12 processes SMBs can automate with AI (and how to get started)

The processes to automate first with AI are the repetitive, frequent and measurable ones: routing requests, reading documents, answering the same questions, sending reminders and follow-ups, moving data between systems that don’t talk to each other. Start with one, measure the result, then expand. Artificial intelligence is needed where a text, an image or a voice has to be understood; for everything else, precise rules are enough.

Italian companies are moving in this direction: according to Istat, Italy’s national statistics institute, 16.4% of companies with at least 10 employees used AI in 2025, and the most common areas were marketing and sales (5.4%) and organisation of administrative processes (4.2%).

How to choose where to start

Score each candidate process from 1 to 3 on four criteria:

• Frequency: how many times does it happen in a month?
• Time: how many minutes does it take each time?
• Errors: how often do mistakes happen or things get forgotten?
• Risk: how serious is a mistake? Here the score works the other way round: less risk, more points.

Start with the process that scores highest. It is usually something unglamorous, such as sorting emails or sending reminders.

The 12 processes

1. Routing incoming emails and requests

AI reads each message, works out what it is about, extracts the useful data and assigns it to the right person or queue, with a draft reply. A word of caution: external messages must be treated as untrusted data, to prevent prompt injection.

2. Answering the phone and taking bookings

An AI phone assistant answers at any hour, books appointments in the calendar and confirms them on WhatsApp. Calls that need a person arrive with the details already collected.

3. Reading invoices and documents

Text recognition and AI extract the supplier, amounts, due dates and line items from PDFs and photos and record them in your business software. Rules check that the totals add up before anything is recorded.

4. Bank reconciliation

Account transactions are matched to invoices and instalments: AI proposes the matches with a confidence score, and a person confirms only the uncertain ones. It is one of the tasks the agents carry out in our property management platform.

5. Payment reminders

Friendly reminders before the due date and follow-ups after it, with tone and timing tailored to each customer. The heavier actions, such as a formal notice, go out only after a person has approved them.

6. Appointment reminders

Automatic messages the day before, with the option to confirm or reschedule. They reduce no-shows and free up slots for other customers.

7. Lead qualification

Leads coming in from the website, campaigns and social media are enriched, classified and assigned to the right salesperson, with an immediate first reply.

8. Quotes and proposals

Starting from a request or a call, AI drafts a quote using the company’s price lists and templates. A person checks it, corrects it and sends it.

9. Management control from invoices

Issued and received invoices are synced and automatically allocated by project and area, with margins, due dates and average collection times always up to date. That is what we built in a management control system connected to FattureInCloud, an Italian invoicing platform.

10. First-line customer support

Answers to frequently asked questions, the status of orders and cases, tickets opened with the details already filled in. When a person is needed, the conversation is handed over with all its context.

11. Moving data between systems

Orders, shipping, inventory and production often live in different software. An automation keeps them in sync: in the pipeline for 3D-printed displays, every piece moves from data entry to tracked shipment with no manual re-keying.

12. Voice notes and meetings turned into data

A note dictated after a visit or a meeting becomes an activity in the CRM, an updated record or a to-do list with deadlines.

Which tool to use

Tool	When it makes sense	Limitations
Zapier and similar services	Simple connections between widely used apps	Costs that grow with volume, data held on third-party services
n8n self-hosted on your servers	More complex workflows, data that must stay under your control	Has to be installed, monitored and updated
Custom code	High volumes, complex logic, strict security requirements	Requires development and maintenance
AI agent	Varied requests to understand and turn into actions	Must be designed with permissions and controls

The best solution often combines several tools: rules and workflows for the predictable steps, an AI agent for those that require understanding.

How to measure the result

• Time saved: minutes per case before and after, multiplied by the volume.
• Response times: how long a customer waits before getting an answer.
• Errors: cases that need correcting, missing data, duplicates.
• Completion rate: how many requests the automation closes on its own and how many it passes to a person.

Mistakes to avoid

• Automating a muddled process: simplify first, then automate.
• No log: if you can’t see what the automation did, you won’t find out when it gets something wrong.
• Too much autonomy too soon: start with proposals approved by a person, and let autonomy grow with the results.
• Processing the same event twice: orders, payments and kitchen tickets must be processed exactly once, even if something is interrupted halfway through.

Want to find out which of your processes is worth automating first? Tell us how you work today.