Ask specific questions - try to be as specific as possible. Use evidence to back up claims. Don't add an emotional element - the more data-driven it can be the better. For example "Joe was too slow getting into position" could be turned into "Fire Team Bravo took 10 minutes to move into position." This removes the emotional layer that Joe (and his team) would react to, while still acknowledging the deficit. Instead of asking why it took so long, ask specifics to build on that why - Did they take too long because of deficit physical fitness? Poor map reading? Unexpected obstacle (wait... that 10' ditch isn't on the map!)? Lack of understanding the mission? Lack of skill in deploying XYZ? The trucks dropped them off 1km out instead of 400m out? This allows for a complete picture to be developed and not just a sounds-good-assumption - too slow because Joe is a fat butt turns into a non-hostile evidence-based problem (which may or may not in fact support the initial claim that Joe needs to lose 20lbs and train his running game).
A strategy for this is asking the 5 Whys.
What happened: Fire Team Bravo took 10 minutes to get into position.
Why? They moved too slow.
Why? They're out of shape.
Why? They haven't been doing PT.
Why? Company PT has been suspended due to COVID.
Why? Potentially unsafe to exercise as a group.
So now we have the problem and, through a course of questions, we have identified the root of the problem. From here we can come up with a real solution that addresses the actual issue. Of course, we could reframe this line of questioning if they hadn't been slow due to fitness, but the example works. Some say this is unnecessary and annoying, and in the given example it might seem like it is overworking the problem, but maintaining this structure can help and make sure we don't miss something. It also leads to solutions at multiple levels - perhaps Joe and his team need to support from their platoon sgt on developing fitness among the restrictions, but also perhaps the company needs to game plan strategies to prevent this from occurring with the less individually motivated members. (And although this tactic is called the 5 why's, you don't always have to ask why 5 times - sometimes you get your root sooner.)
Basically - what was supposed to happen, what did happen, and why was there a difference. That builds the theme. From there, you can look at what worked and what didn't, attempt to identify a cause for those, and then replan how you would do it next time, preserving what went right and improving where possible.
Also, AARs are not just a military thing. They are frequently used in the business setting. I participated with them in the military, but I've conducted them in manufacturing. They're slightly different, but the gist is the same.