I Let AI Manage My Projects for a Week - Here's Why It Didn’t Work

I decided that I wanted to try an experiment that I doubt many have tried before. Replace my job with AI.

If the revolution was coming and artificial intelligence was so powerful, surely it could replace me pretty easily. Save me from doing a bunch of work.

And do it better than me right?

This would be easy. I’d spend my week relaxing and giving instructions to my new robotic employee. But it didn’t end up that way. I’m not sure if that’s a good thing or not but here is what happened.

Lessons learnt (and some tips if you want to try it yourself) at the end.

The Week

This was a fairly typical week for me. Lot’s of meetings, often with very limited direct participation from me. I find that for most of my meetings I am listening for 90% and only giving input occasionally.

Apart from meetings, I process and write a large amount of emails daily. Perhaps in excess of fifty. Some of the emails I write are single sentences, but often they are a few paragraphs.

Report writing is not a daily activity and I spend at most a few hours per week writing reports. These are mostly updates and not full reports. Those will happen once or twice per month. The updates I write are nonetheless important and are used to keep stakeholders informed on project progress.

How I Used Artificial Intelligence

I decided to use the most freely available (and free to use) tools I could find. Most PM’s will not be paying for niche, expensive AI tools. They will be using online versions of ChatGPT or Gemini along with AI features embedded in existing software such as Google Docs or Notion.

Some important lessons-learned:

• I am a bit nervous about privacy. Not being an expert on large language models, I’m not sure where my info is going. Where is all my uploaded project information sitting now?

• Writing prompts takes time. A database of previously used prompts is extremely valuable and saves time. We created a great database of ChatGPT prompts for project managers for you. BONUS - Since you made it this far in the article, use code CHAT20 for 20% off.

• I didn’t fully trust the outputs I was getting. And this slowed me down. For example, I asked ChatGPT to summarise meeting minutes into project status updates. It did so successfully but I still went through the minutes manually and checked them against the summaries. I know this is the responsible thing to do, but it did take time. Using AI saved me time but not a huge amount. Without AI, the summary process would perhaps have taken me 60 minutes. Using AI it took me 40 minutes. A saving, but not massive.

• It would have been very helpful to have a way to integrate notes and to-do’s from various sources into one coherent place. To group these by project would have been helpful. Integrating notes from emails, documents and meeting minutes in one place is a valuable way to keep track of your tasks. However the process of achieving this is extremely time consuming.

• There are a few large time-consuming tasks in everyday project management. Meetings and reading emails are chief amongst them. Most of what happens in these sessions can be categorised as one thing: assimilation. You are reading, listening and speaking to assimilate information. Once you have understood the information, you then spend a small amount of time giving a response, making decisions and seeking feedback (more information to help future decisions). I haven’t yet found a manner in which to make the assimilation of information speedier through the use of artificial intelligence.

• The AI was really great at giving me a wide range of things to think about. It’s thinking was wide, but shallow. I might have come up with these suggestions myself, but it certainly did them faster. I was then able to do deep thinking from its wide outputs. This was of great help to me.

• The majority of my inputs (and all of my outputs) were text based. I used voice typing and AI to clean up the output, but other than transcription, everything required the rather slow use of a keyboard. New AI tools are being released which will make voice-based inputs more accurate and save further time.

  

  Conclusion

  I might be tempted to think of my weeklong experiment as a failure. I could not easily replace myself with a robotic clone. In many ways however, it is heartening. The end of the human project manager is not quite imminent. The work required to input information into my AI tools was extensive. It mostly took more effort than the time it saved.