AI Use Cases

With English not being my first language, I find it really helpful to roughly draft emails and then use AI to improve them. 

This has saved time and increased effective emails, for example, sometimes AI suggests things I hadn't considered before and it's helpful to include from the start. 

Prompt: You are a [job role]. Write a short, professional and friendly email to [recipient] to [aim of the email]. Include the following: [key points to cover]. 

Or I ask AI to: Please respond to [name] in the email thread below. Write a short, professional and friendly reply, including the following: [key points to cover]. 

Sometimes I add other specific requirements, like making sure it's easy to understand, the ask is clear, the email is engaging etc.

I use MS Teams record/transcribe function to document meetings, ensuring all meeting participants are aware. I provided Copilot with existing examples (x 3) and new meeting transcripts with the prompt outlined below to generate meeting summaries that match an established audience, tone and style. 

This did require me to check for accuracy, particularly for names and acronyms, but was a very time efficient way to complete basic admin tasks and frees up my time to add value as an expert (and means I contribute fully during meetings, rather than focusing on note taking). 

Prompt: "Hi Copilot, please can you draft a meeting summary for [audience] based on the transcript of our meeting (attached); I have also included previous summaries, which I would like you to emulate in terms of format, style and level of detail (email samples below)." 

I use Copilot to condense key messaging into helpful headlines; this can be handy where I have existing information or prose that is longer form and needs repurposing for a different audience or format.  

I provided Copilot with three consecutive annual reports and with the following prompt to condense key messages for questions/themes I had already identified: 

"Hi Copilot, digest the following reports and synthesise the biggest (positive) impacts. Can you tell me:1) Where there's evidence [of desired impact]? 2) What's improved the most over the period? 3) What [team] does comparatively well?"  

I find that I get better results if I am explicit and clear about the intended audience, their roles and potential interests and perspectives. 

The key benefits are time saving and ideas generation for tailored messaging. 

I sometime use Copilot as a thinking partner when I'm working through a new idea. Rather than asking it for answers, I use it to help me explore different perspectives, test assumptions and think through options before making decisions. 

For example, when considering how to redesign a development programme, I asked Copilot to take a coaching approach. It helped me clarify what problem I was trying to solve, identify factors I might be overlooking and consider alternative approaches. 

This helped me structure my thinking more quickly and arrive at a clearer way forward Copilot challenging my assumptions and helping me reflect. I have found that often I need to leave any conclusion for a day so I can look at it afresh and make sure I am still using my professional judgement and make the final decisions. 

Prompt: "Act as a coach. Don't give me solutions. Ask me questions that will help me think through this challenge, test my assumptions, explore alternatives and decide on my next steps: [describe challenge]." 

I produced an overview of which topics relating to innovation in the industrial strategy get the most frequent media coverage, what types of stories drive that coverage, which publications the stories appear in, and who the key journalists are. In comms we call this a media landscape review. It's a useful first step when building profile in a given area. It helps you uncover opportunities to get coverage and decide what kind of stories to prioritise and what angle to take to prevent wasted effort. 

This can be done without AI, but it takes a long time. A PR agency might charge 10k for it.  

I used Microsoft Copilot’s “Think deeply” (this gives more thoughtful analysis and complex reasoning) mode using these core prompts: 

Which topics relating to innovation in *Sector* get the most frequent coverage in British news media? Include accurate examples of articles and publications. 

What are the most influential media publications relating to *Sector*? 

I used follow-up prompts to obtain missing information or to expand upon key points. I edited to remove unnecessary information, to add from my own knowledge and experience, verify the sources (as AI can hallucinate), and to give it a general sense check. 

As with any consultant, you will get more out of it if you actively interrogate rather than blindly accepting what they give you. 

Using Copilot to write a complex formula in Excel to pull data from a vast number of fields, from a report which has the same fields but in a different order. Previously, I was able to do this by modifying a vlookup/xlookup, one field at a time to instruct it to pull in a field that I had selected. Copilot was able to provide me the formula to instruct Excel to lookup the field heading from the current reference cell, find that heading in the other report, then return the corresponding field value I required.  

This has saved me considerable time, turning a long-winded, semi-manual process to a very quick copy and paste job. While the formula is not so complex that it could not be learned, but it was difficult to build from scratch without support. Nonetheless, AI tools can handle this request and supply a working formula in seconds. 

This is the specific prompt used for my spreadsheet, but the principles relate to asking Copilot to create any formula for your needs. 

'Write an Excel formula to populate columns D and onwards in my destination table ('Data 19-05-26'), pulling data from a lookup table ('Data 19-06-26'). The key begins cell B2 and corresponding lookup key found in column A. Column positions differ between the two sheets, so the formula must match column headers dynamically.’ 

In excel, I use Copilot to check and cleanse data, in pivot tables and with complex formulas.  

For large sets of data that need to be cleaned, instead of using filters and pivot tables, I may prompt something like ‘I want to retain only staff that have been working for over 5 years, have undertaken management training in the past 2 years, but are not yet managers. I am interested in [add criteria or area of interest] and delete all irrelevant columns. 

In addition, I ask Copilot to repeat the check and compare the results to identify any inconsistencies. I then verify the criteria, formulas and test a sample manually, to check for accuracy.

This saves me a lot of time in my monthly reporting cycle. 

I have been using AI to check for potential missing open and close tags in XML for SQL scripts I have been writing. 

Example Script: 

'<Row><Cell><Data ss:Type="Number">'||TO_CHAR(RANK () OVER (order by substr(spi.spi_sprc,1,9)))||'</Data></Cell><Cell><Data ss:Type="String">' as STRING1, 
spi.spi_sprc as STRING2, 
'</Data> 
</Cell><Cell><Data ss:Type="String">'||STU.STU_SURN||'</Data></Cell> 
<Cell><Data ss:Type="String">'||STU.STU_FNM1||'</Data></Cell> 
<Cell><Data ss:Type="String">'||SPR.SPR_UDF3||'</Data></Cell> 
<Cell><Data ss:Type="String"></Data></Cell> 
<Cell><Data ss:Type="String">'||SPI.SPI_ROUC||'</Data></Cell> 
<Cell><Data ss:Type="String">'||spr.awd_code||' '||awd.awd_name||'</Data></Cell> 
<Cell><Data ss:Type="String">'||SPR.STS_CODE||'</Data></Cell> 
<Cell><Data ss:Type="String">'||PRS1.PRS_NAME||'</Data></Cell> 
<Cell><Data ss:Type="String"></Data></Cell><Cell><Data ss:Type="Number">0</Data></Cell> 
<Cell><Data ss:Type="Number"></Data></Cell><Cell><Data ss:Type="Number">0</Data></Cell> 
<Cell><Data ss:Type="Number"></Data></Cell>                                                                                
<Cell><Data ss:Type="Number">'||taken4.total||'</Data></Cell> 
<Cell><Data ss:Type="Number">'||passes4.total||'</Data></Cell> 
<Cell><Data ss:Type="Number">'||round((y4.y4/taken4.total)/10,2)||'</Data></Cell> 
<Cell><Data ss:Type="Number">'||taken.total||'</Data></Cell> 
<Cell><Data ss:Type="Number">'||passes.total||'</Data></Cell> 
<Cell><Data ss:Type="Number">'||round((all1.all1/taken.total)/10,2)||'</Data></Cell> 
<Cell><Data ss:Type="String">N</Data></Cell> 
<Cell><Data ss:Type="String">N</Data></Cell> 
<Cell><Data ss:Type="String">'||SPI.SPI_PITC||' '||Pit1.pit_name||'</Data></Cell> 
<Cell><Data ss:Type="String"></Data></Cell> 
<Cell><Data ss:Type="String"></Data></Cell> 
</Row>' 

Prompt: I use the prompt “are all open and closed tags accurate” 

An example outcome is outlined below. Please note this is an example, so it has some inaccuracies. 

 

• image tbc

This is very much in the experimentation stage as most of my work is done in SITS and as yes (although slowly changing) the syntax of that particular system isn’t something that AI is particularly aware of, so will give incorrect answers.  

However, it is already offering some time saving benefits in checking errors.  

 

As part of preparing for my next Advance HE Fellowship level via the QMA Teaching Recognition Programme, I created a Copilot Agent to support the collection and organisation of evidence. 

Using a simple prompt in Copilot Chat, I generated a comprehensive set of instructions for the bespoke agent in Copilot Agent Builder: 

"I want to build an agent in Copilot to support my preparation for [relevant level]. Please develop a set of instructions for this that I can use to create the agent." 

The initial instructions required minor amendments to correct some Professional Standards Framework (PSF) inaccuracies, but they provided a strong foundation that was very easily refined. 

Rather than acting as a writing tool, the agent will support long-term evidence-gathering, reflection and critical evaluation of impact. By providing details of projects, initiatives and achievements over time, I can map activities against relevant PSF dimensions, identify the strongest evidence, and reflect on impact. The agent helps connect individual activities to institutional priorities, creating a structured evidence base that can later support development of an initial draft.