--- title: "AI field reports from voice notes, through to the invoice | C-Esium" description: "The technician dictates, AI writes the job report, approval triggers the invoice. How it works, what the AI does, the Tech-O case, cost, timeline and limits." url: https://c-esium.com/en/use-cases/ai-field-reports-from-voice-notes/ --- Use case # AI field reports from voice notes: from the job to the invoice With an AI field report, the technician records a voice note at the end of the job, the AI turns it into a structured report and, once the report is approved, the invoice is triggered without re-keying. C-Esium built this for Tech-O, a multi-branch water-leak detection network in France, where job reports are largely written by AI from the technicians’ voice notes. By [Noé Fantino](https://c-esium.com/en/about/), founder at C-Esium · Updated 6 October 2026 · 6 min read The job report is often the weak link in the chain. It is written in the evening or the next day, from memory, in a form designed for the office. It reaches head office incomplete, someone re-keys it, and the invoice waits. The problem is not the technicians’ goodwill. It is the tool, which asks them to type text on a phone, standing up, between two appointments. Dictating the report to AI solves this. But it is only useful if the report lands in the same tool as the schedule and the invoicing. Otherwise, you have just moved the re-keying. ## How does an AI field report work? Here is the typical journey, as C-Esium builds it into internal software: 1. **Before the job.** The technician opens the job on their phone. Customer, address, reason for the visit, history of the address and any quote are already filled in, from the customer request and the schedule. 2. **On site.** At the end of the job, they record a voice note in their own words: what they found, what they did, the equipment used, what they recommend. They add their photos. 3. **Drafting.** The AI transcribes the note and fits it into your company’s report template: findings, work carried out, materials, recommendations, next steps. It rephrases in clear language your customer can read. 4. **Approval.** The workflow can require the technician to review, complete if needed and approve. The report is filed in the customer’s record with the photos, and can be sent to the customer. 5. **Invoice.** Approval of the report triggers the invoice according to your rules: quote price, fixed fee, time spent or materials used. This step is detailed in the use case [automatic invoicing after each job](https://c-esium.com/en/use-cases/automatic-invoicing-after-each-job/). ### What does the AI do, and what does it not do? - **It does**: transcribe the voice note, structure the report under your headings, rephrase, use the vocabulary of your trade. - **It does not**: make the diagnosis instead of the technician, approve the report, or invoice without human approval. The principle is the same in every C-Esium project: the AI proposes, a person approves. ### Before and after: what changes at each step? | Step | Report typed by hand, then re-keyed | Report dictated into the internal software | | --- | --- | --- | | Writing | In the evening or the next day, from memory | On site, right after the job, by voice | | Content | Varies from one technician to another | Same structure for everyone, in the company’s format | | Photos | Left on the phone, sent separately | Attached to the report and the customer record | | Transmission | Paper, email or re-keying at the office | Available to head office as soon as it is approved | | Invoice | Prepared later, from the report | Triggered by the approval of the report | | Follow-up | Missing reports spotted late | List of reports not yet done, visible at any time | ## At Tech-O [Tech-O](https://c-esium.com/en/case-studies/tech-o/) is a network of non-destructive water-leak detection companies, rooted in rope-access work and organised as a franchise network, with branches in Saint-Étienne, Ambérieu-en-Bugey, Annecy, Limoges, Bourg-en-Bresse and the Lyon area (France). It is a services company. In the internal software built by C-Esium: - job reports are **largely written by AI from the technicians’ voice notes**; - **invoicing is triggered automatically by the job report**; - sales, scheduling, technicians, head office and accounting work in the same tool, with accounting synchronised both ways with Pennylane, the team calendar synchronised both ways with Google Calendar, and built-in SMS and email for customer communication; - management can ask from Claude or ChatGPT which reports have not been done yet: it is one of the 35 or so read-only tools of the MCP connector in production. Across the whole software, not this use case alone, the results are **more than 3,000 hours saved per year, 30% operational gains and 20% fewer human errors**. These are estimates made by C-Esium with Tech-O across field work, organisation, accounting, payroll, HR and management (figures published in October 2026). ## What changes by sector? - **Services**: a job report with findings, photos and recommendations, often sent to the customer. This is the Tech-O case. - **Construction**: site reports, progress, photos and handover records. - **Industry**: maintenance, after-sales service or inspection reports. The approach is the same; we have no published reference in industry yet. - **Health and social care**: a visit or follow-up note can contain health data. The hosting arrangement and the scope of the AI (models hosted within that framework, or pseudonymised data) are defined at scoping, in line with the applicable health-data rules. ## Where do the voice notes and reports go? - **The reports stay in your software.** You own the project code, the database and the documentation, assigned to you by contract. The code is placed in a Git repository in your name at the latest on delivery. - **Hosting.** For clients in the UAE, the software is hosted by default on AWS in its Middle East (UAE) Region, so the data stays in the country. A European region is possible on request. - **AI processing** goes through an AI model. C-Esium recommends business plans or the API (ChatGPT Business or Enterprise, Claude Team or Enterprise), which do not use your data to train the models, and the framework is set out in the contract. - **Access from Claude or ChatGPT** goes through a read-only MCP connector, restricted to authorised users and logged. MCP (Model Context Protocol) is an open-source standard for connecting AI applications to external systems ([official documentation](https://modelcontextprotocol.io/docs/getting-started/intro), accessed 4 October 2026). The details are on the page [connect ChatGPT and Claude to your company data](https://c-esium.com/en/connect-chatgpt-claude-to-company-data/). - **By default, in health and social care**, AI assistants only access data that is not health data. ## How much does it cost and how long does it take? The dictated report can be delivered on its own, as a mobile app for field teams, or inside a complete internal software, where it also feeds the schedule, the invoice and reporting. - **Price**: as a [focused project](https://c-esium.com/en/focused-project/), from AED 40,000; inside a complete internal software, from AED 200,000 (excluding VAT; €10,000 / €50,000 in France). The exact price is set after scoping in a firm quote, paid in milestones ([pricing](https://c-esium.com/en/custom-business-software/#pricing)). - **Timeline**: a V1 on a focused scope is delivered in 1 to 2 months. Jobs, dictated reports and triggered invoices can make up this first scope. Software covering the whole company takes from 6 months to more than a year, delivered in batches. - **Getting started**: a 30-minute call, then a half-day [assessment workshop](https://c-esium.com/en/software-ai-assessment/), on site or by video, both free for a qualified project. ## When is it not the right option? - **You only want to dictate text.** Your phone’s dictation or an AI assistant licence is enough. Licences are the first of the [four levels of AI integration](https://c-esium.com/en/ai-for-business/) and need no project. - **Your off-the-shelf business software already has a report module** that suits your technicians, and your processes are standard. Keep it. - **The content of the report is not defined.** If everyone has their own way of doing it and nobody has decided, AI will not settle the disagreement. Automating a broken process is a known risk: we start by setting the report template with the teams. - **Nobody in-house can lead the project.** An internal lead (project manager or operations manager) is a non-negotiable condition at C-Esium. - **Your budget is below AED 40,000.** Adoption is won in the field. During the assessment, we meet every profile, including technicians who prefer paper. A dictated report that nobody uses saves no hours. ## Frequently asked questions ### Can the AI make mistakes in a job report? Yes, like any automatic transcription. That is why the workflow can include a review and approval by the technician before the report is sent and before the invoice is triggered: the AI drafts, a person approves. ### Do we need a dictation app on top of the software? No. In software built by C-Esium, the voice note can be recorded from the job screen on the technician’s phone. There is no copy-paste between a dictation app and the software: the report is attached to the right customer, address and job. ### How much does an AI field report cost? At C-Esium, it can be delivered as a focused project, for example a mobile app for field teams, from AED 40,000, or inside a complete internal software, from AED 200,000 (excluding VAT). A V1 is delivered in 1 to 2 months; the exact price is set in a firm quote after scoping. ## Read next [### Automatic invoicing after each job Send the invoice as soon as the job report is approved, with no re-keying: how it works, business rules, options compared, the Tech-O case, cost and limits. Read](https://c-esium.com/en/use-cases/automatic-invoicing-after-each-job/)[### Connect ChatGPT and Claude to your data How read-only MCP access works. Read](https://c-esium.com/en/connect-chatgpt-claude-to-company-data/)[### Case study: Tech-O 3,000+ hours saved per year in a multi-branch network. Read](https://c-esium.com/en/case-studies/tech-o/) ## Your software, built for your company. 30 minutes to understand your organisation, your tools and what an internal software with AI would change for you. No commitment. [Book a call](https://c-esium.com/en/contact/)[contact@c-esium.com](mailto:contact@c-esium.com)