Using AI as a reasoning and communication support tool in Career Services and programme planning

My role in CHED’s Careers Service involves developing reports, presentations, concept notes, motivations, and strategic planning material related to student careers and entrepreneurship initiatives.

A large part of my role requires synthesising information clearly and communicating ideas in a structured and accessible way. This includes comparing reports, identifying gaps and overlaps in job structures, preparing presentations, and drafting motivations aligned to institutional processes and policies.

The challenge is often not generating ideas but rather organising and refining them efficiently while still maintaining clarity, alignment, and strategic reasoning.

I also need a way to test and strengthen my thinking independently before engaging others. In many cases, the work required balancing strategic intent, policy alignment, and operational detail within tight timelines.

I mostly use Microsoft Copilot because I find it safer to use, particularly for internal UCT documentation. I also find it useful for locating UCT processes and policies and simplifying my understanding of them.

I use AI mainly to assist with:

• improving the clarity of documents and emails
• summarising reports and highlighting key issues
• comparing reports and generating structured comparison tables
• developing presentation outlines
• drafting motivations aligned to policy and process
• exploring organisational gaps and overlaps in job structures

I used ChatGPT previously, although I use it less now that Copilot is available through UCT.

One example was using prompts to support organisational analysis and strategic thinking. It involved comparing three job descriptions across PC11, PC10, and PC09 levels to better understand how the job family structure worked, where overlaps existed, and whether the strategic role expectations aligned appropriately at Head level. Expanded example:

• “Using the 3 job descriptions create a table format and indicate how the job family works. The positions would be PC11 to PC10 to PC09. I want to understand if the relations work and also identify gaps and overlaps. The PC11 is the Head and should have a strategic role.”

My prompting approach varied depending on the task, but generally combined contextual prompting, instructional prompting, comparative prompting, and reasoning prompts.

For summarisation tasks, I use prompts such as:

• “Using the attached report, summarise it highlighting the key issues.”

For comparative analysis tasks, I use prompts such as:

• “Compare this report to the following report and create a table with the differentiation.”

For presentation preparation, I use prompts such as:

• “Using the attached document produce a presentation outline with the key issues.”

For more detailed planning work, I use contextual prompts to help me conceptualise and explain the purpose and strategic direction of the 2026 Student Entrepreneurship Week. For example, I prompted the tool to use a previous Student Entrepreneurship Week report to help produce a concept note focused on creating faculty-based engagement while still integrating the broader UCT Entrepreneurship Ecosystem:

• “Using the past SEW report, I would like to produce a concept note on the 2026 Student Entrepreneurship Week that will focus the week by Faculty but still manage to integrate the UCT Entrepreneurship Ecosystem.”

I also use prompts requesting aligned institutional reasoning. AI drafted a motivation for a deviation process linked to expenditure for the Global Careers Summit, ensuring that the wording aligned appropriately with UCT process and policy requirements:

• “Write a motivation for a deviation of process for the Global Careers Summit for the Expenditure. The event has taken place already and I need to clear all expenses. Explain that the monies used were not UCT funds but GCS funds which was transferred into the UCT fund for expenditure. A detailed financial and narrated report will be sent to GCS to indicate expenditure. Align this to UCT process and policy.”

I reviewed outputs carefully to ensure the responses aligned with the actual context, intent, and institutional requirements.

I noticed that responses could sometimes be inaccurate or misaligned initially, but that re-prompting generally corrected the issue. Because of this, I found it important to understand the subject matter well enough to evaluate whether the outputs made sense.

I do not treat AI as a replacement for knowledge or expertise.

I prefer using Copilot because I consider it safer for institutional work than some other platforms.

My work generally did not require de-identification of data, but I still remained aware that institutional policies should always be taken into account when using AI tools.

AI worked particularly well as a reasoning and thought-partner tool. For me, it became a way of testing ideas, refining my thinking, and strengthening the logic behind documents before sharing them with others.

It also helped improve efficiency by making it easier to summarise reports, structure presentations, compare documents, and organise complex information more clearly.

One of the most valuable aspects was being able to “bounce off” ideas in a way that supported deeper thinking and preparation.

Human judgment remained essential throughout the process.

I still needed to evaluate whether outputs were accurate, aligned to the intended context, and strategically appropriate. Re-prompting was often necessary when responses were misaligned.

I also found that AI only works effectively when I, as the user, already understand the subject matter well enough to guide and evaluate the outputs critically.

AI became a support tool that strengthened my reasoning process and improved the speed at which I could structure, refine, and communicate ideas.

At the same time, I have not fully integrated AI into my workflow because I still feel the need to remain in control of the process. Instead, I use AI selectively through prompting where it adds value.

I would avoid relying on AI where I could not properly evaluate the accuracy or alignment of the outputs myself.

I also would not treat AI as a replacement for professional knowledge, institutional understanding, or critical reasoning.

Using AI as a tool to get work done quicker, with more thought, is like having a friend you can bounce ideas off.

At the same time, you still need to know your subject matter. AI does not replace your knowledge.

I have also found that using AI consistently over time improves the interaction because the tool begins to better understand your style and approach. For me, effective AI use has been less about automation and more about strengthening reasoning and communication.