In short
- AI is strong at reading unstructured input, drafting, matching and summarising.
- It is weak at exact calculations without a data basis and at decisions that require accountability.
- The best starting points are frequent, time-consuming processes with a clear result and a person who approves.
- Some uses, such as filtering job applications, count as high-risk under the EU AI Act and need particular care.
Where AI really helps in a business
Almost every sensible use follows the same pattern: information arrives unstructured, as an email, PDF, photo or voice message. Someone has to read it, understand it, combine it with data from a program and turn it into something: a reply, an entry, a proposal. That groundwork is exactly what AI can take over. The decision stays with a person.
The following examples describe typical processes, not client projects. They show where AI comes in within each department. The same rule applies to all of them: the AI prepares, a person checks and decides before anything goes out or is posted.
Twelve use cases by department
1. Sales: preparing quotes
The AI reads enquiries, spots missing details, suggests items from your price list and creates a draft. More in our article on automating quotes with AI.
2. Sales: qualifying leads
New contacts are enriched with publicly available company data, classified by your criteria and presented to sales with a short summary.
3. Customer service: answering enquiries
Enquiries are read, assigned and answered with the right data, first as a draft for approval. More in our article on automating customer enquiries. The page on AI customer care shows how we set this up.
4. Purchasing: comparing supplier quotes
Quotes in different formats are extracted and brought into one consistent table: prices, delivery times, terms, deviations from the request.
5. Purchasing: matching order confirmations
Supplier confirmations are compared with the purchase order. Deviating prices, quantities or dates stand out immediately.
6. Accounts: capturing incoming invoices
Invoices as e-invoices, PDFs or scans are read, matched with purchase order and delivery note and handed over ready for processing. More on automating incoming invoices.
7. HR: answering staff questions
An assistant answers questions about holidays, travel expenses or procedures based on your internal rules. New staff find documents faster.
8. Controlling: writing reports
Figures from accounting, ERP and CRM are collected, anomalies flagged and a draft of the monthly report written in plain words.
9. Logistics: reading delivery notes and freight documents
Details from delivery notes, consignment notes and customs documents are captured and transferred into the ERP, with inconsistencies flagged.
10. Marketing: copy and translations
Product texts, articles and translations are created as drafts in your tone. A person checks facts, prices and claims before publication.
11. Management: minutes and decision papers
Meetings are summarised, tasks recorded with owners and deadlines, and documents condensed into a short decision paper.
12. Engineering and service: making knowledge findable
Manuals, data sheets and internal instructions become searchable. Staff ask questions in their own words and get answers with sources.
What does not work well
- Exact figures without a data basis: prices, quantities and deadlines must come from your programs, not from the language model.
- Decisions involving accountability: dismissals, credit decisions or legal assessments stay with people.
- Recruitment: AI that filters or evaluates job applications counts as high-risk AI under the EU AI Act. More under risk categories.
- Unclear processes: if nobody can say exactly how a process works, no AI can take it over reliably either. Clarification comes first.
The technology behind it
- Language models: they read and write text. They are connected through the providers’ interfaces, with data processing in the EU where possible.
- Automation tools: programs such as n8n, Make or Power Automate connect the steps and your systems. More under n8n in your business.
- Knowledge bases: your own documents are prepared so that the AI can give answers with sources instead of writing freely.
- Your programs: ERP, CRM, accounting and document storage stay. They are connected through interfaces.
How to find the right starting point
- Collect time sinks: ask each department which activity costs the most routine time.
- Assess: how often does it occur, are there clear rules, is the data digital, how much harm would a mistake cause?
- Choose one process: frequent, time-consuming, low-risk and with a person who approves.
- Test phase: the solution first runs alongside the existing work, with clear metrics.
If you want to tackle several departments at once, a short AI consulting engagement that assesses processes and sets an order of priority pays off.