AI for small businesses: 8 practical uses in Noumea

AI for small businesses is spreading fast. In 2025, 18% of companies with 10 or more employees based in France were using it, up from 10% a year earlier, according to INSEE, the French statistics institute. The share reaches 15% for firms with 10 to 49 employees and 31% for those with 50 to 249.

There is no equivalent statistic for New Caledonia. In February 2025, the vice-president of the CCI, the local chamber of commerce, said that “we are only at the very beginning of using AI in the workplace”. Here are eight practical uses for a small business in Noumea, and the precautions that go with them.

Where businesses stand, in France and in New Caledonia

The INSEE survey (Insee Première no. 2120, July 2026) covers companies based in France and does not include New Caledonia. It shows rapid adoption held back by skills: 53% of companies using AI say they lack expertise, and 43% cite data protection.

Among companies that do not use it, 71% see no use for it and 54% mention a lack of expertise. The main obstacle is therefore know-how rather than price.

Locally, interest is real. Station N and the digital cluster Open NC held a day on AI for businesses on 28 August 2025, with representatives of the government’s digital directorate. As early as February 2025, the government member in charge of digital affairs called for an AI strategy for New Caledonia.

For a small business, the point is not to adopt “AI” wholesale but to identify two or three tasks where it saves time. Our page on digital tools puts these choices in the context of your whole set-up.

Eight practical uses of AI for small businesses

These uses rely on generative AI assistants, in consumer or business versions. In every case, a person reviews the output before it goes out.

1. Quotes and proposals: structure a proposal from your notes and price list, then rework the text.
2. Emails: draft a reply to a complaint or a payment reminder in the right tone.
3. Meeting minutes: turn meeting notes into minutes with decisions and actions.
4. Monitoring: summarise a long public document, such as an economic note from IEOM, the issuing bank for France’s Pacific territories, or an official resolution, before reading the relevant parts.
5. Translation into English: prepare an exchange with a client or supplier in Australia or New Zealand, then have the sensitive points checked.
6. Customer FAQ: write answers to recurring questions for your website or reception desk.
7. Spreadsheet analysis: spot trends in a sales or expenses export, and check the calculations.
8. Job adverts: write a clear advert from the job description, in line with local employment rules.

Start with tasks that are repetitive, frequent and low-stakes. A mistake in a draft email can be fixed; a mistake in a signed quote is expensive.

Take a typical case: a five-person services firm that writes several quotes and reminders every week. It can start with uses 1 and 2. The manager jots down the essentials, the tool suggests a structure and a first draft, and the final text remains the manager’s own, with its prices and terms.

What AI does badly: local realities and Caledonian law

As CNIL, France’s data protection authority, points out, these systems can “generate inaccurate results that may nonetheless seem plausible”. Extra care is needed on topics specific to New Caledonia. A user interviewed by LNC, the local daily, in 2025 said AI gives her “a huge amount of content” that she then adapts to New Caledonia.

Labour law, taxation and social security are specific to the territory. A tool may talk about VAT where TGC, the local general consumption tax, applies, or cite an article of the French mainland Labour Code that has no force here.

• check any legal, tax or social security rule on juridoc.gouv.nc, the website of the DSF (the local tax office) or that of CAFAT, New Caledonia’s social security fund;
• ask the tool for its sources, then open them;
• do not delegate any decision to it, as CNIL recommends: it prepares, you decide.

Data and confidentiality: the basic rules

According to CNIL, France’s data protection act (loi Informatique et Libertés) has applied in full in New Caledonia since 1 June 2019. The EU’s GDPR does not apply there directly, but it applies wherever that act expressly refers to it. Your clients’ and employees’ data are therefore protected.

For deploying generative AI, CNIL recommends starting from a concrete need, setting rules for use, checking whether the provider reuses submitted data to train its model, and training users. For a small business, that means:

• no personal data (names, contact details, health data) in a free consumer tool;
• documents anonymised before they are submitted, with names and sensitive details replaced;
• settings checked, with the use of your data for training switched off where the tool allows it;
• a one-page usage policy: approved tools, prohibited data, mandatory review.

Choosing tools and setting a budget

There is no need to pile up subscriptions. For a team of fewer than 15 people, one conversational assistant plus the AI features of the office suite you already use often cover all eight uses.

Four criteria for choosing:

• a business plan with written commitments on confidentiality;
• quality in both French and English;
• compatibility with your tools: email, spreadsheets, storage;
• the cost per user per month, often billed in US dollars or euros: allow for the exchange rate.

Set a time-limited test budget for two or three users rather than subscribing the whole team at once. Set aside time for training too: lack of expertise is the main obstacle cited by companies that use AI.

Getting started in 30 days: test, measure, roll out

1. Week 1: pick two of the eight uses, where the team loses the most time, and measure the time spent today.
2. Week 2: test with two or three volunteers, applying the usage policy.
3. Week 3: compare time spent and quality, and note any errors found.
4. Week 4: decide whether to roll out, adjust or stop, then train the rest of the team.

Keep the measures simple: time spent per task before and after, the number of corrections needed and the views of the people using the tool. If a use saves no time after a month, drop it. The aim is not to use AI for its own sake but to free up time for your clients and your business.

After that, add one use at a time. This rhythm suits monthly support: a monthly check-in to track time saved, fix what does not work and decide what comes next.

Frequently asked questions

What can a small business use AI for?

The easiest uses to start with are quotes and emails, meeting minutes, document summaries, translation into English, answers to frequent customer questions, spreadsheet analysis and job adverts. Every text is reviewed before it is sent.

Can you use ChatGPT with customer data?

Avoid entering personal data into a free consumer tool. CNIL recommends checking whether the provider reuses data to train its model and setting rules for use. Choose a business plan and anonymise documents.

Does the GDPR apply in New Caledonia?

Not directly. According to CNIL, France’s data protection act has applied in full there since 1 June 2019, and the GDPR applies wherever that act expressly refers to it. Your clients’ and employees’ personal data are protected.

Sources

• Insee, Les technologies de l’information et de la communication dans les entreprises en 2025, Insee Première n° 2120, 21/07/2026
• Les Nouvelles Calédoniennes, Quelle place pour l’intelligence artificielle en Nouvelle-Calédonie ?, 15/02/2025
• Outremers360, Nouvelle-Calédonie : les entreprises calédoniennes sensibilisées à l’intelligence artificielle, 29/08/2025
• CNIL, Comment déployer une IA générative ? La CNIL apporte de premières précisions, 18/07/2024
• CNIL, Comment la loi Informatique et Libertés s’applique-t-elle à l’outre-mer ?, 04/07/2019