Where should you start with AI in your SME? 7 processes to look at first

AI is everywhere. That quickly creates the feeling that your business needs to “do something with AI”.

But where should you start?

Not with ChatGPT. Not with a chatbot on your website. And not with a list of the latest AI tools.

The best place to start is much simpler: look at the work that already happens today.

Where does your team lose time? Which tasks repeat every day? Where is information copied manually? Where do customers wait longer than necessary for an answer?

Those are usually the most interesting opportunities.

In this article, we look at seven types of processes that are a strong starting point for automation and AI in many SMEs.

1. Processing incoming enquiries

Many companies receive enquiries through different channels:

• a website form;
• email;
• phone;
• WhatsApp;
• social media;
• an external platform.

Someone then has to read the request, understand what it is about, ask for missing information and enter everything into the right system.

At a small scale, that seems harmless. But when it happens dozens of times every week, a lot of time disappears into administration that adds little direct value.

A smarter workflow could:

1. recognise and classify the enquiry automatically;
2. extract the most important information;
3. check whether anything is missing;
4. ask the customer for additional information;
5. create a contact or opportunity in the CRM;
6. assign the enquiry to the right colleague.

AI becomes particularly useful when requests contain a lot of free text, documents or different wording.

2. Preparing quotations

Quotations are still surprisingly manual in many SMEs.

An employee gathers information from emails, notes, price lists and previous quotes. A document is then assembled, checked and sent.

Not every part of this should be handled by AI.

A combination often works better:

• fixed rules for prices and margins;
• integrations with CRM or ERP for customer and product data;
• templates for structure;
• AI to summarise free text or prepare a first description;
• human review before anything is sent.

The goal is not to remove the salesperson from the process.

The goal is to reduce copy-paste work so they can spend more time with customers.

3. Processing documents and attachments

Invoices, purchase orders, technical sheets, contracts, CVs, requests and spreadsheets often arrive by email.

Someone opens the document, finds the relevant information and enters it again in another system.

This is an interesting area for automation.

Modern systems can read documents, structure the data and trigger the next step.

For example:

Email received → attachment recognised → data extracted → customer or project matched → information stored → employee alerted if something is uncertain.

Not every document should automatically be trusted. Financial, legal and operationally critical information often still requires human validation.

4. Following up with customers

A quotation has been sent. What happens next?

In many companies, follow-up depends on someone remembering to contact the customer again a few days later.

That works while the number of opportunities remains small. Once volume grows, leads start slipping through the cracks.

A simple automation can already make a big difference:

• reminders after a set number of days;
• automatic tasks for salespeople;
• personalised follow-up emails;
• different follow-up flows depending on the type of enquiry;
• alerts when an important customer does not respond.

AI can help prepare the communication, but the biggest gain often comes from having a reliable process.

5. Finding internal knowledge

“Where is that document?”

“How did we handle this for that customer?”

“What did we agree on in that project?”

Companies build up a lot of knowledge over the years across SharePoint, Google Drive, Dropbox, email, Notion, Teams, CRM systems and local folders.

The problem is often not that the information is missing.

The problem is that nobody can find it quickly.

An internal search or AI assistant can be valuable when employees regularly need to retrieve the same information.

But first you need to know which sources are reliable, who is allowed to see what and how documents are maintained.

Putting a chatbot on top of poor information does not solve the underlying problem.

6. Reporting and administration

Many reports are still built by combining information from several systems in Excel.

That is often a sign that those systems do not communicate well enough.

Examples include:

• combining CRM sales figures with invoicing data;
• gathering project status from different tools;
• manually preparing weekly reports;
• copying figures into dashboards;
• comparing performance across teams or locations.

Automation can collect and prepare this data automatically.

AI can then help summarise trends or exceptions, but accurate data remains the foundation.

7. Planning and handovers between employees

A lot of time is lost when work moves from one person to another.

Sales closes a deal and sends an email to operations.

A planner reads the email and creates a task.

A technician receives a PDF.

After the intervention, notes are retyped for invoicing.

These are exactly the workflows where integrations can make a significant difference.

A good system lets information be entered once and then move through the process automatically.

Do not start with the technology

The most important question is not:

“Which AI tool should we use?”

A better question is:

“Which task costs our team a lot of time without creating much additional value?”

Make a list of processes that:

• happen frequently;
• require a lot of manual input;
• are prone to errors;
• involve multiple systems;
• make customers wait;
• depend heavily on one employee.

Start there.

Some problems need AI. Others are better solved with a simple integration, workflow or small custom application.

And sometimes the best solution is simply to simplify the existing process.

A practical first step

Choose one process.

Not your entire company.

Then map:

1. how the process starts today;
2. which steps an employee performs;
3. which systems are involved;
4. which information is entered more than once;
5. where errors or delays occur;
6. what absolutely needs human review.

Only then should you choose the technology.

That is also how we approach automation at BOUT LAST WEEK: understand the process first, build second.

Want to know where the biggest automation opportunities are in your company? We are happy to look at the repetitive processes that currently consume the most time.