It is a quarter past three on a Wednesday afternoon and the dispatch phone rings. A clinic has a chiller down, the sterilisation room is warming up and the maintenance contract promises a technician on site within four hours. The coordinator opens what holds the day's visits together: a shared spreadsheet and a WhatsApp group with twenty technicians. The technician who knows that unit is sixty kilometres away, halfway through a service. The one closest has never been there and does not know the compressor was replaced in March. Nobody is quite sure whether the part that usually fails is in any of the vans.
The technician arrives, fixes it, sends three photos to the group and fills in a paper job sheet that the customer signs against their knee. The sheet reaches the office on Friday, someone keys the data into the ERP the following week, and the invoice goes out with whatever hours could be read. The out-of-hours call-out charge, which the contract allowed, was left off because nobody remembered.
Anyone running an HVAC, lift, solar, electronic security or electrical installation business will recognise that morning, and knows the problem is not the people who go out on site.
Short answer
Software for field technical teams has to take every job from the call to the invoice without anyone copying data from one place to another: work orders assigned to the right technician, an app that works offline, the history of each piece of equipment, the parts stock in each van and the rules of each contract applied at close-out. AI comes in afterwards, to turn voice notes and photos into a report, suggest a diagnosis and predict parts, always with the technician confirming. A custom system for this operation typically costs between €15,000 and €40,000, going live between week 10 and week 16.
Money leaks between the phone and the invoice
The technical work itself usually goes well. What goes wrong is everything around it: logging the call, choosing who goes, the part that should have gone and did not, the report that takes days, the invoice that goes out incomplete and the history that lives in one technician's head instead of against the equipment.
A system for field teams is not there to "digitise the job sheet". It is there so that every job starts with the right information and ends with everything needed to bill for it, without anyone copying data from one place to another.
Signs the operation has outgrown paper
If three or more of these are routine, the current process already costs more than it looks:
- reactive jobs are handed out by phone and group chat, and only the coordinator knows who is where;
- a technician arrives on site without knowing what was done last time on that equipment;
- there are second visits because a part was missing that could have been predicted;
- job sheets reach the office days later, and some never arrive at all;
- nobody can say, without an afternoon's work, whether this month's SLAs were met;
- contracted preventive visits are scheduled from a list someone keeps by hand;
- van stock is an estimate;
- one month's invoicing only closes halfway through the next.
Where this kind of operation outgrows generic software
There are off-the-shelf field service management tools, and some are good. But there are five areas where the real operation of these companies tends to be more demanding than the sales demo suggests.
The equipment is the unit, not the customer
A customer has three air conditioning units, two lifts or a solar installation with forty panels and two inverters. Each piece of equipment has its own make, model, serial number, installation date, warranty, preventive plan and history. When the data model revolves around the customer, the history gets mixed up and the technician can no longer answer the most useful question of all: what has already happened to this machine?
That is why, in the systems we design, every piece of equipment carries a QR code label. The technician points the phone at it and opens the asset record, the latest jobs and the work order in progress, without searching for anything.
The app has to work in a basement
Plant rooms, lift motor rooms, rooftops and car parks are exactly the places where mobile signal fails. An app that needs a connection to open a job or save a photo is an app technicians abandon in the first week and replace with paper.
Working offline forces engineering decisions that are best taken early: which data goes onto the phone before the technician leaves, how much space the photos take up, and what happens when a technician closes a job offline while the office has rescheduled it. These conflict rules look like a detail until the day two people turn up at the same customer.
The van is a warehouse
In a company with thirty technicians there are thirty small warehouses on the move. If the parts used on a job are not taken out of the van's stock, the inventory stops adding up within weeks, restocking is done by eye and the real cost of each job is never known. Treating each vehicle as a stock location, with transfers and stock counts, is what separates a system that controls costs from one that merely logs visits.
Contracts have small print
A maintenance contract rarely just says "two visits a year". It says which response time applies during business hours and outside them, which parts are included, which call-outs are chargeable, what overtime costs and what happens if the SLA is missed. If those rules are not in the system, they live in somebody's head and fail every time that person is on holiday.
Closing the job is invoicing
The job closes with a report, photos, parts, hours and the customer's signature. If all of that is in the system, the invoice can be prepared on the spot, with the contract rules already applied. One important detail: in Portugal, invoices issued by software must come from a program certified by the tax authority, so the invoice still comes out of the certified program the company already uses. The system prepares the document and sends it through the ERP integration, instead of someone typing it up again. It is the same integrate-rather-than-replace logic we follow in the custom ERP for wholesale distributors.
Where artificial intelligence genuinely helps
In this sector, AI does not need to be ambitious to be useful. There are three places where the gain is concrete and shows up in technician time and in visits avoided.
Reports from voice and photos. At the end of a job, the technician is tired, has dirty hands and is already running late for the next one. Writing a decent report on a phone is the step most often done badly. Dictating two minutes on what was found and what was done, and letting the system propose a structured report, with the parts identified, the hours and the photos captioned, solves a problem almost every company in the sector has. The technician reviews and corrects it, and the report that reaches the customer becomes legible and complete, which is also a matter of image.
Assisted diagnosis. When a less experienced technician finds an error code on equipment they do not know, the answer is often in the history for that model and in the manufacturer's manual. An assistant that looks up previous jobs on identical equipment and the company's technical manuals can suggest the most likely causes and the procedure to follow. It does not replace the senior technician, but it saves a lot of calls to them, and it captures knowledge that would otherwise be lost when someone leaves the company.
Predicting which parts to take. With the equipment history, the type of fault reported on the call and what is usually needed in those cases, the system can suggest which parts should be in the van before it leaves. Every second visit avoided is technician time recovered and one less frustrated customer.
Two conditions have to hold for any of this to work. The first is data: without a structured history of jobs, the AI has nothing to work with, which is why these modules usually come in once the basic record has been running for a few months. The second is that the technician always has the final say. A wrong suggestion nobody checks is worse than no suggestion at all.
It is also worth separating this from sensor-based predictive maintenance, which is a different conversation. In a factory with its own equipment, reading vibration and temperature continuously makes sense, as we explain in the article on predictive maintenance in manufacturing. In a company that maintains equipment across hundreds of customer sites, the main source is the job history, and that is where it pays to invest first.
A note on location and routing
Planning the order of visits and assigning an urgent job to the nearest technician saves kilometres and hours. But tracking employees' location has legal limits in Portugal. Article 20 of the Labour Code prohibits the use of remote surveillance to monitor professional performance, and the CNPD, the Portuguese data protection authority, has its own guidance on geolocation in the workplace, which should be read alongside the GDPR and Law 58/2019. That guidance accepts external technical assistance as one of the activities where vehicle geolocation can be justified, but it requires a defined purpose, prior information to employees and data that is not used to assess each person's performance.
In practice, we design these features for the minimum purpose: knowing the address of the next job and the area each technician is working in, without continuous tracking or route reports nobody asked for. What is acceptable in your operation should be validated with your legal adviser before it is built, not after.
Three wrong answers to this problem
Buying the tool with the most features. The most complete off-the-shelf products come with dozens of modules and a per-technician licence that grows with the team. If half the modules do not fit the company's real contracts, the result is an expensive tool that lives alongside parallel spreadsheets.
Asking the ERP to do everything. The ERP is good at invoicing and accounting, but it was rarely designed for a technician in gloves, with no signal, closing a job on a rooftop. Stretching the ERP out into the field usually produces screens nobody wants to use on a phone.
Starting with AI. It is tempting, and it is a mistake. Without structured work orders and properly registered equipment, there is no history to feed the diagnosis or to predict parts. The right order is to record things well first and then put AI to work on that record, and it is worth checking whether your company data is ready for AI before counting on it.
From on-site discovery to a regional rollout
1. Free diagnosis
Around half an hour to understand the operation: technicians, equipment, contracts, systems in use and where most time is lost. It is free and carries no commitment to proceed.
2. On-site discovery and proposal
Two to three weeks mapping the real process, ideally riding along on a few jobs and sitting down with dispatch and billing. We list the contract types and their rules, the integrations and the odd cases. At the end, scope, timeline and investment are fixed, with no mid-project revisions.
More and more companies reach this stage with a prototype of the technician app built in ChatGPT, Claude or Gemini. That shortens discovery and lowers the price, because the screens and rules are already visible. What a prototype does not bring is still missing: user accounts for each role, real data, a connection to the ERP, offline use, backups and tests. We explain that step in what an AI-built prototype is missing to reach production.
3. Build in short cycles
Two-week cycles, each with a demo. The technician app is tested early by real technicians, because it is the part that decides whether the project succeeds. The AI modules come in once the core record is stable.
4. Rollout by team or region
Migration of equipment and contracts, short training sessions and a go-live with part of the team for a few weeks. We fix what the field reveals and only then extend it to everyone. The code and the infrastructure belong to the company.
Investment by scope, and what moves it
The figures below are the ones we publish on our solution pages.
| Scope | Typical investment | What it includes |
|---|---|---|
| Core maintenance system | €15,000 to €25,000 | Equipment, preventive plans, work orders and mobile app |
| System with parts management | €25,000 to €40,000 | Adds stock per van, cost per asset and purchasing |
| Asset platform | from €40,000 | Multiple sites, availability metrics and sensor connectivity |
| Multi-role portal | €25,000 to €45,000 | Dispatch, technicians, billing and customers with separate access |
Time to go-live typically falls between week 10 and week 16 (what makes that timeline vary is covered in how long a custom software project takes). What pushes the figure up most is the number of integrations (ERP, invoicing, alarm receiving centres, manufacturer portals), the complexity of the contract rules and the decision to give customers a portal where they can view jobs and raise requests. What brings it down is starting with one contract type and one region, and leaving parts management or AI for a second phase.
When an off-the-shelf tool is the right choice
With fewer than ten technicians, simple contracts and an ERP that already invoices well, a carefully configured off-the-shelf tool is almost always enough, and we say so in the diagnosis. Nor is it worth it when the company has not yet decided how it wants to run the operation: a custom system locks in a process, and locking in a process that will change in six months is money thrown away.
It pays off when the operation has dozens of technicians, contracts with their own rules, parts that weigh on cost and the persistent feeling that you invoice less than you work.
What to know before the first conversation
- How many jobs are carried out per month, and how many are reactive?
- How many contract types are there, and where are their rules written down?
- How many second visits were there last quarter because a part was missing?
- On average, how long passes between closing a job and invoicing it?
- Which systems have to stay connected: ERP, invoicing, purchasing, customer portal?
- Who in the company decides what the technician app shows and asks for?
If you have these answers, half the discovery work is already done. On the page about preventive maintenance management software we describe how we design these systems, and on the one about internal portals how we bring dispatch, technicians and customers together on the same platform. If you want to understand what would make sense for your operation, you can request a free diagnosis. We reply within 24 business hours and, if it makes sense to go ahead, the proposal comes with scope, timeline and investment fixed.
