Using AI in real life - an example of CHAT GPT
Like many others, I have been excited about ChatGPT and have conducted some wild experiments with it. After the initial excitement, I decided to test it with a past digital solution development case that I had been accountable for, naturally with some ambiguity for NDA reasons.
My exercise began by defining the business objective in one sentence with four distinctive words to set the context and the target. Later the only data I entered were example hourly costs for the project team members - all other data was obtained by AI itself. Then I started to ask AI to produce various outcomes.
First the AI created a prioritized product backlog with well-formulated user stories, showing a surprisingly good understanding of the business activities linked to the context. It also created initial estimates and clearly defined acceptance criteria for the stories. It automatically took information security into account on the story level.
The AI listed the project's concrete deliverables, defined the size and needed roles of the cross-functional scrum team with technology skills specific for the case, and created indicative sprint scheduling, taking MVP scope into account there. It identified potential challenges and risks that may arise during the project and offered solutions to mitigate them.
The AI was able to split the user stories into sprint tasks and create valid example code for the coding tasks. It also created an overall technical architecture for the solution with actual product proposals for its elements.
AI was able to propose disruptive and innovative ideas for the solution to stand out in the market - some of them matching the ones we identified during the actual solution development earlier, some went even beyond. AI was also able to propose Sustainability and ESG (Environmental, Social, Governance) related actions for the solution and project execution.
The AI created a good quality Value Proposal and a Business Case Analysis containing the problem statement, proposed solution, market analysis matching with Gartner, benefits, ROI calculation, and conclusion with a suggestion to proceed with the case. It provided several concrete measurements that can be used to measure the value created by the project and the solution itself. It also created a press release for announcing the launch of the solution.
For the customer-vendor perspective, the AI provided a detailed substance description for the Statement of Work and a sales proposal presentation structure matching good practices. When asked, it even created a flamboyant Shark Tank sales pitch.
All this made perfect sense - as I was able to compare to the understanding of the earlier real life case. The ballpark was definitely there, and the answers were very well structured and articulated. All in all AI gave surprisingly good amounts of data points, showed valid business understanding and had good structure in presenting the facts. I have definitely seen worse in real life - and created some of those myself when the time and context expertise has been limited. For a quick & dirty try the results were really impressive - having one sentence as a starting point. Of course I knew pretty well what to ask. And yes: a true case would require f. Ex more context related input data, and the AI solution would have to be able to handle confidential data.
After getting access to the ChatGPT-powered Bing I I checked some of the questions with it. The answers were similar, but the experience with Edge in Win PC was not as fluent as with the ChatGPT interface - after a couple of questions there was constantly a need to switch to a ‘New topic’, and therefore the process was not as iterative and incremental. However, the suggested technology architecture was more advanced with f. Ex. Machine Learning, and it would have been interesting to evaluate that with the past project team.
Of course this kind of result is just the initial understanding and starting point for a case - and the outcomes created by AI would require careful verification before gaining enough trust. The human factor is there strongly - the key aspects of agile development are iteration, human collaboration, and continuous learning and improvement. There always pops up surprises to react to. Due to covid we have learned to work remotely more efficiently, but it still is a very kick-ass move to bring people physically together - especially in a project start - for the people to learn to know each other, establish common ways of working, brainstorming and innovating.
However, it is clear to see where we are going. Just as an example, this could enable evaluating an initial business case immediately in a brainstorming meeting, which otherwise could require hours of meetings over various oceans with distributed organizations. We can make better decisions faster. We can also improve productivity, shorten case lifecycle, boost innovation, and increase customer and market understanding. Just to scratch the surface.
This also suggests where the people should focus on: understanding and defining business opportunity or challenge with related target setting and innovation, people collaboration and leadership.
Comment by ChatGPT itself: “Overall, while AI can provide valuable support to the development process, it is unlikely to replace humans entirely. Human decision-making, leadership, creativity, and problem-solving skills will remain critical to the success of a project.”
However, the progress seems to be quite rapid now, and we need to stay alert for what our competitive edge is as individuals, companies and society, such as where to put education efforts and how to utilize the expanding potential in the best possible way. "To boldly go where no one has gone before."
Blog is written by Mika Lampinen on behalf of Calia IT recruitment company. Mika is seasoned IT leader and has experience in large international organizations such as Amazon, CGI, Nokia, Tieto etc.
