Maksim Osipov, Manager of ride hailing business MENA region
AI is increasingly being integrated into every aspect of ride-hailing to drive efficiency and accuracy, lower costs, upgrade safety, and improve user experience. Here are just some of the areas in which we’re integrating AI into our ride-hailing services at inDrive, and where we’re looking to use it next.
Improving user experience, from support to timing, supply and user feedback
The days of wondering when your taxi will arrive are long gone. Now, ride-hailing services provide an estimated time for the driver’s arrival and when you’re likely to arrive at your destination – even when “unforeseen events” take place. Using pricing and matching models can account for local conditions such as traffic surges, sporting events, weather, and accidents (the more local data collected, the more accurate the prediction becomes).
These conditions can also affect the number of drivers available, and in turn customers’ ability to book rides. We can use this information to create heat maps, guiding drivers to hot spots to increase supply where it’s needed and better serve our customers (and our drivers, who benefit from more work).
AI is also improving customer service and support, much of which can be automated to improve self-service options: for example by reducing the amount of boilerplate material a customer has to read, and instead providing more focused, relevant information. This reduces customer wait times, improves efficiency, and frees up staff to focus on issues where they’re actually needed.
And when dealing with customer feedback, we use AI to cluster and categorize this information into analytical data that allows us to spot trends and infer customer sentiment and tone of voice. This helps us to highlight emerging issues and areas for improvement, so we can direct our efforts where they have the most impact.
In addition, when dealing with customers, AI can expedite decision-making for a variety of requests. For example:
• finding the appropriate specialist to address a customer’s request more quickly.
• resolving disputes about trip charges based on the history of trips, routes, and driver profiles.
• assisting in identifying and addressing unprofessional behavior from drivers or passengers, ensuring better service standards.
Getting the price right
inDrive differs from many of our competitors in that we use a peer-to-peer negotiation model, which lets drivers and passengers directly negotiate the price for a ride. We nevertheless use machine learning in our pricing models to improve the accuracy of our recommended price when customers bid for rides; this provides a starting point for negotiation that is fair to both customers and drivers, all things considered.
By using AI to automate manual pricing, we can react more quickly to dynamic conditions, so that our drivers improve their earnings when demand is high, and passengers can successfully book at a price point that matches their expectations.
In addition, AI can help adjust bidding steps, thereby expediting negotiations and achieving satisfactory results for both passengers and drivers.
Streamlining processes and spotting fakes
Internally, AI can be used to improve operational efficiency by streamlining processes, for example, during security checks. When a driver wants to register in the app, he or she must supply several documents, including ID and driver’s licence. These are manually and digitally verified by a dedicated team of professionals using different filters; now we’re also testing machine learning-based features to better identify fraudulent documents.
Ideally, this will allow us to spot fakes more reliably and more quickly, thereby increasing the safety and security of our users, and speeding up the verification process for legitimate drivers.
We also work with artificial intelligence to strengthen our security ecosystem in other ways: for example, in some countries we use a facial recognition tool to validate our users’ identities; and we use machine learning to review users’ profile images and exclude sensitive, potentially dangerous or commercial content.
The challenges
So AI and machine learning can make a considerable difference to the quality and safety of ride-hailing services; there are, of course, several challenges. For one thing, models tend to “drift”, slowly becoming less relevant over time, so that they require retraining. As such, we are working to improve learning capabilities so that models keep themselves updated – in effect, teaching them to teach themselves.
And because inDrive operates in many countries, we have to adapt to their different laws and regulations, balancing technological advancement with privacy protection and societal well-being on a regional basis. (For example, in countries such as Germany, personal data belongs to an individual, and neither companies nor regulators can approach it. In other countries, data belongs to the government.) So our approach varies in response to local rules and regulations; there’s no one-size-fits-all solution that applies globally.
In addition, as with any undertaking that collects personal data, it is crucial that we protect this. This can be achieved partially by obfuscating data in a way that conserves its contextual value, but hides the customer’s personally identifiable information (PII).
More generally, we ensure the privacy and security of data that we collect by limiting access to it to a strictly need-only basis. Our operations teams can’t access data in bulk, and may only use it for active support requests. In addition, customer-driver exchanges of PII are minimized, and only used to enable drivers and passengers to locate each other, and to improve the ride experience.
As in many other industries, AI and machine learning are enabling ride-hailing to evolve rapidly in quality, safety and efficiency, affecting every aspect of the business. Already, the use of AI has gone from a futuristic technology to a mainstay of the present; you’re experiencing its benefits every time you hail a ride.
