AI Agents in Your CRM: What Changes, and What Yours Must Do

HubSpot has started building AI agents directly into the CRM — not as a bolt-on chatbot, but wired into each customer’s real data — and Salesforce has been on the same path for months with Agentforce. For a small or mid-sized business running sales or support through a CRM, this turns what used to be a purely technical question — “do I wire an AI assistant into my CRM?” — into a legal one too: if that agent talks directly to your customers, the AI Act’s transparency rules that AESIA can now fine you for come into play. Here’s what changed, and what to check before you switch it on.

What HubSpot just launched

On July 23, 2026, HubSpot released Agent Hub and Agent Builder in public beta, as reported in HubSpot’s official announcement and confirmed by MarTech Notes. Agent Hub is a single console where a company sees every active AI agent, grouped by goal — building demand, winning deals, supporting customers, driving retention — with live status and results. Agent Builder lets someone describe in plain language what an agent should do, drawing on the data already sitting in the CRM and defining when it should hand off to a person. It’s available on Professional and Enterprise plans during the beta, and custom agents run on HubSpot Credits every time they take an action.

Salesforce is on the same track

This isn’t an isolated move: Salesforce offers comparable capabilities through Agentforce, including AI-driven call coaching that gives sales reps feedback on their pitch, according to its own product page. The direction is clear: the CRM stops being just a customer database and becomes the place where the agent that negotiates, replies, or follows up on your behalf actually lives.

The detail almost nobody’s checking: Article 50 transparency

Here’s what a lot of SMEs are going to miss when they switch these agents on: if the agent talks directly to a customer — over WhatsApp, web chat, or an automated email triggered from the CRM — it must disclose that it’s an AI, unless that’s already obvious, exactly as we covered when AESIA started enforcing Article 50. A sales agent that negotiates terms or handles a complaint without that disclosure isn’t just bad practice — it’s a sanctionable transparency breach, and the fact that HubSpot or Salesforce turned it on by default doesn’t shift that responsibility to the vendor.

When the CRM’s built-in agent is enough, and when you need something custom

For most SMEs with a standard sales volume, switching on the agent your CRM already ships with — paying by credits — is the fastest, most reasonable option, as we already covered when comparing a custom CRM against Salesforce or HubSpot. The math changes when:

• Conversation volume is high enough that per-credit costs start outrunning the cost of building it yourself.
• The agent needs to reach systems the CRM doesn’t connect to natively — an internal ERP, custom invoicing, a catalog with complex pricing rules.
• The business needs fine-grained control over what the agent decides on its own versus what it hands off to a person, beyond what the standard settings allow.

What to ask before turning on an agent in your CRM

1. What tells the customer they’re talking to an AI, and at exactly what point in the conversation.
2. What gets logged for every action the agent takes — an offer sent, a discount applied, a ticket closed — and who can see it.
3. What it actually costs at real volume, not in the demo: credit pricing scales with usage, and it’s worth projecting against your own CRM’s numbers.
4. What happens when the agent gets it wrong: who reviews it, how it’s corrected, and whether there’s an auditable trail of the error.

How we look at this at Evicron

At Evicron, an AI and custom software studio based in Barcelona, we help businesses make this call without a stake in the outcome: sometimes the right move is switching on and properly configuring the agent your CRM already has; sometimes it’s building something custom because the standard option doesn’t match how you actually sell. In our AI consulting service, we always start by auditing what the agent needs to touch and at what level of supervision, before recommending a platform or a custom build.

The bottom line

HubSpot has brought AI agents inside the CRM with Agent Hub and Agent Builder, Salesforce was already there with Agentforce, and the trend is only growing. Before you switch one on, check that your agent meets Article 50 transparency if it talks to customers, work out the real cost at your volume of credits, and decide whether the standard option fits or you need something custom.

Want us to check whether your CRM’s agent is configured correctly, or whether a custom build would pay off? Get in touch: the first discovery session is free, and we reply within 24 hours.