Pi: The most underutilised AI tool on the market

6th March 2025

ChatGPT is now a part of our lexicon and for some people it’s a big part of their day-to-day work life. Anthropic’s Claude and Google’s Gemini are also prominent, but one tool you may not have heard of is Pi.

Each AI has its own personality, and Pi’s is more human than any other chatbot. It sounds human and it listens to you in a way that other voice-led chatbots don’t. It feels like you are speaking to a real person and one that understands you.

In March 2024, Microsoft paid $650m to Inflection AI (the company that made Pi) in order to lure key players Mustafa Suleyman and Karén Simonyan to the Microsoft team. This means that the unique properties of Pi are now most likely going to be seen in Microsoft’s future tools.

A term you will be hearing a lot more in the AI recruitment space is ‘Prompt Engineering.’ This is another way of saying knowing how to write queries to AIs like ChatGPT in a way that will be more likely to get effective results. For example, if you ask ChatGPT “Is Clive Owen the world’s strongest man?” it is more likely to hallucinate and make something up, whereas if you ask it “Who is Clive Owen?” or “Who is the world’s strongest man?” you are more likely to get an accurate result. Reducing hallucinations remains a key problem for integrating AI into business processes and effective prompt engineering will be part of the solution.

There have been a lot of studies about the methodology of prompt engineering, including one study that showed if a user says to ChatGPT “If you do this task well, I will give you a $1 tip.” Despite not actually giving ChatGPT the $1, it performed better when offered this fictional financial incentive. Another idiosyncratic study showed that if a user says to ChatGPT “Pretend you are Spock from Star Trek,” it will be more likely to respond in an accurate way. 

With Pi, one of its biggest strengths is explaining complex issues in a comprehensible manner. One of the most powerful use cases of AI right now is having it act as each individual’s personal assistant. There is a story about how early on in his career as a CEO, Richard Branson did not know the difference between gross profit and net profit and one of his staff noticed this in a board meeting and then took him aside and politely explained it with a simple metaphor about fishing. With Pi, we all have that helpful assistant to help us understand these concepts.

AI is facing many bottlenecks right now, especially in the availability of data to train it, but one of the biggest bottlenecks is in our understanding of how to utilise it effectively. Many of us are quick to dismiss an AI tool because it can’t do what we want it to do straight away, but learning and challenging ourselves by using tools in new and different ways is something that will separate the people who reap the benefits of AI and those that are stuck with the luddites. We need to meet AI halfway, and just like the people who first used the internet, we need to accept that it isn’t perfect, but if we challenge ourselves, we will be part of the first wave of people who gain the vast advantages that this powerful technology has to offer.

Paul Willis

Head of Product, Bespoke

More Articles

More