26 Jan '22
26 Jan '22
27 Jan '22
“I have been working in the IT industry for over 10 years, and from my own experience I know that it is not always possible to determine the deadline with 100% accuracy. Especially if the project has a multi-level workflow tied to different departments, and the delivery of features is long and unpredictable. Therefore, the idea arose to optimize this process by training artificial intelligence”, - shared Maxim Konin, CEO of the project.The user of the system starts a task and estimates its complexity by a set of criteria that are personally selected for each project taking into account the specifics. For example, the field of programming, the need to make changes to the old code, the complexity of business logic, the need for integration with external systems. Then the system builds a time forecast for the passage of the task through the full working cycle. The workflow is also configurable for the project. The system monitors the progress of tasks and gives a visual warning when the deadline is shifted.
“The deadline for completing tasks is influenced by many factors that are difficult to assess when planning - from banal snacks to force majeure situations. A product must go through many stages before it is available to the end consumer. All this together makes project planning a rather non-trivial task, and the person in charge of forecasting is equated with a “soothsayer”. We decided to build an adequate, and most importantly, an impartial model of working with uncertainty, excluding the human factor.As part of A: START, the team conducted 10 problematic interviews with IT owners-companies both in Russia and in the USA. Four companies received early access to the system for training the neural network and further collaboration. A full-fledged web service is already ready, and now the developers continue to collect a data set for forecasting.
Of course, Agile and Scrum practices are now very popular. However, their disadvantages are a large amount of routine work and an error in estimation, which is done by a person, not a machine. Teams spend a lot of time (about 6 hours per developer) to evaluate tasks and control deadlines,” added Maxim Konin.
“The dataset is assembled for a specific product and company. When a team starts using our service, at the beginning there is a training period for the system. Further, the trained algorithm predicts timelines for tasks, the algorithm self-corrects when new data arrives. Now the system is used by one team of five people throughout the year. Four more companies will join them soon, ”Maxim Konin emphasized.In the future, the developers plan to integrate with Jira, Asana, Youtrack and other popular task trackers and place them in their public marketplaces. There are also plans to improve the algorithm and build more complex forecasts, not only for IT projects, but also for the instrument-making industry.