Common mistakes when applying AI to procurement
Procurement teams overlook data inconsistencies when applying AI, leading to flawed insights and poor decision-making, whereas a thorough data review, such as a category manager checking supplier data, can help identify discrepancies and ensure AI-driven procurement decisions are reliable and effective.
A procurement dashboard that nobody opens is a common sight in many organisations, a testament to the failure of AI deployments in this critical function. The business consequence is clear: procurement decisions are not being informed by data, leading to inefficiencies and wasted resources. The mechanism behind this failure is often a series of mistakes that recur in procurement deployments.
Automating a process nobody mapped
One of the most significant mistakes is automating a process without first mapping it out. This leads to a situation where the AI system is trying to optimise a process that is not fully understood, resulting in suboptimal outcomes. For instance, a category manager may try to automate the process of selecting suppliers without first documenting the criteria used to evaluate them. This can lead to the AI system making decisions that are not aligned with the organisation's goals.
A shift supervisor at a manufacturing plant, for example, may try to automate the process of ordering spare parts without first mapping out the workflow. This can result in the AI system ordering parts that are not needed, leading to waste and inefficiency.
Treating purchase orders as complete when it is not
Another mistake is treating purchase orders as complete when they are not. This can lead to a situation where the AI system is trying to analyse incomplete data, resulting in inaccurate insights. For example, a procurement analyst may try to use AI to analyse purchase orders without first checking if they have been fully approved and executed. This can lead to incorrect conclusions being drawn about the organisation's procurement patterns.
This mistake is often made because people assume that purchase orders are always complete and up-to-date, which is not always the case. In reality, purchase orders can be delayed, cancelled, or modified, which can affect the accuracy of the data.
Giving the capability only to the analyst who already had it
Most people get wrong the idea that giving AI capabilities to the analyst who already has access to the data is sufficient. This is wrong because it does not address the needs of other stakeholders who may also benefit from the insights provided by the AI system. For instance, a procurement manager may give access to the AI system only to the procurement analyst, without considering the needs of other teams, such as the finance team or the logistics team.
This limited approach can lead to a situation where the benefits of the AI system are not fully realised, as only a small group of people are able to use it. To get the most out of AI, it should be made available to all relevant stakeholders, including those in operations and other functions.
Declaring success against a metric with no baseline
Declaring success against a metric with no baseline is another common mistake. This can lead to a situation where the organisation is not able to accurately assess the impact of the AI system, as there is no reference point to compare it to. For example, a procurement team may claim that the AI system has reduced costs by a certain amount, without first establishing a baseline for what the costs were before the AI system was implemented.
This mistake can be avoided by establishing clear metrics and baselines before implementing the AI system. This will enable the organisation to accurately assess the impact of the AI system and make adjustments as needed. If you are unsure about how to establish metrics and baselines, you can contact us for guidance.
When AI is not the right choice
Finally, it is worth recognising that AI is not always the right choice for procurement. There are certain situations where other approaches may be more effective, such as when the data is incomplete or inaccurate, or when the procurement process is highly complex and nuanced. In these situations, it may be better to use other tools and techniques, such as manual analysis or process mapping.
For example, if the organisation is dealing with a highly complex and bespoke procurement process, it may be better to use a manual approach, at least initially. This will enable the organisation to fully understand the process and identify areas where AI can be effectively applied. By taking a thoughtful and considered approach to AI, organisations can avoid common mistakes and get the most out of this powerful technology.
Frequently asked questions
What are common mistakes made when applying AI to procurement?
Mistakes include automating unmapped processes, treating incomplete data as complete, and limiting AI access to only certain stakeholders, which can lead to suboptimal outcomes and wasted resources.
How can procurement teams avoid declaring success without a baseline metric?
Establishing clear metrics and baselines before AI implementation enables accurate assessment of the AI system's impact, allowing teams to make informed adjustments and improvements.
When is AI not the right choice for procurement?
AI may not be suitable for highly complex or nuanced procurement processes, or when data is incomplete or inaccurate, in which case manual analysis or process mapping may be more effective.
How can organisations get the most out of AI in procurement?
By taking a thoughtful approach, mapping processes, reviewing data, and making AI accessible to all relevant stakeholders, organisations can avoid common mistakes and maximise the benefits of AI in procurement.