Integrate HubSpot with a WhatsApp AI Agent: Step by Step
A step-by-step guide to integrating HubSpot with an AI agent on WhatsApp: what to prepare, how to map properties per agent, what the AI does not fill in, and how to test.

Marlos Carmo
Founder of Tolky
Updated on November 16, 2026

Share
Anyone who serves customers on WhatsApp and uses HubSpot as a CRM tends to live the same rework: the conversation happens in one place and the record in another. Someone copies the outcome of the interaction into the ticket, and the history ends up incomplete.
Integrating HubSpot with an AI agent reduces that copying: the conversation generates the record, within rules you define. This step-by-step guide shows how to think about and set up that integration at Tolky, and where it ends.
Updated in November 2026.
What does it mean to integrate HubSpot with an AI agent?
Integrating HubSpot with an AI agent means that the WhatsApp conversation starts creating and updating HubSpot records, with the fields you mapped, and that the agent can use the customer's context instead of starting from zero. The integration does not replace HubSpot: it remains your CRM, and Tolky handles the conversation.
Tolky has its own CRM and a native integration with HubSpot. It is the only native CRM integration; for other tools, such as Pipedrive, Slack, Gmail, Google Sheets, Stripe and Notion, there are connectors. Tolky has no native integration with Salesforce or Pipefy. For the big picture, see AI integration with CRM in companies and AI without integration becomes an FAQ.
What does the HubSpot integration do at Tolky?
At Tolky, the integration maps properties between the agent's conversations and HubSpot objects, configured per agent, and syncs tickets created in the conversation. The table summarizes what is confirmed and what is left to agree on.
| Capability | Status | Note |
|---|---|---|
| Property mapping per agent | Confirmed | Each agent has its own mapping |
| Mappable objects | Deal, ticket, contact and custom object | According to your account setup |
| Pipeline | Confirmed | The mapping is tied to a HubSpot pipeline |
| Fields the AI does not fill | Confirmed | You choose field by field |
| Ticket sync | Confirmed, optional | When a ticket is created in the conversation, it is synced |
| HubSpot required fields | Confirmed | The screen warns when contact data is missing |
| Authentication details and API limits | To agree with the team | Not described here |
Tolky also shows the HubSpot ID in contact information, which helps confirm that both sides are talking about the same customer.
What should you prepare before starting?
Before starting, prepare HubSpot: pipelines, required fields and who owns the records. Most integration problems come from a disorganized HubSpot, not from the integration itself.
- Pipelines and stages: define which pipeline receives records coming from WhatsApp.
- Properties: list the ones the conversation may fill in, such as reason for contact, product of interest or priority.
- Required fields: if the ticket requires a field, the conversation must collect it or HubSpot rejects the record.
- Owner: decide who owns the records and how exceptions are handled.
- Access: who on your team can view and edit the integration at Tolky (the screen has its own view and edit permissions).
On the conversation side, the AI needs a reviewed knowledge base: see how to train an AI agent on your knowledge base.
What are the steps of the integration?
There are seven steps, from design to monitoring. Initial setup is done with the Tolky team, because the integration screen only appears for accounts that already have a mapping configured.
Step 1: define what the conversation should generate in HubSpot
Choose a few clear outcomes, such as "create a support ticket" or "record the reason for contact". The simpler the first goal, the easier to test. Write in one sentence what should happen and in which HubSpot object.
Step 2: organize HubSpot
Review the destination pipeline, its stages and the properties that will be used. Remove duplicate fields and name clearly what the AI will fill in. Note which properties are required to create the object you chose.
Step 3: connect the HubSpot account
Agree the connection of your HubSpot account with the Tolky team. The account is linked to your operation, and authentication details vary by account and are handled at that point, not in this post.
Step 4: map the properties per agent
Each agent has its own mapping: you choose the object (deal, ticket, contact or custom), the pipeline and the properties. For each property, describe what it means and, if there are fixed values, list the options. A clear description helps the AI fill the right field with the right value. A support agent and a sales agent usually have different mappings.
Step 5: define what the AI does not fill in
For sensitive fields or ones that depend on human decision, such as pipeline stage or owner, mark that the AI should not fill them. This is your risk control. When in doubt, leave the field for a person; you can release it later, once you trust it. The post AI governance in customer service covers this logic.
Step 6: decide on ticket sync
If your team opens tickets during the conversation, turn on sync so that a ticket created in the conversation is synced with HubSpot. If HubSpot requires fields, complete them in the mapping; the Tolky screen warns when contact data is missing.
Step 7: test in simulation, publish and monitor
Before publishing, run simulations with real situations: a simple request, a complaint and a case with missing data. Check in HubSpot that the record appeared in the right object, pipeline and with the expected fields. Publish to a small group, monitor for a few days and only then expand.
Keep the first version of the mapping small on purpose. A mapping with three or four well-described properties is easier to test, easier to explain to the sales and support teams and easier to correct than one with thirty fields. Add properties only when the previous ones are returning clean values, and record each change so you know what altered the result.
Which mistakes get in the way most?
The most common mistakes are required fields not collected, poorly described properties, and the AI filling in what should be human. The table links each to its fix.
| Symptom | Likely cause | Fix |
|---|---|---|
| The ticket does not appear in HubSpot | Required field without a value | Collect the data in the conversation or fill it in the mapping |
| Wrong value in a field | Vague property description | Rewrite the description and list the allowed options |
| Duplicate record | Repeated mapping across agents | Review which agent creates each object |
| AI touching the pipeline stage | Field not marked as "no AI" | Mark the field so the AI does not fill it |
| Contact without data | The conversation did not collect it | Adjust the form or the agent's question |
How does the integration compare with a connector?
The native integration is designed for HubSpot and allows mapping per agent; connectors serve other tools, such as Pipedrive, Slack, Gmail, Google Sheets, Stripe and Notion. What each connector does varies by tool; confirm with the Tolky team before depending on a specific action.
If you do not use HubSpot yet, Tolky also works as its own CRM, with pipeline, opportunities and activities. If you are weighing keeping HubSpot or switching, see the comparison between Tolky and HubSpot and the conversational CRM page.
How do you monitor the integration after publishing?
After publishing, monitor the integration by sampling: open a few records created from conversations and check that the object, pipeline and fields are as expected. The routine is short and keeps a small error from repeating for weeks.
A weekly check, at the start, is usually enough:
- Pick five conversations that generated a record and compare the result with what the customer said.
- Look for fields that were frequently left empty. Usually the agent's question needs to change or the property is not relevant.
- Read the conversations where the AI handed off to a person and see whether a "no AI" field could be released, or the reverse.
- Rewrite the descriptions of properties that produced a wrong value.
- Review who has permission to view and edit the integration at Tolky.
Tell the sales or support team what now reaches HubSpot automatically, so nobody duplicates the record by hand. To measure the effect on service, see AI resolution rate: how to measure it. A good sign is the team no longer copying data between systems and spending its time on what needs human conversation.
Where does Tolky fit, and what does it not do?
Tolky is the conversation layer: AI agents on WhatsApp that serve customers, collect data and record the outcome in HubSpot within the rules you set. It does not replace HubSpot and does not promise to fill every field without review.
It is also worth saying what the integration does not promise: we do not claim here two-way sync of every object or reading any HubSpot property during the conversation. For your specific case, ask for a demo on your HubSpot. The platform aggregates more than 5 million AI messages per month and more than 67% AI resolution, platform figures, not a customer's result. See the WhatsApp support solution.
Frequently asked questions
Does Tolky have a native HubSpot integration?
Yes. HubSpot is Tolky's only native CRM integration. For other tools, there are connectors.
Can I map different fields for each agent?
Yes. The mapping is per agent, so a support agent and a sales agent can use different objects, pipelines and properties.
Which HubSpot objects go into the mapping?
Deal, ticket, contact and custom object, according to your account setup. Confirm with the team which ones make sense for your flow.
Can I stop the AI from filling in certain fields?
Yes. You mark, field by field, what the AI should not fill, and leave those fields for a person.
Does a ticket created in the conversation go to HubSpot?
If ticket sync is on, yes: when a ticket is created in the conversation, it is synced with HubSpot. Sync is an option you turn on or off.
Do I need HubSpot to use Tolky?
No. Tolky has its own CRM and can work on its own or alongside HubSpot.
Conclusion
Integrating HubSpot with an AI agent on WhatsApp works best when you start small: one agent, one object, few fields, simulation before publishing. Tolky delivers the native integration and per-agent mapping; HubSpot remains your CRM. Shall we talk?
See the plans, explore Tolky's CRM or talk to the team →
Share
Tags
- integrate HubSpot with a WhatsApp AI agent
- HubSpot WhatsApp integration
- HubSpot and conversational AI
- #conversational AI CRM
- automatic HubSpot tickets
- HubSpot property mapping
- CRM integrated with WhatsApp
- native HubSpot integration
About the author

Marlos Carmo
Founder of Tolky
Marlos Carmo is an AI entrepreneur and founder of Tolky, the conversational-era infrastructure and AI CRM that unifies intelligent service, multi-channel support (such as WhatsApp and voice), live CRM, and operational intelligence in a single ecosystem. He is a finalist for the SXSW Innovation Awards and a member of Francesco's Economy, a global network of young entrepreneurs focused on innovation and social impact. He works connecting Artificial Intelligence and digital transformation in projects for large organizations.
Next step
Want to bring this to your operation?
Tati, Tolky's AI assistant, understands your scenario in a few questions, suggests the best path and connects you with someone from the team, with full context.


