After almost 18 months public, there’s no doubt ChatGPT has irreversibly transformed the world of work since becoming part of the equation. The AI software’s abrupt appearance sparked concerns about making some jobs obsolete— studies suggest that by 2025, AI could take over 52% of jobs globally.

Although this fear isn’t a complete overreaction, we now know that the key to making us better at our jobs (and our jobs better for us) is understanding exactly how to work with it.

The best human input = the best ChatGPT output

Instead of replacing humans, we’ve seen it make workers more productive and win them more time to spend on the stuff that matters by automating those pesky momentum-stunting tasks. A recent study carried out with Boston Consulting Group showed that those with GPT-4 did “more work, more quickly, and of much higher quality”.

From nailing prompts and providing relevant information, to knowing its strengths and weaknesses and how much of a ‘human touch’ to add: the winning combination of tactics needs training and refining in order to squeeze all the juice from our interactions with ChatGPT.

Nail impactful prompts, every time

On top of the obvious tip to “be clear and specific, avoid vague, ambiguous or overly complex language”, there are other necessary ingredients for a great ChatGPT answer: provide context or examples, keep questions open-ended (avoiding yes or no ones), give feedback and preferences along the way… Then adapt and experiment patiently to iterate, iterate, iterate.

Context is everything—but watch what you share

ChatGPT can prove limited in logic and reasoning. That’s why you should always hand over background context and reasoning for questions that require using more human-like logic. Remember: we want to give enough real-life context without sharing any confidential information. For example, presenting your real-life business situation as a hypothetical one.

Both sides should play to their strengths

Taking the model’s strengths and weaknesses into account is fundamental for tailoring our approach.  Like a mediocre job interview, we asked ChatGPT itself what its strengths and weaknesses are. Here’s what it had to say: “Strengths: Knowledge, adaptability, speed, accuracy, versatility. Weaknesses: Lack of emotional intelligence, bias and errors, context understanding, creativity limitations, limited real-world experience.”

Adding a human touch isn’t too much

Knowing the above very well can help us know exactly how and when to get involved. While much human intervention may not be necessary for things like data entry, admin, or menial tasks (except accuracy checking, checking for biases etc), for things like or content creation, writing, data analysis, a human touch is always necessary: AI should only help with the base or structure.

Of course, there’s a ton of human judgment that goes into all kinds of job roles, and upskilling that side is up to you.

ChatGPT’s countless—and counting—capabilities

AI tools will continue shaping our world with its exponential capabilities. So far, we count creative image generation, customer support automation, data-driven decision-making, collaborative writing assistance, and automated note-taking. Even more are developing as we speak.

Staying on top of and being able to truly embrace what they can do is what will mark our success. 5miles’ new ChatGPT learning track will help people with all backgrounds and in all job roles learn how to use ChatGPT like a pro. Keep an eye out for it this week.