How artificial intelligence and automation can streamline procurement

Automation in procurement

Pascal Bensoussan, chief product officer at ivalua, says businesses should start by building a single version of the truth for their procurement data.

“When you are running several ERPs [enterprise resource planning systems] with different logical organisations within those ERP systems, you have to somehow normalise and standardise all of the information into a more unified system,” he says.

Once the data is in a unified form, Bensoussan says the business can begin to use process control to automate parts of the procurement process. “We see a lot of opportunities to automate the process with better calculations and rules that automate certain behaviours,” he says. One example is the ability to send out a new questionnaire automatically if the procurement system identifies an exception in the order, or a recurring pattern in orders.

Clearly, the data needs to be in a consistent format and accessible. There also needs to be a master record for each supplier.

Karolina Hagberg Chinell, chief procurement officer at Ahlsell, a distributor of installation products in the Nordic region, says her company has focused on making the procurement system a system of record, which holds master data.

“To really work, you need a consolidated store with a single version of the system records.” This may start with basic supplier details, but over time, she says, “you need to be open and build integration with other data sources”. This helps to maintain the master data to ensure it remains up to date.

“AI can be used to cut down some of the manual processes that exist in procurement”

The implications of having this single version of the truth for all of a business’s procurement activities means it is easier to identify discrepancies and take advantage of volume purchases from the same supplier. While master data simplifies the work of the procurement team, Karl Poulsen, chief procurement officer at insurance firm Hiscox, a new customer of ivalua, says that once data is in machine-readable form and accessible, it is entirely possible to automate certain tasks.

Poulsen believes that while the industry is probably a long way from fully automated, artificial intelligence (AI)-based purchasing negotiations, automation can identify potential issues with an order far easier than if the task was done manually.

“AI can be used to cut down some of the manual processes that exist in procurement,” he says. “One example is to flag that there is a difference between the order and what has been received. You could then build a bit of code that says, ‘Well, we only ordered 10. If the supplier sent us 15 or something, maybe that’s their fault.”

Depending on the value and type of order, the supplier may be prompted to collect the surplus, or the procurement department could decide that since it is an item that is regularly ordered, an adjustment could be made the next time the item is required. “You could build logic into the system to do this,” adds Poulsen.

Towards full AI-based business software

Microsoft’s recent integration of generative AI into its Office 365 productivity suite shows how ChatGPT can be used to read and summarise documents and email messages, and produce action points from Teams meetings.

Many believe such technology is the way forward for business software, by offering people a way to automate manually intensive and often tedious tasks. The general consensus is that such AI systems can be deployed to augment human workers.

Cyril Pourrat leads procurement for telco BT through its procurement organisation, BT Sourced. Pourrat discusses the role of AI in a Computer Weekly podcast: “We want you to fully leverage AI, machine learning and the digital ecosystem to position ourselves differently compared to other procurement companies. We truly believe that digital is the way to go. This is the way technology will enable procurement to change.”

Pourrat believes enterprise software needs to be “consumer grade” with “everything to be achieved in two clicks, and ideally just by speaking”. In fact, he regards the use of natural language queries in enterprise software as a means to simplify work.

Looking beyond simplifying work, in a book co-authored with Don Scheibenreif, Gartner analyst Mark Raskino discusses the evolution of machines that are capable of making purchasing decisions autonomously.

The book, When machines become customers, explores a world where AI-enabled non-human customers interact with businesses. Some products are already sold this way, such as the office printer, which knows when it is low on ink and automatically reorders from the supplier.

Raskino believes much of the technology to enable machine-based purchasing is already in place and all that is really required is a willingness to accept the technology to make purchases on behalf of people.

“We need a list of providers of a thing. The search engine capability gives you that. You can quite quickly see how AI software can come up with pretty good lists, which is the first part of doing any kind of comparison shop. And comparison shopping itself is already a fairly well understood area,” he says.

People readily use comparison sites to find new providers for gas, water, electricity, insurance and mortgages. “You could shop for these things yourself and get a better deal,” he says, “but we know that you can’t be bothered and it’s hard work, so the algorithm on the comparison site is now working on your behalf, shopping between all the utilities, looking for the best deal.”

While people understand the price comparison site will take a small percentage, Raskino says they will accept this if they are getting an overall saving. “I think we can see the evolution of smart buying,” he adds.

What is clear from the experts Computer Weekly spoke to is that automating procurement starts with a clean data store that holds one version of the truth, in terms of supplier management. So long as the data remains clean, it can form the basis on which simple automation scripts can be run, to manage exceptions without human intervention, highlighting discrepancies in orders and purchases.

AI offers a way to predict demand, enabling an organisation to take advantage of volume discounts or to work around supply chain issues. Beyond this is a world that Gartner’s Raskino envisages, where machines are authorised to make smart purchasing decisions on behalf of people. Much of the technology components to achieve this exist today. The one barrier is cultural. Will people be comfortable allowing AI to make purchases on their behalf?