GHL's AI Agent Action Now Runs on Claude. Here's Why That's Bigger Than It Sounds.

Two updates, one pattern

GoHighLevel shipped two changes this week worth breaking down, because both point at the same thing: the platform is quietly getting harder to outgrow. Neither is a redesign. Both remove a specific piece of friction that used to push people outside the CRM to get real work done. That's the whole argument behind stack discipline, so it's worth looking at closely.

One. AI Agent Action now runs on Claude

Inside Workflows, the AI Agent Action can now run on Anthropic models (Claude Opus 5, Sonnet 5, Haiku 4.5) alongside Google and OpenAI, with a redesigned model picker and a per-action reasoning-effort dial (Low, Medium, High). Stacked on prompt caching and leaner system prompts shipped earlier this month, the same task now runs up to 50% cheaper with no behavior change.

Here's why that matters past the changelog entry. BAMF already standardizes on Claude for anything that requires judgment, not just autocomplete. Until now, a GHL workflow agent meant accepting whatever model GHL defaulted to, or building the logic somewhere else and piping it back in. That gap is closed. You can now spec an AI Agent Action inside a native GHL workflow, running on the same model family BAMF builds everything else on, and dial the reasoning effort against the cost of the task. A quick lead-qualification check doesn't need the same reasoning budget as a contract summary. Now you can tune that inside the platform instead of working around it.

For anyone running or reselling AI agents inside GHL, this is a straightforward upgrade: better model, lower cost, no new tool required. It also strengthens a specific build pattern, a managed AI agent living entirely inside the CRM workflow layer, with no external orchestration layer bolted on. Fewer moving pieces, one place to debug, one login to check.

Two. Calendar AI builds the calendar for you

A new guided flow lets you describe the calendar you need in plain language, team members, availability, meeting location, and Calendar AI recommends and builds it, with a live preview and a test booking before you commit. No more picking a calendar type from a dropdown and hoping you configured it right.

This sounds small until you've watched someone get stuck on calendar setup, which happens constantly. It's one of the more common places a DIY-tier CRM user gets stuck and gives up. Calendar AI removes that friction the same way Ask AI's guided templates already do elsewhere in the platform: describe the outcome, let the system build the steps. The CRM walks you through it instead of you needing to already know how GHL works.

The actual point

Neither update is really about AI or calendars specifically. Both are the platform quietly removing a reason to leave it, or a reason to give up before you've built anything. That's the whole argument behind stack discipline: go one platform deep before you add another, because every extra tool is a login nobody checks and a workflow that can silently break while you're not looking.

Most businesses don't have a marketing problem. They have a systems problem. Updates like these don't fix a broken foundation, but they reward the businesses that already built one. If your CRM, your AI agents, and your calendars all live in one system, every improvement GHL ships compounds automatically. If they're stitched across five tools, you're rebuilding the connection every time something changes.

If you want the foundation built once and built right, that's what BAMF 75 walks you through, free.

Or book a strategy call if you'd rather have it installed than explained.

This post was drafted by Sapper, BAMF's engineering AI agent, from live GoHighLevel changelog data. AI-assisted, reviewed by Justin before publishing, per standing BAMF disclosure policy.