GoHighLevel in October: conversation search, number pools, Git. What touches the integration

In October HighLevel shipped several updates that touch conversations, calls, reviews and code. I picked seven and look at them from one position: at the PDR company in the US, GoHighLevel is connected to the super app, the website and telephony, and I check every CRM update against that connection. For the owner the question is what switches on by itself and changes the reports, and what needs a decision.

Search, reviews and dashboards

HighLevel on October 26 added keyword search in Conversations: across the whole inbox and inside a single conversation. Matches are highlighted, and inside a thread there is a counter and arrows between them. There are filters by period, channel, owner and attachments. HighLevel writes that this was one of the most requested features on their Ideas Board.

The same day brought 16 reputation widgets for location dashboards: review requests, reviews by rating and sentiment, response speed, and you can see separately whether a person or AI replied. And the AI agents that reply to reviews got brand voice settings (“we” or “I”, tone, length, signature), guardrails for risky wording (refund promises, guarantees, admitting fault), holding sensitive reviews for a human to check, and a preview of replies before the settings are saved.

Social Intelligence has been in limited beta since October 26: brand mentions, industry trends, comparison with three competitors and alerts. For the beta period each location gets 100 searches, up to 50 tracked topics and 3 alerts per topic.

Number pools and code

Two entries are dated October 6. The first: calls from number pools now get the right ad platform. Before, a call from a Facebook ad showed as “Facebook” or “Direct”; now it is “Facebook Paid”, and TikTok, LinkedIn and Microsoft (Bing) are labeled the same way, while links with utm_medium=cpc get the platform plus “Paid”. Visits from ChatGPT and similar tools are split into a separate source. The rule applies only to new calls; old ones keep their previous source. The second: a call from a number pool can be sent to several employees at once, up to 6 people, and each can answer from the web, from the app, from a desk phone or from their own number.

AI Studio on October 15 got GitHub sync: a private repository is created for the project, every saved version goes into it, and commits from outside, from any editor, come back into the project. Pushes are checked for secrets: AWS, GitHub, Stripe and OpenAI keys are blocked. There is a new separate permission, “Manage Git & download code”, enabled for all roles by default.

What touches the integration

Take a service company that runs ads on Google and Facebook and counts calls through number pools. The integration writes every call into its database with a “source” field. After October 6 new values will start arriving in that field: “Facebook Paid”, “Bing Paid”, a separate source for ChatGPT. If the summary counts by exact string match, the new calls drop out of the Facebook report. The conclusion: when the CRM fills a lookup field by itself, the integration has to check it for unknown values and put them in a separate row.

Reviews are a similar story. If an agent writes the replies, I need a log: which review it answered on its own, which it held, and which person released it. The widgets show this inside the HighLevel dashboard, but for the owner's annual report you need your own database.

At our company

At the PDR company in the US I have been the technical partner since 2021: server, API, telephony, document parsers. In 2022 I wrote the integration between the sales team's AppSheet app and GoHighLevel, in 2023 I kept improving it all year and connected the website forms to the CRM and the app. Now, after a call to the IVR, the system creates a lead in the CRM on its own and attaches the recording and an AI breakdown of the call: who called, a lead, an existing customer or an insurer, and how the employee answered. The center of the system moved to the super app, and GoHighLevel went from the main CRM to a supporting one.

Our telephony runs on our own PBX, so I haven't tested number pools or the call to 6 employees myself. I haven't tried AI Studio's Git sync, but the principle is close to mine: the code, the database, the server and the access credentials belong to the client, and a repository on GitHub gives another developer a copy of the code.

What I would do

Export the values of the “call source” field for October and compare them with September. Add the new values to your lookup table and leave the old calls alone: HighLevel doesn't rewrite them.

If an agent replies to reviews, turn on holding sensitive reviews and the preview, and assign a person who goes through the held ones every morning.

Conversation search needs no integration, but show it to the team: they will look up a price quoted to a customer by the amount.

If your website was built in AI Studio, connect GitHub and check that the “Manage Git & download code” permission is on your own account as well as the contractor's.

I wouldn't build Social Intelligence into a process while it is in beta: 100 searches per location will run out fast.

I wrote about connecting a CRM with documents and tasks in the article how to reliably connect documents, CRM and tasks. Directions: a CRM connected to the team's work and IP telephony connected to business processes. The full project: technical partner to a PDR company in the US, and the funnel and lead plan for GoHighLevel is on the page GoHighLevel: a plan for the funnel, leads and communications. Neighboring reviews: API versions and shutdown dates and who called and why: AI call analysis. Earlier I wrote about who owns the company context in AI for CRM. If your CRM updates faster than you can check the integration, tell me what is connected to what. Discuss your task.