CoPilot to your Captain
Whilst acknowledging questions and concerns around the use of AI, it must also be acknowledged that AI can introduce greater efficiency simply by assisting in small everyday tasks.
An example task
Researching flight options can be a time-consuming task for assistants. Firstly, we conduct an internet search using sites such as Skyscanner, which then direct us to the individual airline’s websites to verify. Secondly, we present our research in an easy-to-read format so that our client can skim through and choose their preferred option. Thirdly, once the client has confirmed their preferences we can go ahead and make their bookings.
Depending on the journey – length, number of connections – this task can take up to a few hours to complete and diarise all tickets.
The arrival of the internet saved assistants some time telephoning around for travel options. Wouldn’t it be great if we could not only find options, but have them returned in a format that is ready for review by our client?
I’m hearing that tools such as ChatGPT and Claude can do this exact thing, returning terrific results. However, following media reports this summer, I am just a bit wary about putting them onto my devices. I decided to test CoPilot’s capabilities as this is an included, but not really used, application within my Microsoft 365 subscription.
CoPilot
CoPilot is essentially constrained to the Microsoft tenancy in which it exists. I asked CoPilot to look up flights from location X to location Y, outbound date A returning date B. Present airline, departure time, arrival time, duration, cost in a spreadsheet, ordered by departure time earliest to latest. CoPilot (very politely) told me it can’t access all that information. Interestingly, it did tell me how it could help.
Because CoPilot cannot access the scale of information that ChatGPT and Claude can, you must feed it information. To be honest, I quite like this method of working. I felt more in control knowing what information CoPilot had access to. As a first step, from a screenshot of an airline’s website, CoPilot produced a spreadsheet with two worksheets. One worksheet showed outbound flights and the second showed return flights.
Step by step
Having successfully completed this first stage, I then asked CoPilot to update the spreadsheet from screenshots from Skyscanner. CoPilot told me it needed me to upload the original spreadsheet again, which I did. I also asked CoPilot to ignore duplicate flights so that none were entered onto the spreadsheet twice, which it also did.
“Drip” feeding CoPilot produced a spreadsheet that I was able to send to my client to easily choose their preferred flight.
Time savings
This task took half-an-hour to complete with the help of CoPilot. Ordinarily, it might have taken nearer to an hour. On another occasion, the task would likely take less time as I have learnt how to prompt CoPilot to help with it. AI works from prompts, so the better you can prompt it, the better it will work for you. My client was busy today, so greater efficiency with this task has allowed us to focus more on others.
Why this is good use of AI - the nature of the task
This was very much a simple extension of a common research task. The information presented could very quickly and easily be verified by a human. There was no analysis of the information involved – CoPilot did not give an opinion on which option was “best.” That decision still rests with the client. Tickets will not be booked until the client has approved and AI will not make the booking.
To summarise, CoPilot did not do any decision-making or judging. It did not work with or have access to any personal or sensitive information, only that which is publicly available. The task was very much completed with human oversight. However, the benefit of using CoPilot was a definite time saving and identification of further time saving.
Integrate not replace
Using AI does not have to mean the drastic outsourcing of big tasks to a robot! If AI can support with these kinds of tasks, surely this is a beneficial use? What time-consuming, common or day-to-day tasks can you think of where integrating some AI deployment could lead to greater efficiency in your business?
