Skip to content
ConsultEvo

HighLevel Release Radar: Current Status and Safe Setup

As of November 22, 2026, four operational features from HighLevel’s September 14-18 Release Radar still have current support documentation: Ask AI Workflow generation, invoking published Managed Agents from Workflows, Company creation and enrichment, and Local Presence Dialing. The announced $20 AI Assistant Plan was deprecated on November 2, 2026, and is not a current offer.

That distinction matters in practice. Ask AI can draft a support-routing Workflow, but the Workflow owner still needs to inspect it in the builder, test its conditions and exception path, and publish it. This guide separates documented product behavior from recommended setup controls. The September Release Radar is historical context, not a current feature or pricing reference.

HighLevel’s support pages and account navigation can change. Check the current Ask AI Workflow documentation and the target account when labels or availability differ.

Which September 2026 HighLevel updates are still relevant?

The September announcement covered five updates, but the current implementation picture is different. Ask AI Workflow generation, Managed Agent invocation, Company automation, and Local Presence Dialing remain documented operational features. The $20 AI Assistant Plan does not: its retirement notice supersedes the September pricing description.

This is a practical setup guide, not a claim that every feature is available in every HighLevel account. Confirm permissions, plan scope, supported markets, phone inventory, and account-specific navigation before changing a production process.

A generated draft is not a published Workflow, and an AI response is not a validated CRM decision.

Generate a Workflow draft with Ask AI, then review it in the builder

HighLevel documents a sequence in which Ask AI generates a Workflow from a natural-language request, displays a live preview, accepts conversational refinements, and opens the draft in Workflow Builder. Manual edits, testing, and publication happen in the builder, not in the preview. HighLevel also documents starting Workflow generation through Ask AI on WhatsApp.

For a support-intake example, make the business rules explicit rather than asking AI to infer them. A hypothetical prompt could specify that a new inbound support request must contain a Contact ID and message text; no outbound message may be sent without consent; billing requests go to the billing queue; and missing or ambiguous records go to manual review. Include the intended timing, time zone, duplicate-enrollment policy, and suppression conditions.

01Describe the requirementName the business event, required fields, fixed conditions, timing, and exception destination. The Workflow owner supplies and checks the business rules.
02Inspect and refine the previewCheck the proposed trigger and actions, then request conversational changes. Treat the preview as a draft, not as an editable production Workflow.
03Open it in Workflow BuilderThe owner makes required manual edits in the builder and confirms that the referenced fields, queues, messages, and destinations exist in the target account.
04Test, then publishTest normal, missing-data, duplicate-entry, and exception cases. The Workflow owner publishes only after checking the results.

Before publication, verify trigger scope, Contact and Company field mappings, wait timing and time zone, repeat enrollment, suppression, and the route for missing data. A duplicate test should confirm whether repeated events can start a second run and whether the intended conditions prevent unwanted follow-up. Keep consent, country, and lifecycle rules as explicit Workflow conditions.

For account setup or a broader configuration review, see ConsultEvo’s GoHighLevel setup service.

Use Managed Agents for interpretation, not as the Workflow’s control plane

The documented Invoke Managed Agent Workflow action calls a published Managed Agent. Configure the agent with the “Invoked from a workflow” trigger, test and publish it, then add the action in Automation → Workflows. Select the agent, provide appropriate instructions and dynamic context, and use its response in later Workflow steps. Contact data and attachments can be supplied where needed.

This action is distinct from an AI Agent configured directly in a Workflow and from Invoke Agent Studio Agent. Choose according to where the AI configuration is maintained. The support documentation confirms downstream use of the response, but it does not prescribe one universal structured schema, confidence field, retry policy, or approval mechanism.

For a hypothetical support-message flow, an inbound message starts a Workflow. The Workflow passes the selected Contact and message context to an agent that classifies the issue. The Workflow, not the agent, enforces fixed routing rules: accept only billing, support, sales, or other, then branch to the corresponding queue. A missing, unsupported, or review-required classification goes to a human review queue instead of changing a lifecycle stage or sending a customer-facing message.

The following is a proposed design contract, not a HighLevel-guaranteed response schema:

{
  "classification": "billing",
  "confidence": 0.82,
  "review_required": false,
  "rationale": "The message asks about an invoice."
}

Before using a response, confirm that it is available in the actual Workflow context and that any fields used for branching are parseable there. Check that the classification is allow-listed and apply a team-defined confidence policy only if a usable confidence value is actually returned. If the result is missing, malformed, or uncertain, route it to the named support operations owner. Do not write free-form output into authoritative CRM fields. Limit agent tools, Knowledge Base access, CRM permissions, and attachment access to what the task needs. For help designing a bounded agent role and review path, see AI agent implementation services.

Treat Company matching as a CRM data-quality rule

Current documentation places Company automation at Settings → Objects → Companies → Automations. Older Release Radar instructions mention a Labs step; use the current support article and account interface for setup. HighLevel can match Contacts to Companies by Business Name, business email domain, or both, and can create a Company when an appropriate match is not found. Enrichment may add a domain, name, or logo when data is available, but it is best effort.

Review these rules before enabling the automation:

  • Free email providers, including Gmail, Yahoo, Outlook, and iCloud, are excluded from email-domain Company creation.
  • Domain matching is exact. For example, acme.com and mail.acme.com are different domains. A Company stores one domain.
  • If both Business Name and domain matching are enabled, Business Name matching can take precedence. Decide which identity rule should govern your records.
  • Existing Contact-to-Company associations are not automatically moved. Duplicate prevention can prevent future duplicate Company domains, but it does not merge duplicates already in the account.
  • Company creation can succeed even when enrichment cannot identify a name or logo. Disabling automation does not undo records already created.
Decision point

Duplicate prevention is not historical cleanup, and exact matching does not make a subdomain equivalent to its parent. Review existing Company domains and duplicate records before enabling broad processing; send conflicts to a CRM data owner rather than assuming the automation will reconcile them.

A practical preflight is to review internal, test, partner, and shared domains; decide whether subdomains should remain distinct; inspect existing duplicate Companies; and determine whether Business Name matching should run alongside domain matching. A CRM data steward should resolve ambiguous matches and historical associations. For help setting ownership and data-quality rules, see CRM systems consulting.

Configure Local Presence Dialing with realistic expectations

Local Presence Dialing selects a geographically relevant caller ID for outbound calls when a suitable number is available. HighLevel documents location-based matching in the United States, Canada, the United Kingdom, and Australia. Matching can consider postal code, area code, overlay code, nearest available number, same region, and then the account default. The exact selection depends on Contact location data and the sub-account’s available numbers.

Configure the number pool at the sub-account level under the Phone System’s outbound-call settings. The documented setup path is Settings → Phone System → Voice → Other Settings → Outbound calls. Review the available numbers and mark up to 50 numbers as reserved if they should not be selected automatically. Reserved numbers remain available for manual dialing. Local Presence works with the Web Dialer, HighLevel mobile app, and desk phones; a caller can also override caller ID for an individual call.

Repeated outbound attempts to the same Contact use the same selected Local Presence number. If HighLevel cannot find a suitable local match, it uses the account default so the call can continue. HighLevel does not guarantee a higher answer rate, so evaluate the feature as caller-ID selection, not as a call-performance promise.

The $20 AI Assistant Plan is retired

HighLevel deprecated the $20 AI Assistant Plan on November 2, 2026. Existing subscriptions continue only through their current billing cycle and cannot be renewed or repurchased. The retirement notice describes agency users and admins as receiving AI Assistant access through Free AI, subject to fair-use limits; AI Employee subscriptions for sub-accounts are separate.

For current guardrails, use HighLevel’s retirement notice and AI usage-limit documentation, not September’s plan description. Limits can be configured at agency, sub-account, and user levels, with notification or blocking behavior depending on configuration. The support article says no limit is configured by default. Before rollout, check current billing in the account, choose whether reaching a limit should notify or block, and test the behavior with a controlled user or sub-account.

Choose the right control for each update

Update or trigger System or AI responsibility Validation gate Action or fallback
Ask AI Workflow draft Generates and refines a draft Owner checks rules, fields, timing, and test cases Workflow Builder; owner publishes
Managed Agent classification Interprets a message within a bounded task Check allowed value and usable response fields Workflow branch or human review queue
Company association Applies native name or exact-domain matching Review exclusions, conflicts, and existing duplicates Company record or CRM owner
Local Presence call Selects a number using location and availability Check number pool, Contact location, and default Outbound call or account default

Deterministic rules are the better control for fixed conditions such as consent, lifecycle stage, an exact domain, or a permitted classification list. Use AI for bounded interpretation or summarization, then validate its result before consequential CRM changes. Use HighLevel’s native Company matching or caller-ID selection when its documented behavior fits the need.

The reviewed sources do not establish a universal API or export schema, stable event ID, retry contract, exactly-once execution, or transactional rollback for these features. If a separately verified integration stores events externally, define the data grain first: one Managed Agent record per invocation, one Company association per Contact-to-Company decision, and one Local Presence record per call attempt. A proposed Managed Agent uniqueness key could combine the sub-account, verified Workflow execution identifier, agent identifier or version, Contact identifier, and invocation occurrence, but these fields must not be assumed to exist in HighLevel.

For Company data, a domain-only key is insufficient when records are partitioned across sub-accounts or agencies. For Local Presence, do not collapse multiple calls into one daily Contact observation. Use a database uniqueness constraint or atomic upsert on a verified stable identifier, or an appropriately scoped composite key, when concurrent processing matters. A read-then-insert check alone can race.

Before rollout
  • Identify which rules are owned by the Workflow or native HighLevel settings.
  • Define the AI task narrowly, if AI is involved, and verify which output fields are actually available.
  • Test a validation gate and the specific failure route, including missing or repeated input.
  • Name the person or queue responsible for unresolved exceptions and production changes.

Use current HighLevel support documentation and the target account to confirm availability, labels, billing scope, and account-specific behavior before configuring a production process.