Founded in 2015, Olivia AI aims to create a personal financial management assistant to deliver professional and personalized services to personal finance users. Olivia AI aims to serve its customers using a robot chat image which is driven by artificial intelligence, financial management knowledge and insights from behavioral economics. By learning the user's financial and spending habits, the company manages the fund accounts of all users in a unified way. The user's financial management and consumption plans are developed based on user-characteristics. Olivia AI’s team members include a wide range of professionals required for the product, including AI software engineers, financial data scientists, and cognitive researchers, linguists and designers who are proficient in human interaction design.
At present, the platform is still in its initial stage. In April 2017, the platform had just officially launched its business. Currently, it has issued several funds with an average amount of 3,000 yuan. Initially, it only provides services in San Francisco, San Jose, Los Angeles and San Diego in California, and it will expand to the whole United States in the future.
On May 12, 2016, the company received US$500,000 in seed funding.
Cristiano Oliveira, founder of Olivia, is a technical and serial entrepreneur. He graduated from Universidade de São Paulo with a master's degree in industrial engineering and a bachelor's degree in robotics and mechanical engineering. He served as CTO at Spring Wireless and founded Alaska Technology Ventures and online lending platform LendFox. Oliveira is also an advisor to Guidrr and served as vice president at the Commercial Application Department of Kony Solutions. He has strong interest in fintech, artificial intelligence and mobile applications.
Olivia AI is an artificial intelligence-based financial chat assistant. Olivia AI can instantly connect to registered user's all funds account. The user only needs to choose to agree to Olivia-managed accounts, and then Olivia will be able to obtain the user's transaction flow through the bank-level secure channel without obtaining other private personal information. Olivia uses basic financial planning capabilities to monitor consumer spending and give reasonable advice. The operation is simple and the interface is clear and tidy. Olivia provides account summaries, daily spending limits and current costs. Daily wealth management monitoring provides users with more convenient and actionable advice.
Olivia collects user transactions, mobile application data, and geographic location data. Olivia asks the user whether their consumer behavior is normal, and gives consumer recommendations to the user, for example, how much money can I spend today without overconsumption? How much money is free to control this month? By giving visual information including charts, users can see their own consumption easily. Olivia does not give too much appraisal of the user's spending habits and will only give suggestions for reference. Users can consult Olivia before spending again on whether the price of related products is appropriate and affordable. When users are have over-consumed, Olivia will also suggest how to save money to recover sustainable consumption levels.
Olivia helps users save money in more than 100 ways, covering entertainment, travel, transportation, meals and other aspects of life. For example, Olivia will consider recommending Uber to users with a low income when they spend too much time on parking per month. For the user's overhead on daily items, Olivia will advise the user to create a shopping list, analyze the user's shopping preferences, and tell the user when and which nearby stores have discounted activities related to the product. In the future, Olivia will also optimize and adjust the timing and frequency of alerts and warnings to users based on more user feedback data to optimize the user experience.
Self-developed daily consumption advice: Olivia provides account coverage, daily spending limits and current costs. By developing wealth management insights each day, users get convenient and actionable advice, such as daily maximum spending, daily recommended ways to save money, etc.
Oliva’s chat robot interacts with the user more like an option: Unlike other chat robots, Olivia does not perform natural language processing, but communicates and interacts with the user in a selective manner instead.
Below is our interview with Cristiano Oliveira, Founder and CEO of Olivia AI:
Q: Does Olivia also help users achieve deposit goals through automatic deposit methods?
A: Olivia doesn’t use automatic deposit methods. We know that automatic deposit is a common solution, but this mandatory practice can only save 200 to 300 US dollars for each customer per year, and users usually cannot persist. Olivia tries to nudge users through their spending habits.
Q: What is the current user size?
A: Olivia has 30,000 users currently. Moreover, Olivia has also obtained 1 billion transaction data points for the pre-learning of machine learning training. 100,000 users are the base criteria needed for a fintech company to survive. Olivia hopes to reach this goal by the end of this year. The industry admits a company is an excellent fintech project only if it exceeds one million users.
Q: What machine learning models and algorithms is Olivia using to support the business?
A: There are more than 30 machine learning models in Olivia, which can be roughly divided into six steps to implement its business.
The first step is to use a classifier to classify each transaction of the user. For a definite transaction, a decision tree can be used in completing the classification. For transactions that do not have a specific purpose, heuristic algorithms and probability models are used to give the probabilities of various possibilities. Besides, using the learning of pre-existing transaction data, we can judge the specific consumption purpose through machine learning methods.
The second step is to analyze the user's future behavioral through time series analysis.
The third step is to use fuzzy logic to infer where the user usually spends.
The fourth step is to use regression to predict future consumption.
The fifth step is that the expert system gives personal financial advice to users with different consumer preferences and income conditions.
The sixth step is to control the dialogue content that will be presented to the user and the order of presentation using the decision tree.
Q: Does the dialogue interactive model use Natural Language Processing (NLP) technology?
A: Olivia does not use NLP to deal with any of the user's natural language communications. It only communicates with the user by giving communication options. This is because NLP's comprehension ability often doesn’t have such a good performance, and it’s understanding is prone to ambiguity. Instead, giving the user options can guide users to understand Olivia's features more clearly.
It is also worth mentioning that Olivia's interaction process has undergone detailed behavioral design. Olivia does not provide users with options to decide later because many users are prone to delaying or even ignoring the problem. Usually, such options will not be paid attention or be well implemented by the user.