Generative AI concepts that every IT leader needs to know: Part 1

We delve into four key concepts that will help you think about AI strategically, commercially and technologically, to get value from your applications and secure buy-in across your business.

AI is not an IT project. Or at least it shouldn’t be, if you want it to be commercially successful. For AI to be truly transformational it needs to be a collaboration, with input and actions from multiple stakeholders to maximise the commercial results it brings to the business.  

But AI is still a technology, which means that the main driving force behind the project is going to be – you guessed it – you. The IT leader. Regardless of whether the impetus to deploy AI has come from you and your team or from elsewhere in the business, or whether you think the business is ready for AI from a technology and data perspective, the challenge of implementing this next-generation technology and making it a success will fall to IT.  

That challenge won’t just be picking the right software and plugging it in, either. You are going to need to think strategically, commercially and practically. You are going to have to win hearts and minds on one side, whilst managing escalating expectations on the other. You are going to have to address the limitations of your technology stack and your data, while spearheading exciting use-cases and wringing value out of every penny of AI expenditure.  

The question is, what do you need to know in order to do all of this well? How deeply to you need to understand AI to make it a commercial success? You don’t need a PhD in AI to make it work (strategically or technologically) but you do need to have enough depth of knowledge to understand the benefits, challenges, opportunities and practical limitations.  

In our experience, there are eight core AI concepts that every IT leader needs to know to do this. Get to grips with these concepts and you won’t just have a better understanding for deployment: you’ll be in a better position to lead the AI charge, engage stakeholders, steer strategy and secure ROI. 

Let’s start with the basic four first: context, thinking time, models and applications.  

Concept 1: Context  

Let’s start with a bit of background here. For a human being, understanding context in language is a crucial part of brain development. The prefrontal cortex that handles contextual understanding continues to develop well into your mid to late 20s. During that time, you’re constantly absorbing context and refining how you process it.  

An AI doesn’t have decades of nurturing, learning and making mistakes. It can only work with the information it is given. You need to provide your AI with context if you want to maximise how relevant and appropriate its responses are. To do this, you need to provide a source to ground the response in a specific ‘scope’ of information and set clear expectations for the response.  

For example, a human customer service agent might instinctively know how to respond to a frustrated customer, having learned to do so over time.

To get a similarly adjusted response from an AI-powered customer assistance application, you need to provide the context: that means feeding it with product manuals, tone of voice guidelines, response protocols, etc. The more context you provide in your prompt, the more accurate and naturalistic the response will be.  

Tip: Develop and use prompt templates to give consistency to how you provide context across the business. 

Concept 2: Thinking time  

Treat AI prompts like a conversation. The more back and forth you have with your AI, the better the result will be. AI doesn’t think all at once, it builds responses word by word and predicts the most likely next word based on the prompt. You can get more thoughtful, accurate answers by asking the AI to self-reflect or run multiple iterations, prompting the AI to think more broadly and creatively not just predict the most obvious solution.  

 Tip: Build thinking time and reflection into your AI usage. Ask your application to think more deeply after the initial prompt. Train your teams on how to do this and why it matters. 

Concept 3: Custom models  

Everyone needs to understand the difference between your AI model and its applications. The model is the brain that powers the thinking and reasoning behind the task. The applications are all the tasks that the brain can carry out.

To give an example everyone will understand: GPT is a model, ChatGPT is an application. As a model, GPT can be used to power everything from a conversation (like ChatGPT) to automated coding.

A lot of security concerns can be assuaged once your business understands that you can implement a custom model within your business that is isolated from public use. This is your AI. It can be deployed privately, leveraging your own data, documents, systems, and knowledge, and it can be industry, organisation or task specific. 

Tip: The key thing to remember is that the model itself, not just the application, can be isolated: a key point to communicate when considering privacy and security. 

Concept 4: Applications 

Applications are the tools that give you access to Gen AI models. Examples of applications include ChatGPT, Copilot, Midjourney and Synthesia, but applications can also be custom. The user experience of applications is evolving all the time, with conversational, context driven, personalised and dynamic user interfaces developing as AI becomes more advanced.

Models can be integrated into your custom applications or SaaS products like Salesforce, to enhance their efficiency or effectiveness, or they can be built into brand new customer applications.  

Tip: Bring stakeholders across the business into the process of developing use cases, that can then be turned into applications. Our approach is always ‘business first, technology second’. AI really feeds into this, as it has such vast potential that you can deploy it to action all sorts of business challenges or opportunities. Get people on board early and you will see much better engagement and buy-in.  

In our next blog, we will delve into four more concepts that every IT leader needs to know: business data, guardrails, agents and knowledge assets.  

In the meantime, if you’re exploring implementing AI into your business, now is a good time to bring diverse stakeholders together to think practically and creatively about your strategy. 

Whether you already have a clear idea of your AI plans, or aren’t sure where to get started, our AI workshops are a great way to get the ball rolling. We bring together key people from across the business to help you understand AI, apply it to your business, and come away with tangible, commercially viable use cases to explore further.

Book your commercial AI workshop.