I was asked to do some a little thinking and brainstorming recently. Rather than send some emails or get on a call, I decided to ask my local AI a few things. This talks about the responses and reactions from my local DeepSeek AI..
This is part of a series of experiments with AI systems.
Requests for Feedback
I’m a part a research group that is trying to learn how to work better with others inside or outside of an organization. As a part of this, I was asked for some questions we might use to survey people on how they work as a baseline.
I was a little stumped on where to start. I had some ideas, but everything sounds to vague or too specific and I felt as though I couldn’t find a good place to start. So, I decided to ask my AI.
I have a DeepSeek model running in a container, so I opened my browser and asked this question: What are some questions I can ask employees to understand if they collaborate well with others?
The response was interesting. DeepSeek tries to give you some insight into how it approaches the problem. The initial responses showed a little stream-of-consciousness that reminded me of someone thinking out loud. Here’s the first part:
Okay, so I need to figure out how to determine what questions to ask when employees work together. I’m not really sure where to start, but maybe I should break it down step by step.
First, I think about why we’re asking these questions. If it’s to understand collaboration, then we want to know if the team is working well together, communicating effectively, or if there are areas they’re struggling with. So, what kinds of questions would help us assess that?
I remember reading somewhere about gathering feedback from employees. Maybe those are good ones because they can show how each person feels about the group’s work. But how exactly do I phrase that? Like, “How have you felt about the team’s performance so far?” or something along those lines.
Then there’s communication style. If an employee doesn’t contribute much in the moment, maybe asking them to describe what went on before they spoke would help identify if their approach is effective or not. It sounds a bit direct, but it might reveal issues like being quiet when someone speaks up later.
Collaboration skills are another area. I think it’s important that employees know how to work with others, so asking specific questions about group processes and roles could be useful. Maybe something like, “What do we use for planning meetings?” or, “How often do you give feedback within a week?”
From there it had more explanation and then it said “consider the following structured approach:” and it gave me a set of bullets, which you can see some of in the image below.
This wasn’t quite what I was looking for, but it did get me thinking and helped me send a list of possible questions to the group to consider. This was as helpful as if I’d have called someone, but way quicker and I didn’t interrupt anyone else.
I was genuinely pleased by the way this went.
A Recommendation Letter
I had an author ask me for a recommendation letter. This has happened in the past and I’ve written a few of them, but I always struggle to get one written that sounds professional and not like I’m a clown of a DBA/Editor who wears Hawaiian shirts to professional events.
Again, I decided to ask DeepSeek to outline one, giving it the title of an article. It had a few suggestions and then gave me a draft. My prompt was: draft a reference letter for XXX YYY for the article: ipsem lorum ipse lore.
First I got the thinking, making sure I was considering the aspects of the article, should I point out some technical things in my recommendation. It also thinks about including the impact of the article as well. Next, I got a draft with some “insert name” and “insert date” places.
I then added another prompt noting this was for SQL Server Central, which is a respected and long-lived (24 years now) publication well read in the industry. DeepSeek then adjusted its thinking, noting that we want to include info about SSC, how we review articles, etc. This was the end of the thinking, which I really liked. It’s the type of thing a human might say to let me know what they’re going to do:
The draft letter is still a little off, with too much praise for me. This is a good article, but it’s not amazing. In any case, this gave me a decent first draft I could edit and then send off to the author.
It’s not that I couldn’t have done this, but this draft saved me some time and allowed me to clarify my own thinking about what I actually wanted to write.
I think this is slightly better or worse than if I’d asked a human to help give me a draft. Some people might have given me a slightly better draft, some worse, but I’d still be adjusting the tone and editing the content. I know because I’ve used humans for this exact task before.
This was interesting to me to play with an LLM and look for help. On balance, this didn’t take long and I think I did get some help from an AI. It didn’t do my work, but it helped.
