chatbot take actions, or only answer questions?
It can take actions if you allow it, such as updating a case or issuing a refund. Each action gets its own limits and testing, and a person can approve anything that carries risk until you are happy to relax that.
Is a platform's own AI agent ever the better choice?
Often. With one channel, one help desk, modest volume and no data residency rule, it can be live in days and the vendor maintains the model.
Who keeps the chatbot running after launch?
A custom chatbot is a system you own, so it needs someone to keep it current. Our retainers cover that, or your own team can take it over.
Where is Webify.AI based?
Webify.AI is headquartered in Ahmedabad, India, and the first call is free.
Related pages
- Which platforms integrate custom AI chatbots
- What an AI chatbot costs to build
- Zendesk chatbot integration
- WhatsApp chatbot integration
- HubSpot chatbot integration
- Salesforce chatbot integration
- Custom agent vs off-the-shelf chatbot
- Channels and integrations
- AI customer support agents
- Ready-made agents in the Bot Store
- Private and offline AI
- AI agent development
Talk to an engineer about your chatbot integration
A conversation about your goals and systems, and a straight answer on whether an agent is the right tool. Paid work starts only after you have agreed a written scope.
- We sign your NDA before the first call if you need one
- You leave with a next step, even if the answer is "you don't need an agent"
- The code and IP are yours if you go ahead
Book a free first call
Pick a time, or send us the details and we will reply.