The Center for Data Innovation spoke with Milie Taing, president and founder of Lili.ai, an AI-based virtual assistant specialized in project management based in Paris. Taing discussed how Lili assists project teams in securing their projects, and why that is critical in project management.
Eline Chivot: Many organizations have started to use AI virtual assistants. What has led you to create Lili, and what makes it stand out from other similar systems?
Milie Taing: I created Lili after working for an engineering procurement construction management company in Vancouver. At the time, I was a specialist in Excel, I was teaching Excel and was a cost controller. One of my missions consisted in going through 18 years of archive to help trace back the responsibility of clients and my team, to be able to explain who owns how much to whom in order to compile a claim memorandum. As every single day of delay meant penalty, it was important to find evidence to explain what happened and why the company potentially shouldn’t be held liable for the delay, therefore limiting the penalty. In such cases, companies and contractors solve disputes through an informal talk or negotiation and then potentially leading to arbitrage. We had to go through arbitrage, and lost, which led to a loss of millions of dollars, and hundreds of people were laid off. It turns out that, as usual, both the company and the client shared responsibility for the delay, but we couldn’t prove this well enough at the time as the evidence was spread out over 18 years of archived textual information (such as emails, meeting minutes, work site journals, etc.). Had we better documented the project, we would have had more chances to win the case. So I decided to come up with a system that would retain corporate memory and capitalize on it, so that projects become easier and easier.
The positioning is what makes Lili different. There are many AI virtual assistants dedicated to many different segments, including in insurance, customer care, etc. What Lili achieves is to be at the crossing between natural language processing (NLP) and large-scale projects. And that’s unique in itself because even though AI is used for agile project management, very few startups are really working on large-scale projects, especially with NLP. We’re going through the depth of textual data, while many companies use prediction-based figures, data that is already structured/formalized, financial data, dates, relationships between tasks, etc.
Lili is an active player within incubator ecosystems. We are the only French startup that remains as one of the 30 selected finalists of the IBM Watson AI Xprize, in a competition that started with 10,000 startups working on AI to create a better world. Our promise is that if AI manages to solve project management issues we will achieve cheaper energy and better transportation systems.
Chivot: How has project management changed, and how what is the role of AI in this area?
Taing: One aspect of project management that has changed the most in recent years is the complexity, which has increased. I think any project manager can testify of this complexification of project management, it was actually the theme of a Project Management Institute report. Projects are becoming bigger and bigger, and as people have to deal with more and more data in real-time and greater technical complexity, you need more and more people, such as specialists, to deal with this complexity, in turn making everything even more complex. In parallel, another trend has been the professionalization of contract management, especially regarding penalties for project delays. According to an Arcadis report, the average continental European value of disputes is $41 million, and it’s steadily growing. Considering nine in ten projects tend to fail or do not deliver expectations within budget and on time, which has a huge financial impact already, and given every single day of delay adds up to a penalty, the loss in margin can be considerable. Preventing claims for such penalties can therefore make a significant difference for a company’s budget, and that requires a well-documented project.
Both trends are driving up costs and the financial impact of projects. And when you know projects are at the core of any industry and any economy, for instance in the energy sector, oil and gas industry, construction, or for hospitals, anything is a major project, this needs to be taken seriously.
Regarding the role of AI in this, we distinguish four areas where Lili’s AI can be a solution. First, AI can be used as a forensics assistant, helping humans see through gigabytes and years of data. A project can last five to ten years, and every single day more data is created through emails, risk registers, decision logs: This represents lots of data which humans have to crunch and go through to be able—for instance—to trace the responsibility of each other throughout a project. This forensics with AI is not only necessary because humans are completely overwhelmed, but also because from a financial perspective, the fact that we’re able to trace back to what can explain the complexity of a situation is importantly related to the potential penalties companies might be facing, as mentioned earlier.
Second, AI can assist in enhancing the reliability of data. For example, when typing an email or meeting minutes, you may type things such as “There is an issue with the turbine.” It is possible that you won’t add any more detail to this. So the machine will prompt you with “Which turbine are you talking about—the first or the second?” This makes a huge difference, because the better the data, the better the results.
Third, AI can provide real-time alerts to detect whether the tone of communication with contractors is changing and suggests that there may be something to look at. Those alerts can prevent project teams from forgetting or overlooking important risks.
Fourth, AI can help in building scenarios and predictions. Compared with the other types of solutions which can be implemented over a period of six to nine months, this fourth type can be a little trickier to implement and ensure change. But overall, claimers do manage to build a very strong case within a few weeks, depending on how fast they can gather the data.
Chivot: What kind of data do you collect from users, and how do you structure it to run your algorithms?
Taing: The data we collect includes anything that is written, such as decision logs, risk registers, contracts, technical reports, emails, anything where there might be traces of what happens in a project—evidence of complexity, of responsibility. The algorithm then extracts the data as raw text from the initial format (Word, Excel, PowerPoint, etc.). We enrich that, and then we add an automated level of analysis. After which, the algorithm is responsible for the emergence of a profile of the project.
Chivot: Which organizations are most likely to use your system?
Taing: We work with CAC 40 companies (the French stock market index, grouping the country’s 40 largest companies), particularly those of industries such as in military, construction, aeronautics, and energy. Lili is a great fit for any organization that has to manage very large-scale projects. One might assume that most of the time corporations don’t like to collaborate with startups. In our case, this is no problem at all: They are very open to us and accessible, because they are starting to understand the urgency of AI in project management. The success of a startup’s innovation depends on various things, one being the scale of the problem it is aiming to solve. As Lili tackles an issue that is pretty significant in terms of financial impact, we have found a great niche.
Chivot: How have you seen your customers use Lili successfully? What does long-term collaboration with them look like?
Taing: Every day we receive testimonies of clients that are very happy because they find the system incredibly useful. Things that used to take weeks, and elements or issues that may or may not come out depending on how lucky they are, will now appear straight away within seconds. That’s our proudest achievement and the reason why we want to continue disrupt project management. Also, we see cases when claim specialists find strong arguments and gather solid evidence to defend their case in case of undue claims and disputes, including with 360-degree checks over the entire documentation of the project at stake, within minutes.
We focus on claims because there is a huge need there, and it’s a great structured way to decrease the cost of projects. Also, although every project can learn from previous projects, this benefit is often overlooked and that history is rarely used when undertaking a new project. With Lili, companies can build organizational memory and learn with the data gathered. Cross-industrial understanding of project management is actually critical. All of our clients across a range of industries agree to share their models, because they all acknowledge that robust models are important and cross-sectoral learning can make a difference. Each of them needs it.
Lili certainly is aiming to create long-term collaboration as we want to go towards real-time project management and be able to build predictions, not just limit the project assistance we provide to the phase of disputes. We can start intervening as soon as we detect an issue, to trace back responsibilities. With large scale projects there’s lots of delays and budget overrun because we’re at the edge of what humans are able to do in terms of coordination, of management, of techniques, so we do start with the claims, to build the strongest, most robust model possible—which we can then use to transfer slowly these results and knowledge towards real-time project management and predictions, thereby engaging in long-term collaboration with our clients.
The ultimate goal is to have an AI assistant that has lived through 20,000 different projects, can monitor for the project team its blind spots with the flair of a senior project manager, and alert them before the typical problems arise.