Superintelligence Readiness for Refrigeration Crews: Intake and History
How refrigeration service companies can use AI agents for call intake, unit history and refrigerant log reminders, so technicians arrive briefed.

An AI agent for a refrigeration service company is most useful at intake, where a service call becomes a dispatch decision. It reads the request, checks the unit's recorded history and any open refrigerant log items, and gives the technician a short brief before the truck leaves. Be Human designs and runs that kind of agent with written approval boundaries, so the dispatcher and owner keep control over what gets promised and what gets filed.
Key takeaways
- Intake is where a unit's past failures most often get lost, so it is the first place to put an agent.
- Equipment history should be records the agent reads and cites, never details it reconstructs from memory.
- Refrigerant log reminders are obligations, so each one needs a named owner and a visible due state.
- A dispatch brief should state known facts, flag conflicts and mark gaps as unknown.
- Start with one service line and one observable criterion, such as whether every brief is complete before dispatch.
Why the first call decides the visit
A refrigeration call often starts with a sentence like the walk-in is running warm again. The person who takes the call may know the customer, but not which compressor was replaced last year, whether a door gasket was flagged at the previous visit, or whether a refrigerant entry is still waiting for paperwork. The technician then arrives without that context and spends the first part of the visit rebuilding it by phone, which costs time on site and makes the customer repeat the story.
The problem is rarely a lack of data. Most companies already have tickets, technician notes and photos somewhere. The gap is that the information sits in separate systems and in people's memories, and no one is assigned to assemble it before the truck is dispatched. The same pattern appears in other callback-heavy work, as described in the elevator maintenance callbacks guide. Readiness for more capable systems starts with this kind of basic discipline: clear records, named owners and limits that a person can inspect.
What intake should capture before dispatch
A useful intake agent asks for a small, fixed set of facts and writes them in a structured form. It should not interrogate the caller or promise a visit time it cannot confirm. The minimum set of fields looks like this:
- Site address, access hours, gate or roof access notes, and a site contact who can open the door
- Unit identifier as recorded on the last ticket, with make and model where known
- The symptom in the caller's own words, plus any temperature readings the caller can see
- Date and technician of the last visit, and the work that was performed
- Open work orders, quotes awaiting approval, and parts on backorder for this unit
- Refrigerant log entries for the unit and any reminder currently due
- Service agreement or warranty status, if the company tracks it
- Urgency as stated by the caller, and whether stored product is at risk
Anything the agent cannot confirm should be marked unknown. An explicit gap is more useful than a plausible guess, because it tells the technician exactly what to check on arrival.
Equipment history the agent can cite
Equipment history is only helpful if every statement in the brief can be traced to its source. Each line should point to a ticket, a technician note, a parts record or a log entry with a date and an author. When two records disagree, for example a note saying an evaporator fan was replaced while the parts record shows no matching sale, the brief should show both and flag the conflict instead of choosing one.
Keep the history store separate from the model. The agent retrieves the records relevant to the call and works from them. Anthropic's engineering guidance on context engineering makes a related point: a small, well-chosen set of information usually serves an agent better than everything that might be relevant. A brief that runs to several pages will be skipped, so the length of the brief is a design decision, not a detail.
Refrigerant log reminders stay with a named owner
Refrigerant log reminders are obligations, and obligations need an owner. The agent's job is to show which entries are due or overdue for each unit, display the last entry, and route the reminder to the person authorized to act on it. It should not decide whether a particular leak finding requires a particular step, interpret a regulator's thresholds, or file a record on its own. Those rules differ by jurisdiction and change over time, so the company's compliance lead or a certified technician should confirm them, and the agent should cite that confirmation rather than restate the rule from memory.
In practice, the agent can create a reminder, attach the last log entry and notify the assigned person. Writing a new log entry, closing a reminder or changing a unit's record belongs to a human with the right permission. Access to customer and site records should follow role-based rules, so a dispatcher sees what dispatch needs and nothing more. The OWASP authorization guidance is a practical checklist for writing those rules down and testing them.
Dispatch brief: a step-by-step plan
Once intake is complete, the brief is assembled and checked before a technician is assigned. A workable sequence looks like this:
- Step 1: Confirm the caller, site and access details, and mark any field the caller cannot verify.
- Step 2: Match the unit to its record. If two candidate records exist, list both and ask the dispatcher to choose.
- Step 3: Pull the most recent visits for that unit, with the symptom, the work performed and the parts used.
- Step 4: List open items: quotes awaiting approval, backordered parts and refrigerant log reminders that are due.
- Step 5: Write a brief of no more than one screen, with known facts, open questions and a suggested first check.
- Step 6: Send the brief to the assigned technician and the dispatcher, and record that it was sent and when.
- Step 7: After the visit, the technician confirms or corrects the brief, so the next brief for that unit starts from better records.
Keeping a record of what the agent read, what it proposed and who approved each action makes errors traceable. The observability concepts in the OpenTelemetry primer, which describe logs, traces and metrics for software systems, translate well to this kind of audit trail.
Illustrative scenario: a warm walk-in cooler on a Monday
This scenario is illustrative and does not describe a real client. A caller reports that a walk-in cooler is running warm and that the door seems to stay open longer than it should. The intake agent records the site, the access window and the symptom in the caller's words. It then finds that the unit had a door gasket replaced earlier in the year, and that a note from the previous visit mentions a fan motor that was noisy but still working. It also finds a refrigerant log reminder marked as due with no entry after it.
The brief tells the technician what was replaced, what was only observed and what is still open. The reminder goes to the compliance lead, who decides whether it can wait until after the visit. The technician arrives knowing which parts to bring and which questions to ask. Afterward, the dispatcher can see whether the brief matched what was found on site, and that answer shapes the next revision of the intake fields.
Starting with one service line
Start with one service line, one unit type and one success criterion that the owner can check, such as whether every brief was complete before the technician left. Be Human designs that first role, defines the memory and tools it needs, and sets approval boundaries for reminders, customer messages and record changes before anything is switched on. Managed operations then keep the integrations working, review permissions when staff change, monitor brief quality, run recurring evaluations on real examples and handle incidents when a record turns out to be wrong. The work is remote and available in English and Portuguese.
If you want to find the first task worth handing to an agent in your dispatch flow, take the free Be Human AI diagnostic.
Frequently asked questions
Can an AI agent dispatch refrigeration technicians on its own? A well-designed agent proposes briefs and assignments, while a dispatcher keeps authority over who goes where and what gets promised. Approval boundaries should be written down before the agent is switched on, and any change to them should be reviewed like any other access change.
What records does the agent need to build a useful brief? It needs the unit identifier, prior tickets with technician notes, parts records, open quotes, refrigerant log entries and site access notes. Missing or contradictory records should appear as gaps in the brief, and the agent should never fill them with plausible text.
Does this replace the refrigerant log or the compliance work? No. The log remains the company's record, and the people responsible for compliance decide what the rules require where the company operates. The agent only surfaces due items, attaches the last entry and routes the reminder to the right person.
How do we know the agent is actually helping? Define the criterion before launch, for example whether each brief contains the fields the dispatcher agreed on and whether technicians report fewer repeated questions on site. Review a sample of briefs against what the technician actually found, and keep logs of what the agent read and proposed so that any error can be traced.
Your next step
See where an AI agent fits in your company first.
Take the free Be Human AI diagnostic. A few questions about your work show which task to hand to an agent first, what it needs to know and where a person approves.
Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.


