Cognitive Procurement — What Are the Implications?
After first proving its intelligence by beating humans at various games (Chess in 1996, Jeopardy in 2011, Go in 2016 and again in 2017, and Poker very recently), artificial intelligence (AI) is now “blind to the color of the collar”, be it white or blue. It is making its way into the knowledge worker’s workplace, and the stakes are getting high.
People & Technology, not People vs. Technology
The superintelligence described in science-fiction movies and literature has yet to be realized; specialists estimate that singularity — or machines becoming more powerful than humanity — is still a distant concern. Nevertheless, many regard the success of AI as an indication that computers are, in fact, already surpassing humans in every aspect, and as such are destined to replace them.
In the field of Procurement, this cognitive collaboration is known as Cognitive Procurement, and it is capable of far more than the automation of manual and repetitive tasks, which has been around for years.
Implications of Cognitive Procurement
In the last years, consumer demands have been evolving tremendously in terms of immediacy, convenience and personalization, to name just a few. Supply chains have to cope with these new expectations while operating in a global market. It means that Procurement professionals have to manage more complex scenarios in a context with a lower tolerance for mistakes and slowness. To address this challenge, AI is their best ally, and Cognitive Procurement is the answer.
Cognitive Procurement is a new type of cooperation between people and machines that is designed to:
- liberate Procurement teams from even more mundane tasks
- assist them in grasping a more complicated and disrupted business environment
- enable Procurement organizations to do things that were previously impossible
Applications of Cognitive Procurement
Enhancing decision processes is the most visible use of Cognitive Procurement. Supply chain professionals can be overwhelmed by the ever-increasing amount of data and information they must manage, and therefore have to take into account their own actions and decisions. However, with continual increases in their computational power, machines can help. They can make sense of a much larger amount of data than we can, and they do it faster. And, because they process more data and run elaborate algorithms, they are able to generate new knowledge by identifying trends, correlations and tendencies that we cannot grasp.
This is particularly the case with Machine Learning (ML), a branch of AI. Machines are not programmed to do a task; they are programmed to learn. Data analysts teach the computer using data samples and by correcting the AI’s mistakes. Alternatively, the machines learn from historical data to build a model that it applies to any new data with which it is presented.
Experience in domains like healthcare, such as the analysis of x-ray images, has revealed that machines can create new knowledge by identifying correlations that were, until now, unknown. Despite this, it is of notable importance that humans remain involved, because ultimately, the results from these machines are recommendations for actions or decisions, rather than nonnegotiable facts.
In addition to assisting in the analysis process and being a recommendation engine, Procurement technology can become assistants à la Siri or Cortana. As a result, what is possible with phones is also now conceivable in a business context. Machines can understand written and spoken language, and Natural Language Processing (NLP) is a field that has seen enormous progress in the last years. NLP opens the door to new management methods and ways for people to interact with Procurement technology. An application that comes to mind involves transposing the aforementioned assistant to the Procurement context. Such a Procurement assistant could:
- help Procurement teams stay on top of their daily operations
- guide internal customers seeking to make a purchase
For example, let’s imagine a situation where a natural catastrophe impacts a region. The Procurement assistant learns immediately about the event because it is connected to external sources of information (there are many providers of supply chain risk management technology that enable that) or to public sources of information (news outlets or social media, for example). It will then:
- list all impacted suppliers and what the organization buys from them
- identify existing alternate sources
- prepare a report with prioritized recommendations for action
Once the report is rendered, the Procurement assistant will proactively start a conversation with the buyer in charge. A conversation that would look like this:
There are many other use cases. In indirect, the most intuitive one is guided buying. Internal customers discuss with the AI via chat instead of using the classic catalog or free-text that are at the core of most eProcurement solutions.
These are just a few potential upsides that AI represents in the digital transformation of Procurement, with the main business benefits being to:
- make jobs more human than they were by allocating tasks to the human workforce that machines cannot perform
- fuel effectiveness via new efficiencies (i.e. time that has been freed up is dedicated to more valuable tasks)
- produce better results (i.e. providing new knowledge, detecting and identifying new trends and correlations, making contextual decisions and more)
However, even if the benefits are tangible and numerous, the use of AI also raises some important questions.
Challenges & Risks
Machine intelligence quickly moving toward the ability to independently make decisions raises concerns over control, accountability and trust. Who will be responsible if a machine makes a bad decision?
Furthermore, intelligent machines analyze a situation and arrive at a result or recommendation based on a particular type of logic (e.g. coded or learned). People only see the outcome and not how the AI came to it, similar to a black box. Because of this, blindly trusting the AI can be a source of risk, which is exacerbated by the fact that machines are not perfect either. To mitigate these risks:
- machines must not only present results, but also briefly explain how they arrived at them (similar to the discussion in the chat example above regarding how the machine formed its recommendations)
- people should demonstrate critical thinking and have the final say
Another challenge is the necessary investment in developing AI, and it is not programming machines to learn that represents the largest investment. In fact, it is the cost of gathering and obtaining enough data to train the machine that often has an enormous cost. The more data there is, the better the learning phase will be. Furthermore, the qualitative aspect is as important as the quantitative one because learning from bad data will lead to poor results.
AI & ML: The Future of Procurement
AI and ML are certainly the future and represent a substantial improvement in an organization’s capabilities. It is, therefore, important that CPOs and Procurement teams get familiar with Cognitive Procurement and assess its potential. They should then make an informed decision about when to employ AI by understanding its consequences and requirements.
For a printable pdf of "Cognitive Procurement — What Are the Implications?", click here.
Originally published at www.ibisworld.com on March 27, 2017.
