Thought Leadership

Is ChatGPT ready to lead your team?

Written by Will McNelis | Mar 4, 2024 4:56:34 AM

Although it seems like longer ago, ChatGPT was released on 30 November 2022 and very quickly launched itself into just about every conversation and aspect of business life. There was plenty of fear mongering at first about "AI coming for our jobs", and to that end I wrote a post about it taking over Agile Coaching. I thought it was time for a yearly update.

In the year since, AI is no longer synonymous with just ChatGPT, and we have a plethora of tools at our disposal. Most people have incorporated some (or a lot) of these into daily workflows, and it's generally understood that AI is not going to replace us, but rather enhance and amplify our daily work. To that end, I thought I'd write an update, but with a focus on leadership.

So how can AI, large language models (LLM's) specifically, augment leadership capability in your organisation?

LLM Options

It seems that the different options in the AI landscape are doubling every day, but for the purpose of this post, I want to focus on LLM's, and specifically the ones that I use frequently.  The first thing to understand about an LLM is that they are chatbots that can 'understand' and generate language text. They do this by learning statistical relationships from text documents during a training process. The text documents they learn from is called a corpus, and their knowledge recency is limited by the corpus. 

ChatGPT

Whilst not really the first (that honour goes to Eliza in the 1960's), ChatGPT has quickly become the 'Kleenex' of LLM's. The release of GPT-4 on 14 March 2023 updated the corpus to that date (it was limited to 2020 with GPT-3). I subscribe to the Pro version to make use of GPT-4, image creation, and custom GPT creation.

Google Gemini

Google's entry into the market was originally called Bard and recently rebranded to Gemini. Gemini is unique in that it is not trained on a corpus alone; rather it can process multiple types of data simultaneously, such as text, images, audio and video. Gemini also has a leg up in terms of access to Google itself.

Microsoft Copilot

Copilot is the Microsoft entry to the field and replaces Cortana. For corporate users, this may be the LLM that you have access to within your company's toolset. Copilot is useful in that its incorporated into the Microsoft suite of tools and can be used to generate Powerpoint docs, Outlook emails, etc.

So how can a leader use LLM's?

Data Driven Decision Making

One of the strengths of AI is that it can analyse mountains of data, identify trends, and predict outcomes (with startling accuracy). This can help to provide insights that may not be readily apparent and are an incredible input into decision making, especially in a complex environment.

Team Member Development

LLM's are great at suggesting learning resources such as books, videos (in the case of Gemini), and training that might be useful based on a team members specific context (see prompts below for how to incorporate context).

Enhancing Collaboration and Communication

Using AI to assist in communication is probably the use case we are most familiar with at this point. Be careful, though, as people have become incredibly adept at spotting content regurgitated directly from ChatGPT. It is an incredibly useful tool for prompting and discussion points though.

AI is also useful in suggesting collaboration opportunities and ways for team members to work better together, especially in remote situations.

Improve Customer Insights

Again, analysing data and providing trends and insights is a strength of our current crop of AI tools. AI can help leaders and teams to tailor products and services to better meet customer needs. 

Crisis Prediction

By inputting the right data, AI can also help to spot patterns and predict risks that can affect organisations (financial, operational, or reputational). They don't predict the future, but they can be great thought starters for areas potentially requiring risk mitigation.

Innovation

ChatGPT, Gemini, and Copilot are awesome brainstorming inputs for ideation. They won't come up with the 'big idea' for you, but they can suggest opportunities and approaches that the team may not have come up with on their own.

What are the drawbacks?

AI is certainly more than a parlour trick, but not quite 'real'. Some key limitations include:

  • They have zero emotional intelligence. This was a key problem with them taking over a coaching role, and the same is true for leadership.
  • They do not understand ethics or have the ability to make nuanced decisions as humans can.
  • There is inherent bias in the training process, and that bias can come through in their responses. Sometimes dramatically so.
  • They are dependent on the data they have been fed.
  • Regulatory, compliance, privacy and security concerns abound, and will be different  between organisations and sectors. I recommend researching your context with this one, but all organisations I have come across so far have an approved tool or guidelines.
  • Whilst they are brilliant at identifying patterns, they are not able to think creatively, especially in novel-practice type situations.

What's in a prompt?

One of the key skills required for using LLM's is creating the right prompt to generate the results you are looking for. The good news is that prompts are all natural language; in other words, you don't need to learn to code. There is a certain format that yields better results than just asking a question. Key aspects of a good prompt:

  1. Be specific. A generic request will get a generic response.
  2. Use "Act as if..." in your prompt to tell it who you would like it to model its response from.
  3. Tell it what output you want. A bullet list? A table? An image?
  4. Tell it what to do and what not to do in its response. If there's something specifically you do or do not want, tell it. 
    • Subpoint here, but there's no gains to be made in being nice to LLM's - be direct
  5. Use "for example" for specific output format you'd like it to generate.
  6. Build on previous prompts. The first output usually isn't quite right. Build on it with further prompts.
  7. Correct mistakes and give feedback. I tend to use "more like...." or "less like...." a lot.

I am currently working on pulling together a prompt library for useful prompts in achieving the use cases listed above. If you'd find this useful, contact me, and I'll include you in a beta release.

Long story short.

No, ChatGPT is nowhere near ready to lead a team, just as it isn't ready to coach one. Technically speaking, there are still some large limitations before it will be anywhere near ready for that. In the meantime, let's keep learning from each other and making use of the amazing tools we have available to us!