SAP + AI: how the German Mittelstand automates sales in SAP

In many SMEs, SAP is the backbone for orders, deliveries and invoices. Yet the sales work in front of it often still runs on phone, Outlook and Excel. This article shows where AI sales automation with SAP makes sense, which tasks should stay in the ERP and how to set up a first project without rebuilding your system.

Why sales often gets stuck in an SAP environment

SAP is built to record business transactions reliably: customer master data, terms, orders, delivery notes. It is rarely designed for the stage before the order. Whether you run SAP Business One, an older ECC system or S/4HANA, the pattern is similar. Enquiries come in by phone and email, someone in inside sales reads them, looks up item numbers, checks availability and finally types a quote into the system.

Take a metalworking company with 80 employees. Inside sales consists of three people. A large share of their calls isn't about new business but about questions like “When will my delivery arrive?” or “Could you send me the invoice again?”. Each of these questions is quickly answered by looking it up in the system. Added together, though, they mean the very enquiries that could turn into new orders are left waiting.

Where AI sales automation with SAP comes in

The most important principle: SAP remains the leading system. The AI works in front of it, as an extra layer for communication and preparation. It takes enquiries, understands them, collects missing details and hands over a clean record. Bookings are still made in SAP, according to your rules and approvals.

Task	What the AI can take over	What stays in SAP or with people
Order status enquiries	identify the caller, match the order, read out the delivery status from an approved interface	order data, delivery dates, special cases
New enquiries	capture need, quantities, preferred date and contact details in a structured way	review, costing, preparing the quote
Quote preparation	read email enquiries, suggest item numbers, prepare a draft	terms, approval, sending
Follow-up	reminders about open quotes, getting feedback, arranging a call-back	negotiation, price concessions
Scheduling appointments	put visits for field sales into the calendar	route planning, prioritising key customers

Our article on automated quote creation with AI describes in detail what automatic quote preparation can look like.

Architecture: directly to SAP or via a CRM

Option A: the AI talks to SAP

Depending on the system, SAP offers various interfaces, such as OData services, BAPIs or IDocs. A direct connection is possible but almost always requires your SAP consultancy. For read access, for example to the delivery status, the effort is more manageable than for write access. Don't let the AI create orders directly until you have gained some experience with the system.

Option B: a CRM as an intermediate layer

Many SMEs use a CRM such as Microsoft Dynamics, Salesforce or HubSpot alongside SAP, already connected to the ERP. In that case it is often easier to connect the AI to the CRM. Leads, call notes and tasks end up there, and the existing sync with SAP does the rest. You will find notes on Dynamics in the article Microsoft Dynamics 365 and AI for SMEs.

Which option fits depends less on the technology than on your organisation. If sales already lives in the CRM, the AI should dock there too. If SAP is the only system everyone uses, a lean read-only connection for status updates is a good way in.

A first project in five steps

1. Count enquiries for a week. Have inside sales keep a tally of what calls and emails are about: status, new enquiry, complaint, other. This list is your most important basis for decisions.
2. Choose one use case. Status enquiries are often a good fit because they are clearly defined and the answer is in the system.
3. Clarify data access with your SAP consultancy. Which data may the AI read? Via which interface? How is a caller reliably identified before they learn order details?
4. Define handover rules. For complaints, pricing questions or angry customers, the AI puts the call straight through or schedules a call-back. Write these rules down in advance.
5. Pilot for four to six weeks. Then check how many enquiries were answered in full, how often calls were passed on and what inside sales did with the time saved.

Clarifying legal issues, the works council and data protection

As soon as the AI processes customers' personal data, you need a data processing agreement and clarity about where the data is stored. Make sure it is hosted in the EU. With status updates, identification is an extra issue: without a customer or order number, nobody should be told delivery addresses or invoice amounts.

If your company has a works council, involve it early. Systems capable of monitoring employees' performance or behaviour are subject to co-determination. This applies, for example, to reports on who handled how many enquiries. An open conversation at the start saves delays later. Customers who talk to an AI should also be told so; the EU AI Act sets transparency obligations for this.

When the effort isn't worth it

• Your business consists of a few large framework contracts and enquiries are rare. Then a well-organised inside sales team is the better investment.
• Your master data is out of date. An AI that reads out wrong delivery dates does more harm than an answering machine. It is worth looking at preparing your sales data first.
• An SAP migration is imminent. Don't build interfaces for a system that will be replaced within a year. A channel without a system connection, such as structured capture of new enquiries, can still make sense.

Frequently asked questions

Do I absolutely need S/4HANA for AI in sales?

No. What matters is whether your system provides an interface for the data you need. Older systems or Business One can be connected too, although the effort differs. Your SAP consultancy can usually assess this quickly.

May the AI create orders in SAP itself?

Technically, yes. At the beginning, though, an intermediate step is advisable in which an employee reviews and approves the prepared order. That way you keep control over terms and special cases.

What does an SAP connection cost?

That depends on the system, the interface chosen and your service provider, and can't be put in a single figure. Get a quote for a clearly described use case, such as read-only access to the delivery status.

What happens to inside sales?

The team doesn't become redundant, it just gets interrupted less. At a time when inside sales positions are hard to fill, that is the real reason for the project for many businesses.

Conclusion

AI in an SAP environment works best as a front office. It takes enquiries, sorts them and delivers clean data, while SAP remains the system where bookings are made. If you start with a narrowly defined case such as status enquiries, you gain experience without touching the ERP. The AI employees from Neurobots answer calls, emails and chats around the clock, GDPR-compliant with servers in Frankfurt, and pass enquiries to common CRM systems. The Digital assistant for service centres shows what this looks like for enquiries about orders and service.