Analyst Eye on the 6 paths to GenAI implementations in procurement

Incorporating AI into business processes can unlock unparalleled insights and efficiency; we know that from the numerous solution demos we’ve witnessed. Two main trends appear to be driving the development and use of GenAI among the vendor community. Firstly, they can’t afford not to; it’s the future. Secondly, the reality that GenAI (especially ChatGPT) is opening new possibilities for procurement operations.

Whatever the driver, there are two technical realities:

• GenAI is lowering the barrier to entry for AI-enabled solutions. Current applications (ChatGPT and others) are easy to integrate. This means they can open new doors and possibilities for providers that don’t have the capability, expertise or the bandwidth to move forward with AI – it gives them a plug-and-play solution. Its interoperability makes it easier than ever for a procurement solution to embed AI. There is no requirement for developers, and its use case is very broad – you can feed it something, and it will reply with something. This third-party-type integration therefore extends their own solution.

• It is taking natural language processing (NLP) technology, like chatbots and negotiation bots, a step further by generating output that goes beyond simply asking questions or giving prompts. GenAI applications are creating new content based on existing data taken from a body of knowledge.

The reality is that everyone wants to leverage AI and GenAI in particular.

According to our analyst Bertrand Maltaverne, we have encountered six evolutionary steps to getting there, from the very shallow use case to the more advanced and autonomous.

1. The shortcut, and easiest way to adoption, is to append a chatbot to your procurement application – this is nothing more than a window to ChatGPT. And for some providers this is enough, since their technology has other strengths and their customers may not demand or require any more from AI. This type of use is what we call a ‘gimmicky’ approach, a shallow integration, because the user could use ChatGPT without integration. This is the basic level.
2. Then we have different levels where GenAI acts as a co-pilot (a recommender). Users can get an output from GenAI, but it is up to them whether they want to use it; the GenAI output is only informative and not used directly. Within the co-pilot strategy, implementation levels depend on the level of optimization and contextualization and whether the output from GenAI is tailored enough both to ensure quality and prevent certain risks like ‘hallucinations.’
3. The most advanced implementations are the ones where GenAI acts as an agent, a sort of virtual worker/colleague. GenAI is deeply embedded in the application and completes multiple and consecutive (complex) actions on behalf of users.

The next three levels of implementation touch on a much more evolved integration, enhanced perspectives and optimized prompts to allow for more actionable recommendations extracted from a vast source of knowledge. The six steps are explained fully in ‘The 6 different ‘flavors’ of GenAI implementations in procurement applications,’ which is free to read upon very basic registration.

At Spend Matters we’ve seen providers at all stages of this evolutionary process.

Responding to various types of demand, almost every vendor is investigating or launching GenAI features, in most cases using ChatGPT. As our analyst advises in the final thoughts of his ‘autogmentation’ series “there is a varying degree of value that customers can get from applications of GenAI in procurement technology, ranging from somewhat gimmicky to extremely valuable use cases. As such, potential customers need to move quickly with safe experimentation, but also exercise caution in making more strategic commitments without thinking through the broader ramifications.”

The use of GenAI in procurement solutions does not come without risk however. We will be covering the pitfalls to avoid when implementing GenAI in procurement use cases later this quarter.

In the meantime, visit our AI in Procurement page to get the most up-to-date insights on the uses and applications of AI in procurement.