Automation isn't new: we've spent years connecting tools with rules of the "if this happens, do that" kind. What changes with AI is that machines now understand text, documents and images, and that opens the door to automating tasks that until recently needed a person reading. These are ten processes we're automating in real SMEs, ranked from easiest to most ambitious.
1. Sorting and replying to incoming email
The AI reads every email, tags it (quote, support, invoice, spam), assigns it to the right person and drafts a reply with your details. A person reviews and sends. Typical result: first-response time drops from hours to minutes and nobody qualified spends the morning in the inbox.
2. Extracting data from supplier invoices
Invoices arrive as PDFs, photos or emails. The AI extracts the supplier, number, date, net amount, VAT and line items, and records them in your ERP — in our case, Holded — so that admin only has to validate. Goodbye to keying in a hundred invoices a month.
3. Preparing first-round quotes
From a request in plain language ("I need an 8-page website in two languages with bookings"), the AI identifies the services in your catalogue, proposes lines and quantities and generates the draft. The salesperson refines and sends. Especially powerful if your catalogue is well structured.
4. 24/7 customer service with context
Not a menu-driven chatbot, but an assistant that answers with your real information (opening hours, policies, availability, order status) and knows when to hand the case over to a person. Out of hours and in several languages. In hotels, this answers at 3 a.m. what the team would answer at nine.
5. Reports that write themselves
On Monday morning, a summary of sales, outstanding payments, open tickets or campaign performance, written in plain text from your data. Without anyone opening five tools and building a spreadsheet. That's how we ourselves receive the weekly snapshot of our clients.
6. Product sheets and translations
Descriptions, attributes and SEO copy for hundreds of products from technical data, in your shop's languages, with your brand tone as the guide. Human review by sampling. What used to take weeks gets done in days.
7. Lead qualification
Every enquiry that comes in is scored by sector, size, urgency and fit with what you sell, and enriched with public information about the company. The salesperson starts the day with the five contacts that matter, not with fifty forms.
8. Client and employee onboarding
When a sale closes, the welcome documentation is generated, tasks and access are created, and an assistant answers the typical questions of the first few weeks from your internal documentation. The same goes for a new team member.
9. Review and reputation management
The AI groups reviews by topic (cleanliness, service, price), spots trends before they become a problem and proposes personalised responses for approval. In restaurants and hotels it's one of the first things that pays off.
10. Agents that run complete processes
The next step up: an agent that receives an order, checks stock, generates the invoice, notifies logistics and replies to the customer, querying your systems via API at every step and asking for human confirmation when the case falls outside the norm. It's where the most value is and where doing it well matters most.
The rule for getting it right
Automate first what is frequent, boring and low-risk. Always keep a human in the loop where an error is expensive or where the customer would notice the difference. And measure: if you don't know how many hours it saves, you don't know whether it works. If you want to see which of these ten apply to your business, we start with an AI audit; and if you already know, this is how we build automations that don't stop at the demo. Let's talk.
