During the project, we created Slipstream, a solution to support project managers in their role. At the start of the project, we came up with various ideas before deciding on the final idea. After that, we developed this idea further and researched various structures for this model. For our tool, it is important that it runs constantly and functions as an assistant, keeping track of all the project deadlines and requirements by using all information channels to extract the required information. By using this information and the project requirements, the model can provide an insight in how the project is progressing and if action must be taken to keep up with the deadlines.
I was responsible for identifying risks and limitations and developing an approach for future market testing. These tasks required me to not only consider what the tool was supposed to achieve, but also under which conditions it should operate to deliver value to project managers. Therefore, it was important to deeply understand what our model was capable of, this was mostly done through testing, and has identified where the model still has to improve in interpreting fragmented project information and turning it into a reliable support for decision-making. In addition to that, I also supported my teammates in their tasks and examined and explored various structures for our end-product during our meetings.
One important decision was made regarding the dashboard of our model. We had to find a balance between visual appeal and ease of use. This highlighted the importance of user experience alongside the core technical requirements of our tool. This also included considering how to accommodate different projects in the dashboard while keeping a structured overview. I also developed the future market approach to structure how our model can be further developed after initial completion. This included, for example how to structure the role of AI for different projects and how to assess the value created for client companies. Looking back at this process, I have recognised that evaluating an AI solution requires more than technical capabilities, as it also involves usability, reliability, and whether it assesses an actual user need. To further improve in this area, I would like to be able to connect the identified risks and limitations to concrete validation benchmarks and criteria for future development and market evaluation. This would create an easy summary of the performance of the model, creating valuable insights for us and for our client companies, as this proves the value of our model.
At the start of the project, I had no idea what to expect from the team assignments. Although we had a great start with some innovative ideas, the overall project seemed overwhelming, and I had no idea how and where to start. But during the team meetings and discussions we held, the idea became clear, and we started developing. This went well, but we ran into a problem where we had to establish trust from companies to adopt our model, as it uses valuable and maybe sensitive information. This is also the feedback we received on our initial project proposal. To address this, we not only wanted to create terms and conditions to address this, but we wanted to convince people. So, we reached out to current project managers to evaluate how our model performs against humans.
Even though our model is faster than a human, it also detected more errors in our small pilot dataset. This proved that our model worked and confirmed the potential of Slipstream. Although further testing is required to establish reliability across multiple and differing projects, this gave me more confidence in our model and this proof could be used when offering our model to companies.
The collaboration within our team went really well, we were constantly in contact with each other, either in person or online. For most of the project, we had meetings where we divided the work and discussed our progress. We had multiple moments during the project where collaboration and discussion helped resolve uncertainties, for example when we were discussing the structure of the slide deck. We had differing ideas regarding the distribution of slides across the topics. Some topics were more important than others, requiring more context or explanation. But because we had to fit everything in 15 slides, we had to create a distribution where everyone had enough room to explain their topic. But after a short meeting, we were able to create a structure everyone agreed with. This showed how open communication can help unite differing perspectives and foster collective decision-making.
During this project, I worked with students who had a different way of working, this gave me a valuable insight in how ways of working can differ per person. Some immediately shared some ideas we had about the project, while other students were first focused on fully understanding the project before generating ideas. This gave me a valuable insight in how a combination of very differing working styles and structures can work together to create the best outcome. This made me realise that effective teamwork does not require everyone to work in the same way, but rather how different approaches can complement each other. This will be helpful in future projects and my professional career, where working with people with different perspectives and working styles is crucial.
Before working on this project, I perceived AI as a tool that could improve the efficiency of business processes. However, throughout this project I realised that creating an AI solution is not the only thing that matters to an organisation. Organisations must be confident that the AI model produces reliable and accurate results while handling sensitive data appropriately. This made me realise that trust, data security, and reliability are just as important as the technical capabilities of the AI model.
Furthermore, this project made me realise that an AI tool must address a genuine business need to create value. For Slipstream, this means that it should not only identify potential risks and monitor deadlines, but it should also provide new insights that are useful to the person using the tool. While it is sometimes hard to provide measurable benefits, it is sometimes necessary to prove that the model works.
Overall, the project taught me valuable lessons about developing AI-driven business solutions. During demo day, I realised that AI can be applied in industries where I did not realise it could be useful. In future projects, I will focus more on establishing validation criteria early in the process, as this is something we realised too late, and clearly stating the user need we are addressing, as this is what is valuable to the customers/ organisations.
