As a consultant on the subject of AI, I am attending many different companies. What I hear again and again across all departments are statements like “we have to digitize!”, “we urgently need AI!” and recently also “the pandemic is a wake-up call for us to urgently deal with new technologies!”. Especially in purchasing, the desire for new, “safer” methods is very loud! When I then ask where AI is to be used in concrete terms, statements come up such as: “To spread the risk in proactive risk management”, “To reduce time and cost factors” or “in optimized planning”.

And here we are right at the point: most buyers are eager to use AI and recognize its potential, but do not yet have a precise idea of what this might look like. The basic knowledge about the types of AI is missing. That’s why I always recommend companies consider a few general points before even getting into use case planning:

• Analysis: Look for processes with high operational costs that tie up a lot of employees, assess which automation category is suitable.
• Understand opportunities of AI, define pilot projects for yourself and do not lose sight of cost effectiveness.
• Build AI competencies internally but also work with specialized third-party vendors.
• Store granular data wherever possible – it is the fuel for AI applications.
• Combine existing detailed knowledge of own products and manufacturing processes with new AI applications.
• Get small tests up and running quickly; no huge investments are needed, but agility is a prerequisite for success.

What can actually be usefully implemented in purchasing?

When you start planning the use case, you should consider other goals in addition to ROI. These include the sustainability of the solution, the improvement of data quality and the development of new markets. Depending on the type of AI, different projects can be approached. The following use cases have proven successful so far:

– Delivery date forecasting using prediction leads to greater planning accuracy and thus a reduction in capital commitment.

– Initial document capture: Automatic matching of purchase orders, confirmations and invoices leads to the elimination of potential inaccuracies right at the beginning and shortens project lead times.

– Prediction of specific risk factors: Extremely accurate predictions can be made from historical data and qualitative and quantitative sources. Caution: When making predictions, careful consideration should be given in advance to which data (qualitative/quantitative) and factors (internal/external) are to be included. In practice, a combination of qualitative and quantitative data has proven to be the most reliable so far.

And then there is also the master class of all applications, suitable for a strong AI: The takeover of the entire disposition process.  Whether mails, faxes or incoming calls – regardless of the data medium, AI takes over all orders: It reads, analyzes, processes and finishes them on its own. Thanks to the system’s ability to think creatively, decisions are made independently and new solutions are found.

– Incoming customer orders (fax, mail, call) are accepted. In case of unknown customers, the AI automatically contacts them and exchanges master data.

– The AI checks the databases, the warehouse and the suppliers.

– The AI independently creates the entire process flow: Prices, material, suppliers and delivery dates.

– The AI enters all data into database.

– The AI informs the customer about prices, delivery dates and times.

– In case of Okay from customers, the AI organizes the whole order process autonomously.

This results in the following advantages for companies:

– Cost savings up to 90%

– Reduction of errors up to 85

– Acceleration of processes up to 80%

– Acting independently of people and departments

– Clear and structured overview of the status at any time