AI Negotiation and Procurement
The art of the deal reinvented
Procurement professionals know too well that negotiation is a key aspect of the job. Over the years, this task has evolved from a simple buying function to overseeing an integrated set of management functions. What was once a price-oriented negotiation has become a much more complex situation. Indeed, Procurement professionals have more elements to take into consideration before making a deal (sustainability, brand image, etc.). With the rise of AI technologies, negotiation is no exception to the ever-growing possible automated tasks.
Procurement professionals might already familiar with bots. We all know that they are limited to very specific tasks. Indeed, they can only hold short conversations and perform simple tasks such as finding a product in the inventory. The idea of having a “smart” bot or let’s call it assistant is very challenging. To get chatbots to the next level they’ll need to acquire new skills — such as long-term memory and reasoning. Adding these new cognitive abilities is closer than you think. The human language and reasoning are actually very complex for a machine to understand and manipulate. Well, it’s about to change thanks to new AI technologies.
Startups and tech have made great progress in emulating human cognition and communication for automating business processes. Recently, researchers at Facebook Artificial Intelligence Research (FAIR) have open-sourced code and published research introducing dialog agents with a new capability — the ability to negotiate.
In case you didn’t know, bots that can negotiate are nothing new. There are very basic bots on eBay, that bid at the last possible second or others that bid for ad placement online. Perhaps, we could also these negotiation support systems, software that assists humans by suggesting win-win outcomes. These systems were somehow good but with AI, it’s on another level…
AI bots
Let’s start with technical details…
Using CUDA and NVIDIA GPUs, researchers trained a recurrent neural network by teaching it to imitate people’s actions. The models were trained end-to-end from the language and decisions that humans made, meaning that the approach can easily be adapted to other tasks. Reinforcement learning was then used to reward the model when it achieved a good outcome which prevents the AI bot from developing its own language.
To help build their training set, the team created an interface with multi-issue bargaining scenarios and crowdsourced humans on Amazon Mechanical Turk to negotiate in natural language to divide a random set of objects.
It’s actually quite an accomplishment because it is considered difficult is to keep bots on topic and prevent them from saying weird things.
The researchers used supervised learning and reinforcement learning together with hardcoded rules to force the bots to stay on task.
In the reinforcement learning method, the bots are encouraged to reach a deal and penalized with a negative reward when it fails to reach an agreement. The researchers trained the bot by collecting 6,682 dialogues between humans working on the Amazon Mechanical Turk platform.
A long short-term memory network (LSTM) encodes the course dialogue act and another LSTM decodes it.
The use of AI bots to negotiate with other bots or humans to crack the best deal possible is a promising next step for AI and bots. It is safe to assume that AI bots can potentially match or even outperform the best human negotiators.
How to negotiate?
Negotiation involves having a lot of information. The good news is that AI is based on data and can continuously learn. In a negotiation, it is key to know what parties want, what they don’t want, and how much they are willing to compromise. Moreover, your data will be useful to detect your best practices and immediately use them.
The experiment shows how bots can easily be trained to imitate the linguistic and logical patterns of humans and paves the way for AI bots to be used in real-world negotiations. In addition to the linguistic and logical ability, AI bots can also be trained to find the best deals available from various sources for any chosen product or service). This ability to easily switch between sectors can be highly beneficial for Procurement.
Another interesting aspect for Procurement professionals is the way AI can negotiate. Indeed, some bots managed to learn the ability to lie. They feigned interest in objects they didn’t really want and then pretended to give them up during the bargaining process.
What’s next?
This experiment was just a good proof of concept that points the way to building more powerful bots in the future. However, researchers have highlighted two interesting elements. First, during a testing phase, the bots were often mistaken for humans; and second, that, without any human direction, the bots developed nuanced negotiation strategies. Basically, they are learning and improving!
However, it’s important not to get too excited by this research. The AI didn’t perform consistently better than humans. And the researchers were only able to train them in this one specific scenario. They don’t know whether they’ll be able to transfer these skills to other sorts of negotiation.
While AI is certainly more and more impacting the work of Procurement professionals and organizations, it appears there’s a long way to go before autonomous AI negotiators take over. We should first witness the development of AI assistant tools that will automate most repetitive tasks. Procurement professionals will have to work alongside AI bots and not against.
