Posted by Apple Podcasts ¦ Amazon Music ¦ Spotify
Hi, and welcome to Try AI For Growth, a baby podcast out of Make Space for Growth. Here, I share short and maybe surprising stories of how I’ve used technology to tackle everyday challenges – at home, at work, in business. I’m your host, Sara Vicente Barreto, and today, I want to tell you about how AI helped me stop procrastinating on something that was important, but not quite urgent. One of those tasks that seems too big of a beast to start so you never do. Performance reviews with AI. Now, I know this is not the sexiest topic of all, but hear me out!
Background: The Problem to Tackle
Through the years, I have been on and off with the reviews at the charity. Whilst reviews may be less common in the charity sector or in SMEs, my experience in banking taught me that they were an important part of an organization evolving. It is my belief that if you grow and develop people, that will lead to better outcomes for everyone. And even though I am sold on giving feedback on the spot (perhaps a bit too much sometimes), I still find it important to make a stop in time and assess the year and the needs going forward.
So after the last 15 months were dedicated to restructuring and streamlining so many of our processes and reports at the charity, it was time to look at performance reviews. Truth be said, we had not done them since COVID and, with the Mozambique team, we had not done it often. Our project manager did have sit downs for reflection on the year with everyone, but I felt like we needed to align the newly assigned job descriptions with a performance review. Moreover, the team on the ground was asking me for help with it, so I did not want to disappoint (my most common fear)
Part of the transformation this year was to bring structure, clarity and direction closer to the ground, so the team could be empowered and make more decisions. And we took important steps in this direction:
- We made the job descriptions aligned and clear with activity expectations for everyone (I wish I had done that with ChatGPT but unfortunately I did not realize at the time this could be done);
- We hired a project manager on the ground that works with all field officers to help them work towards the expected activities and ensure we have the right reporting in place;
- We drafted a set of common goals across all projects with related activities and performance indicators in a way that we had not done before (and yes, after the first ones were done manually and then ChatGPT has been a good fellow in helping us with this as well);
- We shared the individual project reports with the detailed analytics we send to the donours to the project team so they could see how we assess the projects and how transparent we are about what goes well and what does not go well.
With this in hand, it was time to set the stage for the next phase. Performance reviews and goal settings time, so we can achieve our goals faster and the team can take the initiative to assess how they are doing and how they can change a project for the best. Ultimately, we don’t want them to fear the process, but rather to engage in a constructive way to improve our results. But how to start?
I was in constant fire fighting mode at the charity, covering for maternity leave and departures, hiring and training new people, while trying to fundraise and run the day to day all at once. Sounds familiar for an entrepreneur? How could I conceive even getting back with performance reviews? But I had 3 key assets. First, I had the history of what we did before. Second, I had a full-time volunteer who was a former head of HR in a corporate. Third, I had AI. It had to work.
Performance reviews with AI
So I did what I often do when I am stuck in a problem these days, and I need something that helps me start. I go to ChatGPT and try it out. Here is how I started
- I added 2 documents to the chat- First, an old performance review that we had created back in 2020 with the help of a corporate partner project at the time. It was quite detailed but a bit vague in terms of the concepts that the local team might relate to. I also added a job description I had done earlier in the year.
- I prompted as detailed as I could:
“I am preparing the performance review process for our Mozambique team. I am including our prior 2020 model which I would like to review in light of its complexity. I would also like to align the competencies assessment with the priorities reflected in the job description the team received earlier this year. In the voice of an HR manager, can you provide me with a set of revised questions in language that would be simple enough to implement with the operational team on the ground?”
I was mind-boggled by the first answer. It came split into sections – Planning & Organization, Project Execution, Monitoring and Relationship Building, Data collection, Behaviour, Personal Development. Not bad for a start. Rule #1 accomplished – I had provided good context, included relevant data and was specific in my ask.
Time to move to rule #2 – refine
I got excited with the answer and suddenly saw a small light at the end of the tunnel for this project. So I uploaded all of our job descriptions and asked for revised questions that would allow me to have a single consistent questionnaire across all the field officers and project manager. The answers were duly reviewed and it came with the extra recommendation of how each one could have a personalized section on the basis of the project that each person was covering.
But I was still not happy. A lot of the questions were open, and I wanted to be able to have a good amount of closed questions that would facilitate the process and also allow comparability over time. Moreover, I knew culturally that would make implementation more successful. So I went back at it and tried again. That is when the revised answer was much closer to the final product.
Was this it? No!
Rule #3, challenge the model
My next question was
“How does this questionnaire compare with the original model in terms of the areas included and what have you left out?”.
At this stage, the model struggled a little bit with analysing the original document I had provided in excel to give me a detailed enough answer. The way the columns were organized were not allowing it to read the data accurately. So I provided a PDF of the old performance reviews and got a great assessment of the areas the AI had maintained, refined or excluded in this new proposal. The model also justified why it had done this. I was impressed.
With this, I was able to adjust further, add back the areas that I felt were important, tweak the sections and number of questions and have an excellent draft to get the process going,
Time to get the humans
Don’t get me wrong, the process does not end here. The next step is pretty human. Our volunteer reviewed the document and provided a few thoughts (though not many, to her great surprise) and is now working on implementing it with the ground. That will be the hardest part and that is where humans will need to be involved all the way.
As a bonus thought, one of our elements of discussion as we finalized the process was how to match the self-review with the downward review from the local coordinators. I feel pretty strongly these need to be done independently, but my colleague felt like we needed a space to match the 2 and see where the differences are to be discussed. We agreed to do it in the performance review meeting with the project manager, who will be tasked with making that assessment.
Do you want to bet that I could put the 2 documents in ChatGPT and ask for a preliminary assessment of the differences between evaluator and evaluatee and get a great analysis?
Lessons
So what have I learned with this experience?
- The Power of a Clear Prompt
Providing detailed context in my initial request made all the difference. The more specific I was, the better the output. I felt like my first prompt was pretty strong and well thought of. - AI Saves Time, But You Still Need to Guide It
ChatGPT gave me a solid draft, but I had to refine it, review the changes, and make decisions about what worked best for my team. - AI Can’t Always Read the Data
When AI struggled to read the document, I was fearful this would not work. I had some issues before with excel files and was not keen to repeat the series. But I tried to find a solution that allowed AI to work with me. Pretty much as you have to do when a colleague is not able to work with a document right? - AI Can Do More Than Draft
AI did not just create the form—it helped me analyze and improve it by comparing versions and offering suggestions. I am finding it more and more it can help me analyse and think, not just create.
For me, performance reviews aren’t just a corporate checkbox—they’re an opportunity to align your team, give and receive valuable feedback, and support growth. But it is true they can be daunting for to start, especially for small teams or busy entrepreneurs. By doing performance reviews with AI, I turned a process I’d been putting off into something manageable and even insightful.
So, if you’ve been avoiding performance reviews or another company process because you feel like they’re too overwhelming, try this:
- Start with your current resources.
- Use AI to draft a structure.
- Review, refine, and adapt it to your needs.
That’s it for today’s experiment! I hope this story inspires you to tackle those tasks you’ve been procrastinating on—with a little help from AI. If you’ve used AI to simplify or improve your workflows, I’d love to hear about it. Send me your stories or questions, and let’s keep experimenting together.
Thanks for listening to Try AI for Growth. Don’t forget to subscribe, and if you enjoyed this episode, leave a review! Until next time, keep experimenting and keep having fun.
If you missed the prior episodes, go here!