AI Use Cases
AI Use Cases
There are a wide number of real-world problems being solved with AI with those in automotive, financial services, and healthcare leading the way.
Join the DZone community and get the full member experience.Join For Free
To gather insights on the state of artificial intelligence (AI) and all of its sub-segments — machine learning (ML), natural language processing (NLP), deep learning (DL), robotic process automation (RPA), regression, et al, we talked to 21 executives who are implementing AI in their own organization and helping others understand how AI can help their business. We began by asking, "What are some real-world problems you or your clients are solving with AI?" Here's what they told us:
- Little things that are surprising to some people is that ML is being used in really tiny situations to make things cleaner, smoother, smaller, and nicer. Accelerators on cars are no longer connected to a throttle. Your foot is pushing on data, not on a cable. The thing that turns what you push is partly physics-based engineering programs and partly little pieces of AI/ML. Your motion has the effect you expect. That is not simple. Control motions that feel intimately strongly connected are great applications of ML. Mega things that integrate with other kinds of software. ML is learning the emergent behavior of the large-scale systems so it can predict without doing the detailed simulations and we can move through the design space more fluidly. I continue to be astonished by the progress of ML in linguistics applications. The voice to text and text to voice are getting so incredibly good that people are beginning to use image recognition in some really innovative ways. They’re integrating them with other systems as well. A lot of sensors on cars are image based with an inherent estimate of the state with multiple views of the scene. The best AI/ML learning isn’t even visible.
- 1) Industrial use case Ham-let with a mechanical valve with Siemens with sensors in the valve and sensor outputs one of the sensors is a microphone we listen to hear if the valve is opening and closing and measuring the time to open and close and be able to predict when it will need service. 2) Koito, leading maker of lighting for automotive smart adaptive driving beam (ADB) headlights, uses an array of LED lights. When cars are coming, cut out the beam that would make it hard for that person to see. Make algorithm work on an inexpensive piece of hardware real-time on the edge on cheap hardware.
- We are involved in a diverse set of projects related to AI use cases, including training models for autonomous vehicles, medical imaging, and homeland security, genomic analysis, risk, and fraud detection to name a few. In one case, a premier autonomous vehicle company is using machine learning to train and refine algorithms for use in self-driving cars. In another case, genomic analysis is being used as an early detection system for bio-terrorism. In yet another case, image data from pathological samples is being rapidly scanned to detect early stages of cancer. Through earlier detection and treatment, the hope is that these cancers can be eliminated before they can gain a strong foothold.
- A large number of our clients are in the financial services companies with applications for fraud detection and loan approval. We help institute a practice in their organization where they can focus on day-to-day nuts and bolts of operational ML but also the compliance and governance aspects. Understand what ML initiatives are being deployed, understand the datasets being used, check the health of the algorithms, if there are odd prediction patterns they can tell immediately. How to manage ML pipeline for compliance reports. Multiple assessments: A/B testing, new models challenging existing models, canary with a simple algorithm to match more sophisticated algorithms to look for shifts in behavior. Another category is the notion of drift. Most ML are supervised and trained. In production, you don’t control data and all sorts of scenarios. Help customers understand if algorithms are seeing data patterns that were not seen in training.
- Switch from vertical to horizontal solution. We’re in banks with mortgage approval from 45 days to one day. OEM with ML on cars. Telco’s predicting for customer support and customer care. Should not be writing a platform for monolithic solutions. As long as the model is written and workable upon Spark or TensorFlow, we fit in the middle.
- Drug discovery with GSK. Combination of new and old uses. New outcomes for current drugs. Reuse and recycle existing knowledge. You have the data and can analyze it in 100 different ways. Financial services around risk XVA algorithms adjusting for credits and debits. Retail with chatbot and business process. Reducing job repetitive tasks and automating those.
- The problems and use cases vary by market. 1) Our healthcare customers, which consist of physicians, hospitals, and clinical groups, aim to use AI to deliver efficiency and compliance with medical and clinical documentation and transcriptions and lighten human workload. 2) Our Automotive customers look to AI as an enabling tool for creating their own branded automotive assistants that help to perfect the driving experience for their customers. With AI, automotive assistants understand individual driver preferences to personalize, entertain, and execute complex, contextual commands. AI will also play a significant role in the not-too-distant future of autonomous vehicles. 3) Our telco customers are utilizing AI solutions to transform their businesses into that of Digital Service Providers. Communication Service Providers — both mobile and cable/wireline — are facing regulatory and market pressure on traditional revenue sources and increased competition from over-the-top providers. To keep from losing more market share or from becoming commoditized, there’s never been a more critical time for building strong relationships with subscribers through contextual and personalized interactions enabled by AI 4) Our large enterprise customers are utilizing AI to provide effortless and intuitive customer self-service across a variety of channels, including phone, Web, mobile app, messaging, and IoT devices. Each customer uses AI to solve for the specific challenges within their own industry — from banking to insurance, logistics, airlines, government, and major consumer brands.
- One of our core competencies is website classification. We analyze tens of millions of URLs and content each day to provide a classification of whether a website is about news, sports, or one of more than 500 other categories. These classifications can be used by companies that provide web content filtering (parental controls) and advertising companies to better identify potential customers. It is better to sell basketball shoes on a website about basketball than a website about sailing. As you can imagine, processing millions of URLs by humans would require an army of people to look at all of the individual websites, so automation is key. Our system is completely automated, and we only use humans to evaluate how the system is performing. We have recently released our IoT Security Platform that serves to protect consumers from being hacked through their vulnerable devices. For that we use AI to analyze the communications that occur on a network to identify the types of devices that are connected and determine if any of them are behaving abnormally or maliciously. This information can then be used to alert users and IT professionals to disconnect the device or update firmware.
- Repetitive tasks such as using data from a form and infer from previous behavior. Buy the same thing every week we pick and add exceptions. The use case isn’t getting rid of a ton of people, how do you reduce the amount of time expensive people spend doing commodity work. Give you all the analyst reports. Now we produce analyst report with the automatic decision with what you’ll do.
- Customer runs a bunch of compressors with 70 sensors and astounding data get one half a percent better of performance on a $100MM of capex. Need to safely record current failure to find anomalies you don’t need to remember, learn and able to predict what will happen. Reduce the need to throw huge storage batch at these workloads. Deploying in a large manufacturing environment for aircraft all parts with RFID tags. Data center exploding SAP and Oracle too expensive. Runs on Raspberry Pi digital twin RFID tag two lines of code tells where the tag is in the factory and can see tags come together instead of tracking nuts and bolts we are tracking a sub-assembly. Can draw map of manufacturing 100% performance improvement with 50% less to Oracle and SAP.
- 1) A U.S. defense agency 100,000 contracts per year are digitally encoded into contract management system and analyzed with NLP to examine for terms, concepts numerical patterns to determine which were most successful and beneficial to the agency. Extra efficiency reduces headcounts, manual labor, increase efficiency, and reduce errors. Millions of contracts per year you see trends. Bearing on efficiency has multiple implications. 2) Government organization in Asia to Improve public service to improve life for citizens. Monitor social media and local news for ML. Gather info from diverse sources across multiple source, multiple languages, and combined with operational data. Monitor sentiment as a gauge for success and to learn where improvements are needed. They see this as a way to engage citizens better and develop higher quality programs and eliminate those that do not bring value. Voice of the Citizen. 3) Large airport in Asia wants to retain passengers and airlines to be a modern hub watching complex sources of data to make the passenger experience better to give them competitive advantage. Door sensors, traffic in hallways, security lines, maintenance, clean restrooms, tie in with merchants to guide passengers to gates and offers from merchants to provide a better CX, improve own operations and improve business for merchants. Will differentiate from other airports in three years.
- Helping companies work in an agile way in product development and across the entire organization. Visualizing flow. We’re trying to take product management out of product management and agile out of agile and just make it intuitive. Autonomous teams managing themselves. Tool to help them do it without thinking much about what they are doing. Using simple visual flows. AI augments ability to manage stuff. If trying to predict how long a project will take and have AI to augment vision that is powerful. Working in a more organic way. 2,000 people all working together it becomes a complex system of notes and relations. It’s hard for a single person to see patterns and directions in massive body of activity. AI can help augment your vision for what’s going on. A city is even more complex system. How AI can help govern society in a good way – augment the decisions of our politicians. AI will ruin or enhance humanity. It’s up to us to learn to handle.
- Our platform offers content creators a new way to automatically distribute their content on new devices, apps and websites. With the ability to process million on minutes of audio content in real time, we provide the best tool to index and optimize the audio content for search engines. With our technology, any audio content from radio shows or podcasts, can be available to a wider audience with minimum effort from the content creator side. In addition, we extend the shelf life of the audio content and allows listeners to consume it on-demand, wherever and whenever they want. And the most important thing, we enable a new monetization option for content creators by creating a perfect ecosystem, that delivers unmatched audio experience for OEMs to deliver to their customers. Our technology is already incorporated into devices from Samsung, Bose and Harman, as well as virtual assistants - including Amazon Alexa and Google Home - in apps and in-car infotainment.
- Daunting tasks like tax compliance and cash applications can be addressed with the use of AI and ML, as they can learn from historical data to increase efficiencies and accuracy. Within finance, accounting processes can be automated using machines including everything from processing expense reports to accounts payable. Even risk assessments can be made by machines taking advantage of advanced analytics to assign risk scores.
- With email, one of the biggest challenges is it is so broadly used and it’s too big for manual work on an ongoing basis. AI makes sure we can keep up with dynamic environments at scale, and making sure the right email gets delivered, to the right person, at the right time. AI can also help in detecting security and can identify sending patterns of spammers. If one gets through, AI helps identify the issue in real-time and shut it down as quickly as possible.
- Customer service. Telecom company with call centers worldwide. Handle customers in multiple time zones. Three issues: 1) what’s my bill, why is it so high; 2) I’m moving; or, 3) I want to report an outage = 80% of the calls. Use bots to solve. Let humans handle the higher-level problems. Utility to make transformation to have bots to be call center reps. Seamless hand conversation over to the live agent if the caller is getting frustrated and then hand back to the bot for a survey. FAQ use case. Bot scrape site and build an autoboot that digested FAQs on the site. Using pure software and augment with transactional capabilities. E-commerce bot how to return an item. Budget allocation in marketing and attribution. There are tactical aspects where we are using ML to sort through data or find patterns. We find opportunities, score them, and then apply methodology which is very similar to methods deployed in hedge funds to optimally manage assets. We can maximize returns and virtually eliminate the risk for wasted spend. We don’t do marketing. We do investing. That is the number one differentiator between DemandJump and other MarTech solutions. We know the inputs to the investment
- Our product makes Wi-Fi more reliable, which in and of itself is a major problem faced by companies in almost every industry. As wireless becomes the predominant access technology, making Wi-Fi better and more business critical is a huge initiative. In addition, we add indoor location services to wireless using AI. For example, we can help hospitals detect if a dementia patient is wandering off site without a nurse and lock doors, so they don't get into harm’s way. We can help sick patients easily navigate through a hospital, or weary travelers navigate an airport or hotel. We can also help schools, companies etc. locate students and employees in an emergency so emergency responders can act quickly and appropriately.
Here's who we spoke to:
Opinions expressed by DZone contributors are their own.