How to Track Parts Inventory and Job Photos With an AI Assistant

Your AI assistant needs a place to put the evidence

An AI assistant can remind techs, read notes, summarize work, and prepare invoices. But it cannot keep parts and photos straight if everything lands in texts, camera rolls, folders, and random spreadsheet tabs. The assistant needs a structured place to record what part moved, from where, to which job, by whom, and with what photo documentation.

For HVAC, electrical, plumbing, AV, and smart-home work, the details matter. A missing photo can delay billing or weaken a warranty claim. A missing part record can make a profitable job look fine while inventory quietly disappears. The fix is to treat parts and photos as job records, not afterthoughts.

Track parts as movements, not just a list

A simple parts list tells you what items exist. It does not tell you what happened. Real inventory tracking needs movement records. Every part should be able to move from supplier to shop, shop to van, van to job, job to return, or job to warranty claim.

At minimum, your assistant should capture five facts any time inventory changes: the part, the quantity, the source location, the destination or job, and the timestamp. If the part is billable, that should be marked at the job level. If it came from a technician's van, the van stock should be reduced. If it was ordered for a specific job and not used, it should not vanish into the shop shelf without a record.

This is the difference between asking, “Did we use a 45 amp breaker?” and asking, “Which job consumed the 45 amp breaker that was loaded onto Mike's van on Tuesday?” The second question is the one owners actually need answered.

Use job photos as structured documentation

Job photos should not be a pile of images named IMG_4027. Each photo needs context. Was it before work, during work, after completion, a serial plate, a damaged part, a code issue, a ceiling cut, a panel label, or a customer approval item? A dispatcher should not need to open twenty thumbnails to guess.

Train the assistant to ask for photo categories at upload time. Common categories include before, after, equipment label, wiring, access condition, damage, installed part, safety issue, permit, and closeout. The categories can be simple, but they need to be consistent. Consistency is what lets the assistant later say, “This completed job is missing an after photo,” or “Here are all serial plate photos for heat pumps installed this month.”

The operating workflow that works in the field

The workflow should be short enough that a busy technician will follow it. Do not ask techs to become data clerks. Ask for the few items that protect the job, the customer, and the invoice.

1. The dispatcher creates or confirms the job record before the visit.
2. The technician opens the job on arrival and sees required photos or checklist items.
3. Parts used are selected from a known list or entered as a temporary part for office review.
4. The assistant records quantity, location, and billable status.
5. The technician uploads required photos under simple categories.
6. The assistant flags missing documentation before the job is closed.
7. The office reviews parts and photos before invoicing.

That last step is important. AI can reduce the chase, but the office still needs a review point before money goes out the door or a customer receives a final bill.

What the database should store

For parts, store item name, SKU, supplier, category, unit cost, active status, and stock location. Then store every movement separately. For job use, connect the part to the job, the technician, the quantity, and whether it should appear on the invoice.

For photos, store the job, visit, technician, file URL, caption, category, and timestamp. If you work on equipment, connect photos to assets when possible. A serial plate photo attached only to a file folder is easy to lose. A serial plate photo attached to the actual equipment record becomes useful for years.

This is why a real PostgreSQL operations database matters. Your assistant needs relationships: job to property, property to client, job to visit, visit to tech, job to parts, job to photos, job to invoice. Without those relationships, it can only summarize loose material.

Where SQL Agent fits

SQL Agent is built for this exact kind of service-business recordkeeping. It is a $295 one-time pre-built 38-table PostgreSQL operations database that an AI assistant can install in one command. Instead of starting with empty tables or messy spreadsheets, your agent gets a structure for dispatch, clients, parts inventory, job photos, and invoices.

That matters because parts and photos touch several parts of the business. A used part affects inventory, job costing, billing, warranty, and sometimes a future callback. A photo affects closeout, customer proof, QA, and invoice support. SQL Agent gives those records defined places to live so your assistant can manage them instead of hunting for them.

Questions your assistant should be able to answer

Once the data is structured, the assistant becomes useful in plain operational language. You should be able to ask:

• Which completed jobs are missing after photos?
• Which parts were used on yesterday's calls but not marked billable?
• Which van stock locations are below reorder level?
• Which jobs used customer-supplied parts?
• Show every photo and part tied to this callback.
• Which invoices are waiting on documentation?

Those questions are not fancy. They are the daily questions that keep a service company from bleeding time in the office.

Bottom line

If you want an AI assistant to track parts inventory and job photos, do not start with prompts. Start with structure. Decide what must be captured, where it belongs, and how it connects to the job. Then let the assistant enforce the routine: parts used, photos attached, missing items flagged, invoice ready.