Field Services enters private beta

PRIVATE BETA / JOB OPERATIONS

1. Field Services links requests, jobs and visits

Announced 3 November 2026 · Enrolled beta accounts only

What changed

HighLevel announced Field Services in private beta, organised around requests, jobs and visits. The module covers estimates, scheduling, dispatch and billing, with web and mobile working views. Enrolled accounts see a Beta-labelled navigation entry; the announcement is not a general release.

The vendor says availability suggestions consider staff hours, services, service areas, existing bookings and travel time. Manual scheduling can still place work outside a service area. Route changes are proposed for review before application, and online payments require a connected payment provider.

The same post describes an AI receptionist and booking through Google Business Profile. Those are HighLevel’s claims. This research does not independently confirm Google eligibility, Australian coverage or availability in any particular Growthcenter account.

Who this affects

Plumbers, electricians, cleaners, landscapers and other businesses sending staff to customer locations may find this relevant if they have beta access. A consultancy working remotely, a business without field staff, or an operator satisfied with its existing dispatch system can leave it alone.

Our analysis: what it means for your business

Hypothetical example: a plumbing business accepts an enquiry for an outer-suburban address. A spare calendar slot does not settle whether the technician has the right equipment, whether travel fits the day, or whether a second visit will be needed. The dispatcher still needs a clear rule for accepting the work and a way to tell the customer what has actually been confirmed.

Our recommendation is to define the difference between a request and a confirmed visit before exposing a new booking page. Name the person responsible for checking uncertain addresses, unusual work and timing exceptions. If staff can schedule outside the normal area, make the approval rule explicit rather than assuming the service-area setting prevents every exception.

For an authorised beta trial, use a small set of representative dummy jobs: a routine visit, an address near the coverage boundary, a job requiring two staff, and work that needs a return visit. Record the expected assignment and customer message first. Compare the resulting schedule with the dispatcher’s decision; do not treat a suggested route as evidence of time saved in this business.

A trial should answer an operational question, such as whether the office and technician can reliably distinguish quoted work, accepted work and the next visit. Keep the current working process available until those handoffs are understood. The announcement does not establish a reason to replace an existing dispatch product, connect a payment provider or enrol in beta today.

What to do next

1. Check beta eligibility and the actual features visible in the account.
2. Write the acceptance rule for requests, service-area exceptions and confirmed visits.
3. If a trial is authorised, compare a small dummy-job set with the dispatcher’s expected decisions.

PRODUCT UPDATE / CONVERSATION AI

2. A model recommendation needs an agent check

Announced 3 November 2026 · Prompt-based agents

What changed

HighLevel now recommends GPT-6 Sol for prompt-based Conversation AI and uses it by default for newly created agents. The post says existing model selections remain, with a stated exception: agents previously using GPT-5 or GPT-5 Mini have already moved to GPT-6 Sol.

HighLevel also says it will retire GPT-4.1 within Conversation AI, without giving a date in this announcement. Its assertion about OpenAI deprecation was not independently confirmed for this edition. Vendor evaluation, speed and cost claims are not evidence of a particular business’s outcomes or bill.

Who this affects

Businesses already using prompt-based Conversation AI for written customer enquiries should check their configured agents. This is not evidence of a Voice AI migration. Businesses without this product can ignore the model-menu change; Growthcenter account applicability remains unverified.

Our analysis: what it means for your business

Hypothetical example: a cleaning business’s assistant receives an enquiry for a service it does not offer, followed by a request for a same-day booking. A useful check is whether the configured agent respects both limits and hands the conversation to the right person. A more fluent reply is not enough if it creates an unsupported service or timing promise.

Our recommendation is to inspect the model actually selected in each relevant agent. The new default matters when someone creates a replacement or additional agent; the migration exception matters when staff assume an old agent has stayed unchanged. Record the observed setting and the date instead of relying on the model originally used during setup.

Before an authorised production change, compare a fixed set of enquiry examples using the product’s test facility or an isolated copy where available. Include an ordinary question, an unavailable service, an after-hours request and a request to speak to a person. Mark whether each response follows the business’s supplied rules, asks for missing information and avoids making a booking commitment it cannot substantiate.

Keep model choice separate from the information and actions the agent can use. A wrong service area, outdated price or unclear escalation instruction needs correction at its source. Changing the model is not evidence that those inputs have been fixed. Record any failed examples for the existing account owner to investigate; no such testing was performed for this edition.

The undated retirement statement is a reason to watch the vendor’s account notices and plan a controlled review, not to invent a deadline or rush an untested switch. If a business is already satisfied with its configured agent, its immediate task can be documenting the current setting and the small acceptance test it would use for a future change.

What to do next

1. Record the actual model setting on existing prompt-based agents.
2. Compare realistic enquiry and human-handoff examples before an authorised change.
3. Keep vendor retirement notices, test results and account costs distinct from marketing benchmarks.