In the commercial departments of mid-sized companies, the same activities repeat day after day: capturing documents, checking invoices, comparing supplier offers, reconciling contract versions. Each individual task is manageable — taken together they tie up a substantial share of working time of experienced staff who are professionally qualified for more demanding work.
Three areas are particularly suitable for pragmatic AI support.
Accounting — pre-capturing incoming invoices
Incoming invoices are a classic. A PDF arrives by email or supplier portal, someone opens it, reads the relevant fields, enters them into the accounting system. The five to ten fields are always the same: supplier, invoice number, date, net, VAT, gross, due date, cost centre.
AI support takes over three steps:
- Reading: structured extraction from PDF, image or embedded text — even with changing suppliers and invoice layouts.
- Plausibility check: verifying that net plus VAT equals gross, that the invoice number follows the expected format, that the supplier exists in the master data.
- Suggestion: booking line, cost centre, due date — based on historical bookings of similar invoices from the same supplier.
The accountant sees the complete pre-capture, checks, approves or corrects. Responsibility stays — the mechanical typing disappears. In the projects we have seen, processing time per invoice drops from four to eight minutes to under one.
Purchasing — comparing supplier offers
When three suppliers send offers for the same item, three different table formats, three different line item structures and three different labels for the same thing arrive. The comparison costs time — and is prone to mis-allocation of items or overlooked discounts.
AI normalises: three differently structured offers become a comparable table. Similar items are mapped to each other, quantities converted (pack of 100 vs. piece), discounts and terms shown consistently, deviating delivery and payment terms highlighted.
The buyer sees the comparison at a glance and can make the commercial decision — instead of spending the first two hours just making the offers comparable.
Contract management — finding and comparing clauses
The real risks in contracts hide in the details — liability caps, termination notice periods, jurisdiction clauses, penalty clauses, service-level agreements. With every new contract, these points should be checked against in-house standards. In practice this happens only on a spot-check basis — because time is short.
AI can do two things here:
- Extraction and structuring: relevant clauses are identified in a contract and transformed into a standard form. Who is liable for what, at what level, with what exceptions? Which deadlines apply, with what triggers?
- Comparison against in-house standards: extracted clauses are compared with the minimum and maximum values defined by the company. Deviations are highlighted — with reference to the corresponding standard.
This does not replace legal review, but makes it faster and more systematic. What was previously checked against personal experience is now checked against an explicit reference matrix — which, as a side effect, forces that matrix to be defined in the first place.
How the activity profile changes
In all three areas the staff’s activity shifts in the same direction. Before: lots of entry, lots of transcription, little judgement. After: less entry, more judgement. Professional depth — distinguishing a critical invoice from an unclear one, recognising an aggressive contract, telling a truly good offer from one that merely looks cheap — becomes more visible and more in demand.
For staff, this is in most cases a win. For the company it is a shift of value creation: fewer hours on low-qualified work, more hours on judgements that require experience and context.
Where AI does not belong
Three areas in commercial operations stay human:
- Final approvals (payment release, contract signing, order placement) — responsibility belongs to the person with the authority.
- Negotiation decisions (which price is acceptable, which supplier to favour strategically) — AI delivers the comparison basis; the human decides.
- Conversations (with suppliers, customers, internal stakeholders) — the relational layer stays human; AI at most prepares the information.
This line is not drawn historically, but structurally: at these points it is about responsibility, not about case processing.
What becomes visible in the end
An AI-supported commercial department has less typing, less searching through PDFs, less double entry. What it has: better data quality, faster throughput, a clearer picture of its own risks. These are not the headlines AI projects usually produce — but they are the effects that, over time, make the biggest difference inside a company.