The previous four parts were diagnostic. This one is prescriptive. Here is how to get real value out of AI tools without getting burned by them.
Use it for first drafts, not final answers
AI is fantastic at producing a shape. A shape of an email. A shape of a function. A shape of a meeting summary. Take the shape, then bring it the rest of the way yourself. The first 60% of any creative task gets easier. The last 40% is still on you, and that's where the value is.
Verify anything that matters
If a wrong answer would cost you money, time, or trust, verify it. Names, dates, numbers, quotes, citations, code that touches production — all of these need a second pair of eyes. Yours or someone else's.
A useful habit: ask the model where it got the information. If it cannot tell you, treat the answer as a guess. If it can, check the source actually says what the model claims.
Give it context, not just instructions
The single biggest improvement most people can make is giving the model more of the relevant material up front. Not "write a follow-up email" — instead, "here is the original thread, here is what we agreed to, here is the deadline, write a follow-up email." The more context, the less the model has to invent.
Notice when it's pretending
Watch for the small tells:
- Suspicious specificity ("the 2017 study by Lee et al.") that you can't immediately verify
- Wishy-washy hedges that sound thoughtful but say nothing
- Lists of seven things when there are clearly only three
These are the signs the model has run out of real information and is filling in.
Don't outsource your judgment
The most common failure mode is not the model being wrong. It is you trusting the model when you would not have trusted a human stranger giving you the same answer. Your instincts about what to trust are still good. Use them.
The bottom line
AI is a tool. A genuinely useful one, in the right hands, for the right tasks. Use it for the unglamorous middle 60% of your work — the boilerplate, the first drafts, the boring research. Keep your judgment for the parts that matter.
That's the series. If you read all five, you now know more about how this stuff actually works than 95% of the people writing breathless articles about it. Go forth and be skeptical.