One thing about Copilot: It does not lack confidence. We should all take a page from its programming and aim for that same level of confidence.
Just like with humans, that level of confidence can be awesome. Confidence is key to making things happen, being heard, taking the next step forward. A confident human can get a lot done.
And on the flip side....
Confidence without substance is bluster. It's a false narrative or overshooting your skill set. It's noisy and, well, it can be a little obnoxious, you know?
Last week's blog was all about using our Microsoft tools and Copilot to improve research.
A very innovative way to use AI. And, it is partly what it was designed for. We have to be thoughtful though. Think through the output and ask yourself if the confidence you see in the reply is true.
Ask yourself: How do I know what to trust?
You'll get fast, confident output with polished precision. AI is great at this. GREAT.
Your job? Make sure you don't confuse polished with proven.
Make verification part of your workflow now. Start here and create good, solid habits that won't leave you looking at a confident, blustery, incorrect mess of output.
Where exactly did this come from? And when?
You're looking at the output already (I told you it was fast). Now your critical thinking needs to kick in. This can be in the form of prompts, of course, to dig a little deeper than question, answer, done.
Ask Copilot, where did you get this information? I need you to share the source. Then YOU have to decide if the source is credible and if it actually does back up the output.
Copilot will send you the source, but your work isn't done there. You have to LOOK at the source. Open it and start your human due diligence.
Also, check the date on this source. AI can surface information that was accurate at one point but is no longer current.
That's important for all kinds of topics. Things like Microsoft features, pricing, regulations, security guidance and more.
If the information you are looking for is fluid and has the potential to change, you need to verify that you have the most up-to-date output from your AI friend. That's your job.
Who does this impact and why should I care?
When you get that output, think about the recipient.
What does this claim influence? Who does it influence?
Let this be another guide in your verification process.
If it's for a client recommendation, a financial decision, a presentation (your credibility is on the line here), a policy or an executive decision, it is worth your time and energy to vet the output.
Period.
These are not the only scenarios, just some of them. The point is, consider where you're sharing this and go back to the primary source when possible.
Because really, it's your name going on it, right?
Yes. It's your name going on it.
Incorrect, outdated and just plain false material cannot be 'blamed' on Copilot. The output is only as good as the input, and the verification process belongs to the humans in the room.
Which brings us back to confidence.
Ask yourself if you are really confident in putting your name on this. If your boss, client, colleague or even friend wanted to know where the information came from, could YOU stamp your name on it?
If the answer is no, you're not finished yet.
We shouldn't be afraid to use AI because it can be wrong. Humans get things wrong too.
(Ahem. If you say you haven't gotten anything wrong, I'll be challenging that output friend)
We just need to make sure that we are using AI responsibly, verifying and vetting before we turn this out into the world.
There's plenty of unclear and false information out there, let's not contribute to it.
It's our job to help AI get it right. Confidently.