Which Platforms Integrate Custom AI Chatbots With Support Software?

Which platforms integrate custom AI chatbots with existing customer communication software? Almost all of them. Every mainstream help desk, CRM, contact-centre platform and messaging channel now takes a custom AI chatbot, in one of three ways. Zendesk, Intercom, Salesforce, HubSpot, Freshworks, Zoho, Gorgias, Microsoft, Google, Amazon, Genesys and Twilio each expose an API or a bot connector that lets you plug in a model you control. Most of them also sell their own AI agent, billed per resolution, per action or per session. And the messaging channels themselves, led by WhatsApp with more than 3 billion monthly users, accept any bot that speaks their API.

So the question is rarely whether your software can take a custom chatbot. It is which route to take, what each route costs, and where your customer data goes once the bot is live. This guide answers those three questions for the platforms businesses actually run, with every price and rule linked to the vendor's own page.

It is written for support and operations leads, IT managers and founders who already have a help desk, a CRM or a WhatsApp number, and want AI in it without replacing it. We are Webify.AI, and we build and integrate these chatbots, so we have an interest; where a vendor's own AI agent is the better choice, we say so. Every figure was checked on 19 October 2026 and will change, so confirm the price on the vendor's page before you sign.

The short answer

A custom AI chatbot connects to existing customer communication software in one of three ways:

• Through the platform's own AI agent. You configure the vendor's assistant inside the product and pay per outcome. No code, fast, and the data stays inside that vendor's cloud.
• Through the platform's API or bot connector. Your own model, hosted where you choose, receives each message by webhook and replies through the API. Zendesk's Sunshine Conversations, Salesforce's Agent API, HubSpot's Custom Channels API and Genesys's Digital Bot Connector are examples.
• Through middleware. A bot framework, an integration platform or a messaging API sits between your model and several channels at once, so one brain answers on WhatsApp, web chat, email and the help desk.

The rest of this article goes platform by platform, then prices, then how to choose.

What a custom AI chatbot integration is

A custom AI chatbot integration is a connection between a language model you control and the software your customers and staff already use, so the bot can read a conversation, answer or act, and write the result back where your team works. It differs from a vendor's built-in AI agent in two ways: you choose the model and where it runs, and one bot can serve several systems at once.

Key facts at a glance

• WhatsApp has more than 3 billion monthly users, and its Business Platform accepts any bot (TechCrunch).
• Vendors' own AI agents cost about $0.50 to $2.00 per resolution; Salesforce charges $0.10 per action; Meta's Business Agent costs $2.00 per million tokens.
• 74% of consumers find re-explaining their issue very frustrating, and 76% would choose a company that keeps one conversation thread (Zendesk CX Trends 2026).
• Gartner expects agentic AI to resolve 80% of common customer service issues on its own by 2029.
• Zendesk, Salesforce, HubSpot, Genesys, Zoho and Google each document a route for bringing your own model; the links are in the platform section.

Why one chatbot across every channel matters

Customers do not think in channels. Zendesk's CX Trends 2026 report, based on more than 6,000 consumers and 5,000 service leaders across 22 countries, found that 74% of consumers find having to re-explain their situation to different agents very frustrating, and 76% would choose a company that lets them keep text, images and video in one conversation thread without restarting (Zendesk CX Trends). The same report says nearly 8 in 10 consumers find AI bots helpful for simple issues, and 51% prefer a bot when they want immediate service.

The volume is moving to automation. Gartner predicts that by 2029 agentic AI will autonomously resolve 80% of common customer service issues without human intervention, cutting operational costs by 30% (Gartner). Salesforce's State of Service report expects 50% of service cases to be resolved by AI by 2027, up from 30% in 2025 (Salesforce).

A chatbot that lives in one channel, with its own memory, works against all of that. The integration is the product.

Which platforms integrate custom AI chatbots with existing customer communication software?

Help desks and shared inboxes

• Zendesk. Zendesk sells its own AI agents and also takes outside bots. Sunshine Conversations is its messaging platform and API, and Zendesk documents connecting "another app or program, such as a custom bot" to it (Zendesk). The switchboard decides who holds each conversation: your bot passes control to Agent Workspace when a human is needed, which makes a ticket eligible for assignment at once (Zendesk Developers). Some Sunshine Conversations methods need a separate licence rather than the one bundled with Suite Professional, so check your plan (Zendesk).
• Intercom. Fin, Intercom's AI agent, is priced at $0.99 per outcome and can be used "with your current helpdesk including Salesforce and more", with a minimum monthly commitment such as 50 outcomes on other help desks (Intercom). Intercom's Messenger and inbox also connect to outside systems through its APIs and webhooks.
• HubSpot. HubSpot's Customer Agent is priced per resolved conversation. For a custom bot, the Custom Channels API registers your bot as a channel: HubSpot posts outgoing messages to your webhook URL, and you publish incoming replies back into the shared inbox, so agents see the whole thread (HubSpot Developers). The rule-based chatbot's "Trigger a webhook" action needs a Data Hub Professional or Enterprise subscription.
• Freshdesk and Freshworks. Freddy AI Agent is sold in session packs, and Freshworks' products expose REST APIs for ticket and conversation access. Session packs expire within the billing cycle, and when sessions run out the AI agent stops responding and chats go to live agents (Freshworks).
• Zoho SalesIQ. Zoho's Enterprise plan lets you "deploy a Gen AI-driven chatbot powered by in-built Zia or BYOK third-party AI", and the pricing page says you can configure "Anthropic Claude, Google Gemini, DeepSeek, or self-hosted LLMs to power your bots" (Zoho). That self-hosted option matters if your data must stay in-country.
• Gorgias. Built for e-commerce, Gorgias prices its AI Agent per resolved conversation, with no seat or per-message fee, and is built around a two-way Shopify sync plus 100 or more commerce apps (Gorgias).

CRMs and service clouds

• Salesforce. Agentforce is Salesforce's own agent layer. For custom work, the Agent API lets an outside application start a session, send messages synchronously or as a stream, and end the session (Salesforce Developers). Agents can also be invoked from Apex and Flow, so your own model can be called from inside a Service Cloud process.
• Microsoft Dynamics 365. Dynamics 365 Contact Center and Customer Service use Copilot Studio for agents. Copilot Studio agents can call external APIs and connectors, and a custom bot built on the Bot Framework publishes to Microsoft Teams for free on the standard channels (Microsoft).

Messaging channels

• WhatsApp Business Platform. The Cloud API accepts any bot. Meta charges per delivered message by category: marketing, utility, authentication and service. Replies inside the 24-hour customer service window are service messages, and each business number gets 1,000 free service messages a month. From 1 October 2026 service messages beyond that free tier are billed again, after being free since November 2024 (Meta). Meta also sells its own Business Agent, billed at $2.00 per million tokens, roughly 4 to 5 cents a message by Meta's estimate, and launched Business AI for small businesses in India in May 2026 (Meta).
• Messenger and Instagram. Both run on Meta's Messenger Platform APIs with webhooks, the same pattern as WhatsApp (Meta).
• Web chat, SMS and in-app. Twilio's Conversations API carries chat, SMS and WhatsApp in one thread and bills per monthly active user, from $0.05, with the first 200 free, plus the channel fees (Twilio).
• Microsoft Teams and Slack. Internal help desks live here. Teams bots on the Bot Framework's standard channels carry unlimited messages at no charge; the premium Web Chat and Direct Line channels are metered (Microsoft).

Contact centres and voice

• Genesys Cloud. The Digital Bot Connector lets end users "communicate directly with third-party bots" through Genesys Cloud Messaging, and Architect flows call your bot with a Call Digital Connector action (Genesys Developers). Older community posts say this needs a bring-your-own-technology bundle; confirm with Genesys.
• Amazon Connect and Lex. Lex is billed per request, $0.00075 for text and $0.004 for speech (AWS), and a Lex bot can hand to a Connect queue. Your own model can sit behind Lex through a Lambda function.
• Google Conversational Agents, the product formerly called Dialogflow CX. Chat requests are $0.007 on Flows and $0.012 on Playbooks; voice is $0.001 or $0.002 a second (Google Cloud). Webhook fulfilment lets your own service answer.

Middleware and standards

• Integration platforms. Zapier, Make and n8n move messages and records between a chatbot and thousands of apps without a custom connector, which suits a first version or a long tail of small systems.
• The Model Context Protocol. MCP is an open standard for describing the tools and data sources an AI model may use. Anthropic donated it to the Agentic AI Foundation under the Linux Foundation on 9 December 2025, co-founded with Block and OpenAI and supported by Google, Microsoft, AWS, Cloudflare and Bloomberg, with more than 10,000 public MCP servers at the time (Anthropic). For an integrator it means one connector to your CRM can serve several models.

The custom-bot route on each platform

What the platforms' own AI agents cost

Every vendor has moved to outcome pricing. The unit differs, which makes the bills hard to compare, so here they are side by side.

Two details decide what you actually pay. The first is what counts as a resolution. Zendesk's blog describes a resolution as a ticket that stays closed through a 72-hour quiet period (Zendesk); HubSpot's definition is support provided and no hand-off to a human for 72 hours (HubSpot). A customer who gives up is a paid resolution under both. The second is the platform fee underneath: Intercom's seats, Salesforce's editions, Dynamics 365 Contact Center at $110 a user a month (Microsoft).

The arithmetic at volume is simple. A support team closing 10,000 conversations a month through a $1.50 agent pays $15,000 a month for the AI alone, before seats. The same volume through a custom model on your own infrastructure is the cost of the server and the people who run it, whatever the volume. We set out that comparison, with hardware and electricity prices, in our article on private AI for UAE businesses.

Native AI agent or custom chatbot: how to choose

Use the platform's own agent when you have one channel, one help desk, modest volume and no residency rule. It is the fastest path, and the vendor keeps the model current.

Build a custom chatbot and integrate it when one or more of these is true:

• You need one assistant across several systems. The native agent in your help desk does not answer on WhatsApp, in your mobile app and in Salesforce with the same memory. A custom bot behind the APIs does, which is what the 74% who hate repeating themselves are asking for.
• Your data cannot leave the country or the building. A vendor's agent runs in the vendor's cloud. A custom model can run on your own servers or in an in-country tenancy, and only the message text passes through the channel. Zoho's self-hosted LLM option is the rare native route that allows this; elsewhere it means a custom integration. Our Private AI page covers what that involves.
• Arabic, or another language the vendor's model handles poorly. Gulf-built models such as Jais 2 lead the Arabic leaderboards, and a custom integration lets you use them. Our comparison of open-weight models has the scores.
• Volume. Above a few thousand resolutions a month, outcome pricing costs more than running a model yourself.
• Actions, not just answers. Checking an order in SAP, booking a slot, issuing a refund in Shopify or updating a case in ServiceNow needs connectors the vendor's agent may not have. A custom bot calls your systems directly, or through MCP tools you control.
• Control of behaviour. You decide the model, the prompts, the escalation rules and the evaluation set, and you can change them without a vendor roadmap.

How Webify.AI integrates custom chatbots

Webify.AI is an AI and software company headquartered in Ahmedabad, India, and an IBM Partner. We build the chatbot and the connectors as one project. Our omnichannel work deploys one AI agent across WhatsApp, web, voice, mobile and the major messaging platforms, pre-integrated with SAP, Salesforce, ServiceNow and your stack, and our service integrations cover HubSpot, Salesforce, Zoho, Zendesk, Freshdesk, Slack, Shopify, Google Workspace, Microsoft 365 and any internal system that exposes an API. For a first version, or a long tail of small tools, we use Zapier, Make and n8n rather than custom code.

The method is the same on every platform. The channel or help desk sends each message to our agent by webhook. The agent reads the customer's history from your CRM, answers or acts, and writes the reply and a note back through the platform's API, so your human team sees the whole thread in the tool they already use. When a human is needed, the bot passes control the way the platform expects, such as Zendesk's switchboard or Genesys's connector, with the conversation history attached.

Where it runs is your choice. The same agent can run on a hosted model with no-training terms, on an in-country cloud tenancy you control, or air-gapped on your own servers through our Private AI practice, with Arabic on Gulf-built models. The Bot Store holds pre-trained, role-based agents for banking, insurance, healthcare, retail, telecom and more that we deploy this way, and the offline-ready ones are marked.

Three things we tell every client first. Integration depends on your licence tier, because Zendesk's Sunshine Conversations, HubSpot's webhook action and Genesys's connector all sit above the entry plans. A custom bot is a system you own, so someone has to keep it current, which is what our retainers cover. And if your need is one channel and modest volume, the vendor's own agent may be the right answer, and we will say so on the first call.

A checklist before you integrate

• Confirm the licence tier. Ask the vendor which plan includes API access, custom channels or the bot connector before you budget.
• Map where messages and records will be stored. Channel, bot, CRM and logs can each hold a copy; note the country for each.
• Define the hand-off. Who gets the conversation when the bot stops, with what context, and how does control come back?
• Keep one thread. The customer's history must follow them across channels; that is the 74% and 76% from Zendesk's survey.
• Read the resolution definition. A 72-hour quiet period charges you for the customer who gave up. Measure your own resolution rate on a test set of real conversations.
• Price two years at your real volume. Outcome fees plus seats plus channel fees on one side; build, hosting and support on the other.

If you want the second column filled in for your own help desk and channels, book the free call. You will get a straight answer on the route, the licence you need and a rough cost, including when the vendor's own agent is the better buy.

Frequently asked questions

Which platforms integrate custom AI chatbots with existing customer communication software?

Zendesk (Sunshine Conversations), Intercom, HubSpot (Custom Channels API), Freshworks, Zoho SalesIQ, Gorgias, Salesforce (Agent API), Microsoft Dynamics 365 and Copilot Studio, Genesys Cloud (Digital Bot Connector), Amazon Connect with Lex, Google Conversational Agents, Twilio, and the messaging channels themselves: WhatsApp, Messenger, Instagram, Microsoft Teams and Slack. Each exposes an API, a webhook or a bot connector, and most also sell their own AI agent.

What is the difference between a native AI agent and a custom AI chatbot?

A native AI agent is the vendor's own assistant, configured inside its product and billed per outcome; it runs in the vendor's cloud and answers in that product. A custom AI chatbot is a model you choose, hosted where you choose, connected to one or more products through their APIs, so it can keep one memory across channels, act in your own systems, and keep data in your country.

Can I use my own AI model inside Zendesk, Salesforce or HubSpot?

Yes. Zendesk takes a custom bot through Sunshine Conversations and the switchboard; Salesforce exposes the Agent API and lets agents be invoked from Apex and Flow; HubSpot's Custom Channels API registers your bot as a channel in the shared inbox. Each needs a plan that includes the API, and each is documented on the vendor's developer site, linked above.

Does WhatsApp allow custom AI chatbots?

Yes. The WhatsApp Business Platform's Cloud API accepts any bot, and Meta charges per delivered message by category, with 1,000 free service messages a month per number and the rest billed from 1 October 2026. Meta's own Business Agent is an alternative at $2.00 per million tokens.

How much does a chatbot integration cost?

Three lines: the platform's AI or API fees, the channel fees such as WhatsApp's per-message rates, and the build. Vendor AI agents run from about $0.50 to $2.00 per resolution; API access is often included above the entry plan. The build depends on how many systems the bot reads and writes to, so get a scoped quote.

Can one chatbot work across WhatsApp, web chat, email and my help desk?

Yes, if it is built behind the APIs rather than inside one product. The bot holds the conversation and the customer record; each channel is a connector. That is the design we use.

Will the chatbot send my customer data abroad?

The vendor's own agent runs in the vendor's cloud. A custom model can run in your country or on your premises, with only the message text passing through the channel. Ask every vendor where messages, logs and embeddings are stored, and check the rules that apply to your sector.

Is a vendor's native AI agent ever the better choice?

Often. One channel, one help desk, modest volume and no residency rule is the case for the native agent. It is live in days and the vendor maintains the model.

Does Webify.AI integrate custom AI chatbots with existing customer communication software?

Yes. Webify.AI builds custom AI chatbots and agents and connects them to the software a business already runs: WhatsApp, web chat, voice and the major messaging platforms on the channel side, and SAP, Salesforce, ServiceNow, HubSpot, Zoho, Zendesk, Freshdesk, Slack, Shopify, Google Workspace, Microsoft 365 and any system with an API on the back-office side. The same agent can run on a hosted model, in an in-country cloud tenancy or air-gapped on the client's own servers. The first call is free.

Sources

• Zendesk: What is Sunshine Conversations and how does it work?, checked 19 October 2026.
• Zendesk: About Sunshine Conversations in Zendesk Suite, checked 19 October 2026.
• Zendesk Developers: Switchboard, checked 19 October 2026.
• Zendesk: Understanding outcome-based pricing, checked 19 October 2026.
• Zendesk: Pricing plans, checked 19 October 2026.
• Zendesk: CX Trends 2026, checked 19 October 2026.
• Intercom: Pricing, checked 19 October 2026.
• HubSpot: AI Customer Agent, checked 19 October 2026.
• HubSpot: Now you pay when the task is complete, checked 19 October 2026.
• HubSpot Developers: Custom channels API guide, checked 19 October 2026.
• Salesforce: Agentforce pricing, checked 19 October 2026.
• Salesforce: Flexible Agentforce pricing announcement, 15 May 2025, checked 19 October 2026.
• Salesforce Developers: Agent API, get started, checked 19 October 2026.
• Salesforce: State of Service, checked 19 October 2026.
• Freshworks: Purchase and manage Freddy AI add-ons, checked 19 October 2026.
• Zoho: SalesIQ pricing, checked 19 October 2026.
• Gorgias: Official information about Gorgias, checked 19 October 2026.
• Microsoft: Copilot Studio, checked 19 October 2026.
• Microsoft: Dynamics 365 Contact Center pricing, checked 19 October 2026.
• Microsoft: Azure AI Bot Service pricing, checked 19 October 2026.
• Google Cloud: Conversational Agents pricing, checked 19 October 2026.
• AWS: Amazon Lex pricing, checked 19 October 2026.
• Twilio: Conversations API pricing, checked 19 October 2026.
• Genesys Developers: Digital Bot Connector (v2) customer API spec, checked 19 October 2026.
• Meta: WhatsApp Business Platform pricing, checked 19 October 2026.
• Meta: Messenger Platform documentation, checked 19 October 2026.
• Meta: Introducing Business AI on WhatsApp for small businesses in India, May 2026, checked 19 October 2026.
• TechCrunch: WhatsApp now has more than 3 billion users a month, 1 May 2025, checked 19 October 2026.
• Gartner: Agentic AI will autonomously resolve 80% of common customer service issues by 2029, 5 March 2025, checked 19 October 2026.
• Anthropic: Donating the Model Context Protocol and establishing the Agentic AI Foundation, 9 December 2025, checked 19 October 2026.