doings: Project Management with AI Success Guarantee

When projects fail, it is often due to poor project management. doings puts an end to this. The Berlin-based start-up uses AI to make projects more efficient. The doings team built the training pipeline in the AWS cloud together with the PCG team.

doings Software GmbH was founded in 2023 by Andreas Schäfer, who is now Managing Director. Johannes Ottmann later joined as co-founder and is CEO. doings builds AI-based “employees” that double the efficiency of project management. doings has been accepted into the exclusive Kiez.ai Accelerator, in which only a few start-ups are accepted.

We can read minds. Need proof? What do you think when you hear BER airport and Stuttgart 21? Years of delay and massive cost overruns? That wasn't too hard to guess because these two projects are textbook examples of disastrous project management. But they’re just the tip of the iceberg. Surveys show that approximately 60 percent of all projects fail. No company, no public administration is spared. Wherever people try to execute plans, failure is often inevitable.

There are many reasons for this, but one central issue is resource planning. Time and budget estimates are often calculated too tight and lead to overruns. Sometimes, the goals aren’t clearly defined from the start and keep changing. Other times, key decision-makers become involved too late and end up shutting down the project because they weren’t part of the original idea. Digitalization also seems to play a role. While it can help to maintain an overview, it can also make projects even more complex if you rely too heavily on it to solve all problems.

Artificial intelligence to the rescue! There are now thousands of tools that sift through data and create summaries or presentations. But much is still piecemeal. Microsoft's Copilot can do a lot, but it only has an overview of partial tasks. In addition, one disruptive factor remains: the human being. They have the unpleasant characteristic of misunderstanding colleagues or misusing projects for their own career plans. What is needed is an AI that takes a holistic view of a project and takes the human factor into account.

Meet Alex,the project manager, and Max, the product manager. They introduce themselves to potential customers on the website of the Berlin-based start-up doings. Alex writes reports, assigns tasks to project team members based on their skills and manages project plans. Max collects and consolidates feedback, ensuring that project goals are met. Alex and Max are available 24/7, and can be booked as often as needed because they’re AI bots specialized in project management.

Managing Director and co-founder Andreas Schäfer well knows the challenges faced by many software engineers, including those in his own circle. They often complain that, in addition to their development work, they have to manage projects - tasks they have neither the time nor the inclination to handle. It’s common for developers to be handed responsibility only after a project has been running for a while, typically when it’s in danger of failing or a leader exits the company. “Read the 100 Slack channels, look through 500 emails - and write an interim report on the project by Monday.” Schäfer frequently hears about these “suicide missions” from his acquaintances. The computer scientist wanted to develop an AI for them that would take some of the burden of project management off their shoulders.

Powered by doings software, Alex and Max are the solution to these challenges. Their AI can access any data, including outdated or unstructured information, through various interfaces. If an employee is thrown into a project at the deep end, the AI can sift through the Slack channels and emails, filtering out the noise to extract and summarize the key information into a concise dossier. “AI is predestined for this task, and it's a low-hanging fruit that brings a lot of efficiency gains,” says Schäfer.

doings goes further. One of its strengths is that the AI can use schedules and calendars to assess who is currently working at what capacity. At one customer - an IT consultancy with 700 employees - the AI even helps to curb the cutthroat mentality that is widespread there. When project requests came into Jira, team leaders rushed to land the best jobs and keep their team busy - on a first-come, first-served basis. “It was like a bazaar,” reports Schäfer. doings AI, on the other hand, distributes the requests according to the actual workload and the skills of the employees. “With us, orders are awarded more efficiently and fairly.” Employees also appreciate this. “We receive very positive feedback.”

According to Schäfer, this also has to do with the fact that the doings solution is easy to implement and use. Users don't need special AI knowledge. doings can be connected to countless systems and automatically triggered by them and then pass on data and results to them. The AI understands which inputs and outputs it needs, and how these are generated using natural language without programming. The AI agents break down your tasks, search for their tools independently and check the completeness and quality of your outputs - in other words, much more differentiated than a simple ChatGPT prompt.

Even if using it reminds one of ChatGPT, the customer does not have to be dependent on OpenAI. Many start-ups dock their AI solution to its large language model GPT4 via API. doings can do this too, but can also load more than 10,000 AI models, including many open source models such as LLaMA2 from Meta, and integrate them into its low-code modules - always according to the customer's requirements. Some of these only need a thousandth of ChatGPT's resources, but they can only do one specific task very well and not all others. In addition, the doings software does not answer one request at a time, but many at the same time. The system evaluates itself and improves the quality of its results over time. doings can work multimodally and combine several AI models with different types of data, such as a language model for written input and output with image recognition for quality assurance in a factory.

doings is still in its infancy. Schäfer and his colleague Johannes Ottmann launched the start-up in August 2023. The seven member team brought PCG on board to train the AI. “They are capable people,” says Schäfer with a wink - and by that he also means himself a little. Before founding doings, the computer scientist was a Cloud Solution Architect at PCG (former kreuzwerker) in Berlin for one and a half years. So they knew each other well and when Schäfer met Kristine Jetzke again at an AWS event, the collaboration was quickly agreed.

The result was a proof of concept generously funded by AWS. Now, 90 percent of the doings software operates in containers in the AWS cloud, including the backend with the AI model, and large parts of the frontend. The PCG team assisted in fine-tuning the Large Language Model LLaMA2 and developed scripts for communication. Requests are now processed through standardized workflows, though they handle different types of data. Another example is the IT consulting mentioned earlier. When a project request comes in, the document is uploaded into the AI, which calls on the various models as needed. The AI recognizes whether there are certain technical requirements and works through a checklist.

Development at doings is on-going. In the future, AI agents will automatically understand text tasks and divide them into sub-tasks for different models to maximize efficiency and tailor solutions to specific industries and use cases. The proof of concept is complete, the collaboration with PCG is over for now - but not forever. Andreas Schäfer: “If we need support again, PCG will definitely be our first choice.”

Public Cloud Group (PCG) supports companies in their digital transformation through the use of public cloud solutions.

With a product portfolio designed to accompany organisations of all sizes in their cloud journey and competence that is a synonym for highly qualified staff that clients and partners like to work with, PCG is positioned as a reliable and trustworthy partner for the hyperscalers, relevant and with repeatedly validated competence and credibility.

We have the highest partnership status with the three relevant hyperscalers: Amazon Web Services (AWS), Google, and Microsoft. As experienced providers, we advise our customers independently with cloud implementation, application development, and managed services.