Wondering just how good AI has become, I asked OpenAI’s ChatGPT to write a negotiating strategy for securing cost reductions from existing clients. What it provided was a little scary, because it was both very good in some areas and horribly wrong in others at the same time.
A friend had a similar experience when experimenting with AI to supplement his company’s technical helpdesk. The first question they asked ChatGPT was answered more correctly and concisely than his human call centre, however the second answer was a total wreck. Left unsupervised, AI was clearly going to cause just as many problems as it solved.
So, what does AI have to do with negotiations?
We already use complex analytical systems and decision support systems in the process of negotiations, so how is this different and does it matter? Depending on who your legal and contracts advisers are, you may already be using AI to assist with drafting for human review of the text.
Before you rush off and make AI a part of your negotiation preparation and conduct, there are a few issues that you should be aware of:
- As a natural language processing system, the underlying AI algorithm is trained by exposure to data on a massive scale. The data used is based on what is publicly available and does not in this case include sensitive commercial negotiation plans and outcomes. The training is often ‘unsupervised’ meaning basically that the AI learns patterns in the occurrence of words based on their frequency, not which examples are the best and most relevant examples of their type... So, questions about recent or sensitive issues just won’t generate answers.
- The system cannot provide any kind of compact traceability to how it assembled your answer and what specific data created that element of the model... So, it isn’t clear what is a ‘good’ or ‘trustworthy’ answer unless you are a discipline expert in the relevant area.
- Any data you provide becomes part of the AI data set and can be used to provide answers to other queries... So, think carefully about any sensitive data you include.
Notwithstanding these issues, there are some potential uses for even the current form of the system:
- Ask for a strategy dealing with the issue at hand (without providing any sensitive information), it may produce things you hadn’t thought of! For example, “how can we negotiate a position with less risk”
- When working in an unfamiliar industry or with a new customer, try asking it for a precis, and use that as a basis for further and more conventional research (i.e., google😊).
- Ask for a summary of the sorts of issue that might arise in a negotiation of the type you are contemplating and then use this as a conversation starter among your subject matter experts. You might just find some things that hadn’t been considered enough...
- Ask for a summary of information about your organisation and see what might be available to other parties.
Overall, the pace of development in this area is high, and we need to stay on top of developments to understand what we and our partners might be able to employ in negotiation support. For now, recognise we are in a transitional time for AI so move cautiously, stay involved and keep your human experts close!
Next time you need to conduct a complex negotiation, consider giving your Scotwork negotiation adviser a call for some professional help preparing for the negotiation! You will be surprised what an experienced human practitioner (with an objective understanding of your situation in all its commercially sensitive dimensions) can help you achieve.
Happy negotiating 😊