It wasn’t long ago that if I mentioned ChatGPT, the response would be “What’s that?”. But at my wife’s 2023 Christmas dinner with her friends, who come from a diverse range of industries, ChatGPT and how it could change the world was the hot topic. This shift illustrates the deep penetration of the technology into everyday discourse. Except for Threads, owned by Meta, which already had 3 billion users worldwide using its other products, ChatGPT is the fastest-growing app of all time.

ChatGPT is a commercial product developed and owned by OpenAI, a non-profit organisation dedicated to advancing AI research and is one of many such AIs now on the market. ChatGPT is the tool we all see and use but behind this is the OpenAI AI LLM’s (Large Language Models) and LMM’s (Large Multimodal Models) known as GPT (Generative Pre-trained Transformer) models. 

In my opinion, ChatGPT has a bit of an image problem right now. So much of the content and so many of the webinars we see out in the wild are focused on using ChatGPT for marketing content generation and copyrighting, but these use cases only utilise a small part of what this technology is truly capable of. Sure, they are good use cases for introducing people to the technology, but a cautious approach should be taken when your voice and identity are at risk. It’s too easy to lose authenticity when an AI tool is writing all your content for you and you simply copy, paste and publish. The reality is that for many businesses adopting GPT models it will form only a minor portion of their GPT model usage. It’s also highly likely that GPT models won’t be adopted in isolation and other AIs will long term form part of a business's strategy.

Before exploring some potential use cases there’s a couple of important OpenAI API features that I’d like to briefly call out: -

• Assistants - allows you to build AI assistants within your own applications. You can define an assistant and then leverage the GPT models, tools, and knowledge to respond to user queries.
• Knowledge Retrieval - allows you to create solutions that are augmented with knowledge from outside the GPT model, for example you can upload documents specific to your company and an OpenAI API assistant can reference those in the conversation. 
• Function Calling - allows you to tell the chat or assistant about code functions available to your solution, for example you might have an API to check a customer’s order status that you want the AI to know about. While the AI itself won’t call your check order API, it can instruct your application to do so and then make use of the results.
• Code Interpreter - allows the assistants API to write and run code in a sandboxed environment (i.e. an environment that’s only used for your session). This tool can process data files and do several things such as check data quality, create graphs from the data based on your requirements or use the data to create another file – perhaps an Excel spreadsheet with a consolidated view or a power point presentation. 

So, how can businesses adopt GPT models effectively?

Sales
If you're selling B2B, then the likelihood is you create proposals or tenders for some projects. Depending on the industry and your business, these proposals can vary in complexity, can be daunting, and are certainly time-consuming.

What if you could take all your old proposals and tenders, securely provide them to the AI, and then use that AI to accelerate the proposal and tender generation process? Businesses are doing that now; developing solutions using the OpenAI APIs. These businesses can respond to RFQs quicker than those that haven’t, and be more prolific in the number of RFQs being responded to. These businesses have reduced the constraints they faced in proposal and tender preparation.

Online Customer Experience 
For those operating e-commerce platforms or online services, enhancing customer experience and support is a perpetual goal. Traditional chatbots have their limitations, often able to handle only a narrow set of queries before requiring human intervention. GPT-powered chatbots are a game-changer, capable of pulling data securely from your systems to answer a wide array of inquiries regarding products, orders, and more. Additionally, GPT models and other specialist search AIs can be used to enhance search functionality, whether that’s searching a company's published documents for the answer to your question or returning more relevant product search results.

Call Centres
The integration of GPT models into call centres can revolutionise customer service, presenting various levels of adoption tailored to the needs and strategies of businesses. At one end of the spectrum, we have a fully automated AI system capable of handling Tier 1 inquiries, using voice to text and text to voice and being able to translate to many different languages. This AI acts as the first line of interaction, efficiently managing common questions and issues, allowing human agents to focus on more complex and nuanced customer needs.

For organisations favouring a more balanced approach, partial adoption comes into play. In this setup, GPT models act as a dynamic support tool for customer service agents. It quickly sifts through information and provides agents with accurate data, enabling them to address customer queries more effectively. This not only enhances the efficiency of the agents but also improves the overall customer experience by reducing wait times and increasing the quality of support provided.

Beyond these direct customer interactions, GPT models bring value in evaluating and refining the quality of customer service. The technology can be employed to analyse and rate the performance of service calls, offering insights into the effectiveness of agents and identifying areas for improvement. This continuous feedback loop can drive training and development, ensuring that customer service teams are not only well-informed and responsive but also consistently evolving and improving in their roles.

HR & Recruitment
The recruitment process, often consumed with the task of sifting through countless CV’s and conducting numerous interviews, can be streamlined with a GPT model. It can swiftly analyse CV’s, match job descriptions with candidates' qualifications, and even conduct preliminary screening conversations, enhancing the overall efficiency and effectiveness of the hiring process.

Companies often end up with many documents covering various aspects of HR procedures and policies, not to mention on-boarding guides. Companies can securely upload these documents to the AI so employees can get instant responses regarding HR matters. This alleviates HR teams to focus on the important work and the work where human interaction is really needed rather than answering a question about holiday entitlement or social media usage guidelines. 

Finance
Companies often handle large amounts of customer data, perhaps it’s customers providing recent utility bills and other documents for KYC regulations, or perhaps it’s bank statements to prove income and expenditure. While document digitisation technology has been around for a long time, GPT models can be used to better contextualise this information, for example by understanding the nuance between an account number and a reference number on a letter, increasing the accuracy of the information collected and reducing the need for manual intervention.

Knowledge Retrieval
Training and knowledge base material could be securely uploaded to the AI so that users can ask topic specific questions e.g. You are an international transportation company, having to adhere to many rules concerning export. Your GPT model could allow users to ask what the rules are when shipping product x to destination y and the AI will use the documents available to it to answer the question.

Data Analytics
GPT models are the Swiss army knives of AI; they are good at doing many things but in some situations, they won’t excel. There are simply better AI solutions on the market, or the GPT model isn’t capable of fulfilling the requirement. As an example, a manufacturer who wants to perform anomaly detection on their production data could use a GPT model to help with this but the reality is it would be a very poor choice – it wouldn’t allow real-time scenarios, it would be expensive, and the model isn’t optimised for this use case. Instead, that factory should be looking at solutions like the anomaly detection service within Microsoft Azure AI Services. Using a GPT for ad-hoc data analysis, however can save time and provide some powerful analytical capabilities for people not trained in data science. 

Conclusion
The above examples of how businesses could adopt GPT models isn’t exhaustive by any means, but instead is aimed at giving you a few examples to get you thinking about how your business could better utilise the technology. 

Of course, there are many considerations when evaluating the fit of any AI into your business, but the most important aspects should always be the privacy, security and ethical considerations including bias. You’ll notice when I mentioned use cases that involve providing access to your data it was prefixed with the word securely.

Firstly the security and privacy terms of ChatGPT vs the OpenAI API’s are different so it’s worth checking out the latest information at GPT base and OpenAI to understand risks, but in essence when using the OpenAI API’s your conversational data and uploaded files are not used to train future models, which is enabling large enterprises to build secure OpenAI API based solutions. 

The potential of GPT models and other AI’s stretches far beyond what we’ve touched upon today, promising not just to revolutionise business operations but also to redefine the landscape of industry competitiveness. Remember, AI probably won't take your job, but someone using AI probably will.

Don't wait for the future to come to you. Start exploring how you can harness the power of AI today. At Intelligent Industries we offer a free initial consultation so you can do just that.