AI CRM in 2026: built-in AI or your own assistant?
An AI CRM is a customer relationship manager where AI does part of the work: summarizing deals, drafting follow-ups, updating records and suggesting what to do next. In 2026 there are two kinds. Some CRMs build AI agents into the product and charge for usage. Others let the assistant you already pay for, such as Claude or ChatGPT, work inside the CRM through MCP.
As of October 2026. Vendor details come from the official pages linked below and can change.
What counts as an AI CRM?
Almost every CRM now has an AI button. That alone does not make it useful. The question that matters for a small business is: who does the work, and who pays for it?
There are two models.
Model 1: AI built into the CRM. The vendor runs its own agents inside the product. You turn them on, they act on your records, and the vendor meters the usage. HubSpot, Salesforce and Pipedrive all work this way to some degree.
Model 2: the CRM your assistant works in. The CRM exposes its records through the Model Context Protocol (MCP), an open standard for connecting AI apps to other systems. You connect Claude, ChatGPT or another assistant, give it permission, and it reads and updates the CRM on the plan you already have. This is sometimes called an agent CRM. If MCP is new to you, start with what MCP is.
How do CRM vendors charge for built-in AI?
CRM vendors usually charge for built-in AI through credits or add-ons on top of seats.
Built-in AI usually means credits or add-ons on top of seats. Three examples, all from the vendors' own pages:
- HubSpot. HubSpot says on its credits page that all paid plans include monthly credits, extra credits cost $10 per 1,000, and unused credits expire at the end of the month. Its prospecting agent uses 100 credits per lead and its data agent uses 10 credits per response. HubSpot's Agent Hub page describes marketing, sales and service agents built on CRM data.
- Salesforce. The Agentforce pricing page lists Flex Credits at $500 per 100,000 credits, and Agentforce add-ons for Sales and Service at $125 per user per month for unmetered employee use.
- Pipedrive. Pipedrive's pricing page shows AI report creation from the Lite plan (US$14 per seat per month, billed annually) and AI email writing and summaries from Premium (US$59 per seat per month, billed annually). Its Sales Assistant summarizes deals and predicts win probability.
None of this is bad. It is just a different bill: you pay per seat for the CRM, then again for the AI that works in it.
What each model is good for
| Question | Built-in AI CRM | CRM your own assistant works in (MCP) |
|---|---|---|
| Who runs the AI? | The CRM vendor | Your Claude, ChatGPT or other assistant |
| How is AI billed? | Credits, add-ons or higher tiers | Your existing assistant plan |
| Setup | Turn features on inside the CRM | Connect the assistant once with sign-in (OAuth) |
| Best at | Packaged jobs: lead scoring, support bots, email sequences at volume | Open-ended work: planning, drafting, research, updating records as you talk |
| Context the AI sees | What is in that CRM | The CRM plus whatever else you connect to the assistant |
| Control | Vendor settings and spend caps | Per-assistant access, read-only or update, your review |
| Switching assistants | Not possible; the AI is the vendor's | Connect a different assistant to the same records |
What should a small business check before buying an AI CRM?
A small business should check AI costs, spending limits, assistant access, permissions, review controls, edit history, delivery context and data export before choosing an AI CRM.
Use these questions on any vendor, including us.
- What does the AI cost after the trial? Find the per-credit or per-seat price on the vendor's own page, and check whether unused credits roll over.
- Can I set a hard spend limit? Ask for account-level and feature-level caps.
- Can I use the assistant I already pay for? Look for a documented MCP server or connector, not just an "AI" label.
- What can the AI change? Read-only access for summaries and update access for edits should be separate choices.
- Do I review before it changes things? A review queue or approval step matters more than the model behind it.
- Can I see who changed what? Every AI edit should be recorded, with an undo where possible.
- Does the AI see the work, not just the deal? For service businesses the delivery side (projects, tasks, approvals) is half the customer record.
- What happens when I leave? Check that you can export your data.
When is a CRM with built-in AI the better choice?
A CRM with built-in AI fits high-volume sales teams, businesses needing customer-facing bots and teams that rarely use separate assistants.
Be honest about this. A built-in AI CRM is the better pick when:
- You have a sales team at volume. Hundreds of leads a month, sequences and routing rules are where packaged agents earn their cost.
- You need a customer-facing bot. Support agents that answer customers on your site are a vendor product, not something you assemble from a chat assistant.
- Nobody on the team uses an assistant daily. If the AI has to show up inside the tool or it will not be used, built-in wins.
When should a small business use Claude or ChatGPT with its CRM?
Using Claude or ChatGPT with a CRM suits small businesses that already use an assistant daily for call prep, proposals, follow-ups and project planning.
Bringing your own assistant tends to fit a 1-20 person business when:
- You already pay for Claude or ChatGPT and use it every day.
- Most of your "CRM work" is thinking and writing: call prep, proposals, follow-ups, project plans.
- You want one place for deals and the client work that follows, so the assistant does not plan a project without knowing what was sold.
- You want to try a different assistant later without moving your data.
The trade-off: you set up the connection, you write good prompts, and assistant plans have their own usage limits. Your assistant is also only as good as the records it can reach.
How Flori fits
Flori is a CRM with project management for freelancers, small agencies and small teams, built for Model 2. It has accounts, people, deals, leads and a deal pipeline board next to projects, tasks, docs and a client page in one workspace.
There are no Flori AI credits. You connect Claude, ChatGPT, Grok, Cursor, Codex or Gemini CLI over remote MCP with browser sign-in, and choose per workspace or per project whether each assistant gets read-only or update access. Tasks marked Ready for AI are reserved by one assistant at a time to reduce overlapping work among assistants that honor the claim. Submitted work can wait in a review queue; authorized direct updates can save immediately. Saves record the assistant that made them. In-app drafting runs on your own Anthropic or xAI key with spending limits.
Flori does not send invoices, make calls or run email sequences at volume, so a sales-heavy team should look at a dedicated sales CRM. See how the CRM with project management works, the setup in the MCP project management guide, or compare with HubSpot and Pipedrive. Current status is on the pricing page.
FAQ
What is an AI CRM? An AI CRM is customer relationship software where AI handles part of the work, such as summarizing deals, drafting follow-ups, enriching records or suggesting next steps. Some vendors build their own agents into the CRM and bill by credits or add-ons. Others let your own assistant, such as Claude or ChatGPT, work in the CRM through MCP.
What is an agent CRM? An agent CRM is a CRM designed for AI agents to read and update, not only for people to click through. In practice it exposes its records through a standard like MCP, controls what each agent may change, and records which agent made each edit so a person can review and undo the work.
What is the best AI CRM for a small business? It depends on who sells. A team with dedicated salespeople and high lead volume usually gets more from a sales CRM with built-in agents. A freelancer or small agency that already uses Claude or ChatGPT daily often does better with a CRM that assistant can work in, so it does not pay twice for AI.
Do AI CRM features cost extra? Often, yes. As of October 2026, HubSpot meters AI agents with credits at $10 per 1,000 beyond the included amount, and Salesforce sells Agentforce through Flex Credits and per-user add-ons. Check each vendor's own pricing page, because included amounts vary by plan and change often.
Is it safe to let ChatGPT or Claude update my CRM? It can be, with controls. Give read-only access until you trust the setup, use update access only for the projects you choose, keep the assistant's tool approvals on, and use a CRM that records every AI edit and lets you review or undo it. Never connect an assistant you would not trust with the same records.
Start your workspace