There are many companies that have an extensive history of unsuccessful attempts at implementing technology tools in procurement. Not only do these organizations suffer from lackluster results, but their procurement teams end up disillusioned.
Now, chief procurement officers increasingly have set their sights on adopting artificial intelligence (AI) tools. Most organizations have yet to fully integrate these technologies into their core procurement processes though, only pursuing them as far as the pilot phase or obtaining a few key user licenses. Going forward, it’s clear that to succeed companies must go beyond merely "integrating AI" — especially for generative AI, a technology that’s rapidly evolving and could have wide-reaching potential in procurement. Instead, new tools must serve a larger purpose within the broader scope of procurement activities.
How to manage common obstacles when implementing AI in procurement
Looking at procurement from an external perspective, a significant number of processes emerge as promising candidates for AI tool application. However, it is essential to temper our enthusiasm and examine why past failures occurred, to avoid repeating them in the future. The causes fall into four categories: purpose, IT, people, and data:
Purpose — ensuring effective use cases and cost management
- Weak use cases: Digitization can greatly enhance the way of working for procurement in selected process steps. But the digitization use cases have to be chosen wisely to support the buyers in the end — reducing unnecessary tasks and supporting daily work without adding new data entry or control activities.
- Costs: Implementing AI can be expensive, including costs for the AI tool itself, data preparation, infrastructure upgrades, and training. Therefore, it's important to carefully plan and budget for the implementation.
IT — building a solid foundation for ai implementation
- Technical infrastructure: The IT infrastructure of your organization needs to be robust enough to support AI applications. That includes the necessary hardware and software, reliable internet connectivity, and the integration of the AI tool with your existing procurement systems.
- Regulatory compliance: Depending on your industry, you may need to consider regulatory compliance issues related to the use of AI. For instance, certain types of data processing may be restricted or require specific permissions.
- Technology deficits: Many AI tools on the market serve a good purpose but are not always mature enough, integrated well, or designed to serve the dedicated purpose. The chosen technology has to match the use case in full, being easy to integrate and simple to use.
People — Managing change and developing AI skills
- Change management: Like any new technology or process, AI implementation may be met with resistance from employees. It's important to effectively communicate the benefits of AI and provide adequate training to staff to facilitate a smooth transition.
- Capability and skills, including those specific to AI: In addition to traditional training, AI tools require instilling an understanding of the workings and limitations of the specific AI application being implemented. This knowledge is fundamental so that users know when AI outputs need to be questioned critically.
Data — ensuring data quality and security
- Data quality: AI tools and machine learning models are only as good as the data you feed them. Incomplete, outdated, or incorrect data can lead to erroneous outputs and misguided decision-making. Thus, it's crucial to ensure that your procurement data is well-structured, clean, and comprehensive.
- Data privacy and security: The procurement process often involves handling sensitive information such as supplier details, contracts, and pricing. Hence, robust data security measures are needed when integrating AI tools that process and analyze that data.
When it comes to implementing AI-based procurement tools, however, some of these obstacles can be more easily overcome. For instance, advanced AI can work with incomplete and false data by proactively identifying and actively managing it.
On the other hand, some of the other causes for unsatisfactory AI implementations mentioned above have risen in importance. Taking the example of data privacy and security, the adaptation of open language models represents significant challenges. Handling of sensitive and competitive data with these tools often requires either a major increase in costs (say, to set up a separate mode) or limitations to the data’s functionality (such as having no direct access to the open-source internet and therefore no access to real time, outside information).
Use case-driven AI development and implementation are key to success
When analyzing various AI implementations, a pattern emerges. They tend to fail when two critical criteria are not met:
- The selected use case did not meet the buyers’ needs to reduce their workload and support their everyday procedures.
- The scope of the tool did not match the use case. It tried to solve too many problems and was not focused on the proper issue.
Despite the challenges and obstacles, some companies and teams have found ways to use AI tools in their daily processes. Successful adoption depends on matching an available tool and a relevant process step.
Aligning AI tools with relevant procurement processes
Step 1: Identify procurement processes with potential
The first task is to identify the dedicated process step on which AI can generate the most value-add to procurement. Our analysis of the procurement process showed considerable potential for AI-driven value-add across all process steps, varying by the kind of AI. (See Exhibit).
Descriptive and diagnostic AI, as well as predictive and prescriptive AI, are already well established in procurement, for use cases including spend analysis, supplier risk management and monitoring, and demand pattern analysis.
When we asked generative AI itself — in this case our Marsh McLennan solution LenAI — how relevant generative AI currently is for procurement, the response was “limited relevance” on all process steps. Still, we see high generative AI value-add for procurement in the future. Operational procurement especially can benefit from generative AI applications, in use cases such as the creation of RFPs (request for proposals) for a new commodity, AI negotiations with suppliers for small and common parts (c-parts), and automatic replies to supplier notifications on price changes.
Step 2: Focus on use cases already supported by mature AI
The second key step is to verify that AI technology for a selected use case is available, which, especially for new tools such as large language models, may be unlikely. We suggest matching the long list of possible use cases against the maturity of digitization solutions (which can be assessed by asking how well they overcome the obstacles mentioned earlier).
It’s also necessary to ask if there is a designated IT solution already market-ready and available, taking into account the amount of customization needed. The easier the technological realization will be, the higher the acceptance and likeliness for success. Additionally, you should evaluate early on if your technology partner is also your implementation partner, which will reduce the effort.
Once the use cases with value-add to procurement are defined, further implementation steps will come. Among others, these include verifying the use cases regarding data availability, the fit into the IT landscape, and the fit into the company’s specific procurement processes.
Take advantage of AI’s momentum immediately
Digitization in general and generative AI specifically has value-add to procurement processes. Its implementation has common obstacles, which companies can overcome with the right strategy. The time to get involved with AI is right now: Solutions both large and small are available for the procurement process to facilitate everyday work and generate value. We currently see many companies integrating new ideas, priming the procurement landscape to change rapidly in the coming days.