Artificial intelligence in procurement

Artificial intelligence in procurement

• Artificial intelligence (AI) is a quickly growing field that can be applied to many areas of business, including procurement.

• AI can increase efficiency by quickly absorbing, analysing and summarising information. 

• AI should never fully replace human oversight – automation of procurement tasks can lead to inaccessibility and discrimination if not properly managed. 

• Organisations using AI in procurement should seek to understand the risks, mitigate them where possible, and avoid using AI in areas where risks cannot be mitigated. 

What is artificial intelligence?

Artificial intelligence (AI) refers to machines that have been trained to perform tasks that require human intelligence and problem-solving skills. AI often involves the development of algorithms, modelled on how humans make decisions, and ‘trained’ from banks of data to be increasingly accurate over time. 

Our resource ‘Artificial intelligence and emerging technologies’ has more information about AI and disability inclusion. 

How do businesses use artificial intelligence in procurement?

Artificial intelligence is starting to appear in numerous areas of business operation, including procurement. The Chartered Institute of Procurement and Supply has identified three broad categories of AI uses in procurement: 

Machine learning

AI can analyse data sets using algorithms to identify patterns and make predictions about future trends to support and advise decision-making. It can do this far more quickly and accurately than humans can. 

In procurement, AI could look at data related to spending or service level agreements and spot recent trends. It could then recommend actions based on how they predict those trends will develop. 

Natural language processing

Natural language processing AI can interpret human languages, and are often able to mimic human language fairly convincingly based on the probability of one letter or word following another in any given context.  

In procurement, natural language processing AI might be used to analyse a Request for Proposal or a contract and summarise the key information. 

Robotic process automation

Robotic process automation AI creates algorithms that automate repetitive, simple tasks, often completing them more quickly and accurately than a human could. 

In procurement, this could be used to automate the processing of invoices or purchase orders. 

How could AI in procurement affect disabled people?

Opportunities 

Machine learning AI can analyse large amounts of data, which could help to highlight problems with accessibility and inclusion – for example, through looking at reviews or complaints relating to a product or service. This would give organisations regular and accurate information about the accessibility of their products and services that they could use as evidence if the issue stems from a supplier. AI could also offer recommendations on how to improve the situation (although human procurement experts should always assess and research these). 

An organisation might use natural language processing AI to generate summaries of how procurement-related documents affect disability inclusion. That would allow procurement teams to compare the inclusiveness of suppliers and bids more easily. 

Robotic process automation could automate procurement processes that a disabled employee finds difficult. For example, automating calculation tasks that an employee with dyscalculia may struggle to complete. 

Risks

AI can replicate biases that are present in its training data. AI is only as good as the information that has gone into it. If AI is trained on information that contains disability bias, it will replicate this bias. For example, if it sees that previous procurement trends have selected suppliers with poor disability inclusion records, it will see that as a feature to look for in future suppliers, and thus embed that practice. It may notice that bids using language such as ‘inclusive’ and ‘diversity’ have not been successful in the past, and associate those words with a lower value in its ranking. 

Additionally, if an organisation’s review or complaints processes are inaccessible, then disabled people may be less able to make complaints. As a result, the AI analysis of user experiences will only include feedback from people not experiencing the barrier. 

AI also sometimes ‘hallucinates.’ This is the name for when it produces false information, or perceives patterns that do not exist, resulting in inaccurate or nonsensical outputs. This feature of AI also means uses can manipulate it, for example by feeding it misleading information, to divert it from its original intended purpose.  

Therefore, expert procurement teams should always oversee and factcheck any AI analysis. 

Recommendations for businesses using AI in procurement

• Start small – introduce AI to simpler and more discrete parts of the procurement process first. Do not make it an integral component before you are sure you can use it safely. You must be sure that AI in procurement does not lead to worse outcomes for disabled end users or disabled-owned suppliers. 

• Ensure the AI you are using has been ‘trained’ in the way you want – for example, using training data and operational parameters that support and demonstrate disability-inclusive practices.  

• Conduct algorithmic auditing on a regular basis to assess whether the AI is displaying bias. This should be done by an independent auditor. 

• Capture as much data as possible about disabled end users’ experiences. Monitor changes and trends, and intervene if negative outcomes are happening. 

• AI is no substitute for human expertise – ensure strict oversight of any AI processes by competent professionals. 

AI is itself an outsourced product for most organisations. Apply the same high standards of disability inclusion to AI in procurement as you would to any other type of outsourced product or service. If it does not work for disabled people, fix it or do not use it.