How to Start with Generative AI

One of the hot topics with Generative AI is how-to implement it in your organisation. Where and how to use it. Everyone is talking about generative AI, but few are using it more than just offering ChatGPT Team subsription or maybe Co-Pilot from Office 365 family. Don’t get me wrong, thats good if you are even that far. Means that you’ve been thinking about this and have taken concrete steps towards it. But when we think further than just offering these great AI tools to help our employees to be more productive, we need to look deeper. What different processes, value propositions and brand values do we have? How should we arrange and change the way we work, to be called generative AI driven organisation in the future?

Starting with Generative AI

As you know your business best, it means you have a job here. It’s hard for outsiders come into your business, understand it fully and then make suggestions. For sure there’s a lot of consultants on the markets now telling you how to prompt, how to use AI and all sorts of things. But very few of them goes really deeper with their offering. And when going deeper it starts to cost a lot of money. So I think the big four and companies like McKinsey love to offer you AI consulting tailored for your business processes, but the price tag is astronomical. So as always, you need to get a hang of the basic in on the technology side and the concepts. This will help you understand the capabilities and limitations. There are many things that are already, let’s say “there” with generative AI, but not maybe ready for full enterprise usage still. So get yourself up-to-speed with the current state and what’s happening on the markets. Spring time might well be the time you’re reviewing your strategies and I’m pretty sure there’s not a single leadership team that will now cover generative AI somehow in their strategies for the coming years. Write down the short- and long-term goals for generative AI to build the basics for your path forward to implementation.

Moving forward to testing

When you’ve established the basic knowledge and you’ve written down a plan what would be the first steps to take it’s time to move forward to actually planning how to start. As in every technology project you need to start somewhere and start testing it. Crucial part for generative AI is the current status of your data and application landscape you have. I would not recommend to start with the most demanding use-case involving several data sources and API’s, as this will increase the cost just in the needed man hours for the work. Of course if you have e.g. and enterprise application integration platform in place and could easily add data flows towards, you might actually get the data sources in order more easily. My recommendation is to also choose a platform to operate your Generative AI processes. If you select point-to-point solutions for different use-cases that will increase you costs and scatter your organisation data in different places. When you expand on this, it will lead to a situation where you don’t know more where you’re data is going. I’d like to compare this to application integration. It started with point-to-point integrations and then moved towards integration platforms to manage the processes, monitor them and allow application replacement easily.

First Pilot Usage

You’ve all set up with your data and processes and published the internal chat interface for you model. That’s the phase where the testing and pilot usage starts from. In this phase you need have your team in place who’s responsible of testing and analysing the solution. The easy use-cases will probably work out-of-the-box, but then users move to more advanced use-cases, you need to closely monitor first that everything is working as planned. Testing generative AI models is something that it’s not talked so much, but it’s a huge part of high quality solutions. People are still quite forgiving to hallusination and silly answers, but moving forward let’s say 12-24 months that’s not the case anymore. Generative AI will come a commodity that is expected to work correct all the time. And if that user experience is not met, it will lower the trust score of the solution and lead to people not using the solutions provided to them anymore. Or that users will find alternative solutions to solve the problems they have. So remember to test!