ChatGPT Elevated Our Expectations

February 15, 2023

ChatGPT is a revolutionary new language processing technology that has revolutionized how artificial intelligence can interact with people. Developed by OpenAI, ChatGPT uses natural language processing techniques to allow AI agents to engage in conversation with users in a more human-like way.

But, generative AI like this eases one step in a business process.

Your business processes span multiple people, teams, technologies, and AI. How will you implement conversational tech like ChatGPT into a process?

John Michelsen, Chris Kraus, and I discuss how ChatGPT helped the world realize how far this technology has come and what you need to prepare for when bringing it into your enterprise.

Where is AI like ChatGPT headed?

ChatGPT is a fantastic example of how AI can raise the bar for human-computer interactions. It’s inspiring to see it maturing due to Moore’s Law and exciting to see more general AI models like Google’s A.I. Test Kitchen being developed as well. It is important not to get carried away with overhyping AI, but the potential of this technology is clear. We must continue to strive for intelligent machines that can understand human needs, rather than just build understandable technology.

ChatGPT created a new level of technology awareness

ChatGPT has created a new level of technology awareness for the general population. It’s different from Google because it gives you an answer that sounds more like everyday conversations, such as asking questions about homework and children. Even non-technical neighbors and friends are amazed by this technology, and I’m sure offshore coders can use ChatGPT to write their products. This has changed everyone’s perception of AI and I believe many can see how it can apply to their businesses.

What ChatGPT can’t do

ChatGPT has certain limitations due to it being trained with general knowledge from the internet. From a business perspective, it won’t be able to answer very particular questions that are not available online, such as an employee handbook or internal resources. Additionally, while some data on the internet is reliable, other data might not be trustworthy. For example, we asked ‘How many gold medals did the US win in the Olympics?’ and ChatGPT answered ‘113’, this would be a reasonable answer but not correct as that was from the Tokyo Olympics, not Beijing. This accuracy is what concerns leaders looking to implement this type of AI at the enterprise level.

What is the enterprise use case?

The enterprise use case of generative AI like this is similar to Google or Bing search. However, identifying sources today is lost but that will improve. OpenAI is focused on the right problems like taking on bias and cleanliness into consideration so it mistakes when the context doesn’t provide enough information for the question. But, enterprise technology leaders need to transition from building apps with this technology and creating more maintenance issues to creating efficient systems.

Are other AI models going to proliferate inside a business and cause more issues like too many apps?

The real value of AI technology comes from using it with a process automation fabric that envelops all of the existing system capabilities and people capabilities. We’re not just talking to one computer, but rather weaving together all the computers within a business with all its people. This is why we are leveraging AI from Open AI – it’s enriching our product and the products of others. We are still at the beginning of this journey, but it’s clear to us that this is where we need to go – a conversational interface that can reach beyond basic task-level satisfaction. That means being able to integrate with production planning systems and manager approval processes for a user to get their answer.

We take the first step by seeing what a chatbot can do, but this is only the beginning. We need to create an enterprise platform that can handle complex tasks and conversations, understanding nuances of policy and procedure, while also being able to confirm data pulled from a production planning system. To do this, we need to combine the people and their capabilities within an organization with the systems that are in place. We can’t just rely on conversations with a computer – it needs to go beyond that. This is our target, where we want to be. We are looking forward to the possibilities this kind of technology can bring.

The goal is to get these technologies working together so that they can automate processes, while also understanding policy and procedure nuances like when a human would use their experience and intuition. This combination of people, systems, and AI will provide an incredibly powerful enterprise platform that can help automate processes in a way that is accurate, efficient, and reliable.

AI-generated code doesn’t deliver business outcomes

AI code may seem like an easy solution to many business problems, but it won’t actually deliver the desired outcomes. Code is only useful when it comes to creating something new, not maintaining it – 90% of tech work is maintenance! Working with AI-generated code can be likened to a dog trick: impressive on the surface, but ultimately not going to get the job done. What’s needed is a higher level of abstraction: modeling processes end-to-end, and connecting tech to form the fabric that can provide all the underlying capabilities – this is something that cannot be spoken into existence. It goes beyond code, compilers, and dev environments. AI code may seem impressive but it’s not enough to solve the most complex business challenges.

People are still part of the process

People are still part of the process and so data look-up doesn’t have the concept of having someone be asked for decisions or approvals. It could simulate your boss but it doesn’t actually get you into the enterprise, I need someone to approve this right. Systems need to be updated via a conversation and that requires integrating processes with this type of interface. So clearly we’ve got work to do here and that’s not an open AI issue or a generative AI issue, but how do we best consume this tech and weave it into the things that we already have in terms of people and system capabilities to create better outcomes? That’s what we need to be thinking and until then it’s making it so the kids don’t have to take tests anymore.

The AI needs to understand the context to deliver a customer’s outcome

Google taught us that a simple text box would be all we need to fetch information instead of Yahoo’s indexing of the entire internet. This increases the demand for this type of solution-oriented approach in the business community and it is great, however, there is still much left to do. For AI to deliver customer outcomes, it needs to understand context. Even in our cognitive issue resolution case which is simplified email triage, the AI needs to understand what is being discussed in the email; whether it be moving a delivery date or having certain items in stock and then delivering the outcome. There are a bunch of chatbot-based tools popping up on the market that can recuperate the email content that it has already seen or been told, but it cannot reach out to the back-end system to do any more than that. The real value in understanding is so it can do something; linking together the ‘do something’ part. This is why we need an interface other than a huge screen displaying dashboard digits; we need an interface that can ask us, and understand the customer’s outcome. This is the future of AI.

Where should people start with chat-like AI models?

If you want to start with chat-like AI models, we suggest starting by curating the specific information and training it on your company’s data. That way you can tell people that it’s on the website or in their internal employee policies which they would not have read otherwise. And this way of having a conversation will be easier for them as it doesn’t solve outcomes but it can help people understand the information easily. If you want to make the conversation tied to actions and multiple steps, then that’s when we get into an interesting part. But if you want to take small steps, it is possible to do this and gain some success simply. What I mean by this is shifting the burden from someone in HR asking the questions to having an AI model answer them. That’s a very simple use case that everyone can understand and get behind.

Closing thoughts

I love ChatGPT and all the tech that follows in its footsteps. We are already playing with one from Google, and it’s bringing us closer to that Star Trek experience we know is where we belong. We’re not quite at the level of hype yet, but it’s still a good thing! Not only are we building our own AI, but we’re also consuming it to enhance our products which I love. If you have a problem in your app, try solving it by pulling in ChatGPT – that way it can proliferate into another great mess! All-in-all, the AI revolution is definitely here and will bring us closer to the future we’ve always dreamed of.

Links and Resources


Scott King

Scott King

Chief Marketer @ Krista

John Michelsen

Chief Geek @ Krista

Chris Kraus

VP Product @ Krista

Close Bitnami banner
Close Bitnami banner