DZone Research: How Organizations Benefit From AI
DZone Research: How Organizations Benefit From AI
Using automation to improve the efficiency of the operations and empowering employees to focus on more meaningful and less repetitive tasks.
Join the DZone community and get the full member experience.Join For Free
Bias comes in a variety of forms, all of them potentially damaging to the efficacy of your ML algorithm. Read how Alegion's Chief Data Scientist discusses the source of most headlines about AI failures here.
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, "How can companies benefit from A.I.?" Here's what they told us:
Automate to Improve Efficiency
- There was a time when AI was really bad at routine tasks such as speech recognition, object classification, and natural language processing. However, now the only limitations to automating a process are: 1) Is the process stable? Is it unlikely there will be unexpected changes that require adaptation? 2) Do you have enough examples/scenarios to train an AI algorithm? If the answer to the above questions is "yes," then that process is a prime candidate for automation — even if the process is complex and requires sophisticated abstraction. Additionally, even the first question is becoming less of an issue as systems learn how to become more flexible.
- The non-technical business point of view is the return on investment to enhance an existing product they already have like a recommendation engine for a financial services company. Another thing is to improve the efficacy of their operations. Banking customers use AI to reduce the turnaround time of loan approval for their customers. Others are just providing services they were not able to provide before like fraud detection, customer service, improvements in reliability, scale, and things like that. The company wants ROI, and a robust AI/ML practice helps this. Understand the investments being made while seeing how the algorithms help improve efficiencies, drive revenue, or reduce costs. Also, manage risks, avoid catastrophic failures and exposure to legal liabilities, stay compliant with the rules of the industry, and ensure employees are working efficiently and staying compliant. Risk management and compliance feed into this.
- Make people more efficient and manage time and projects efficiently.
- AI, ML, and analytics make a business process smarter, but it is how we connect the function and automate the process from the get-go to make it impactful. One of the largest benefits of A.I. implementation in the enterprise is the ability to handle routine processes with a time and cost efficient and accurate approach.
- Enterprises that implement AIOps see a number of major benefits — they can:
- Increase end-to-end business application assurance and uptime
- Manage an integrated set of business and operational metrics
- Predict and prevent outages
- Dramatically reduce Mean Time to Detect and Mean Time to Repair
- Lower the number of IT FTEs dedicated to troubleshooting
- Optimize IT and reduce IT costs
- Gain visibility into the hybrid IT environment
- Accelerate migration to the hybrid cloud
- Expedite the adoption of hyper-convergence and microservices architecture
- Reduce risk in consolidating and migrating data centers
- Free up resources to enable IT operations to become a proactive source of innovation
- Automate and reduce the cost of audits and compliance
- Simplify IT processes
- Break down silos across their IT teams
- Enable less experienced staff to become more productive, faster
- IT has never been more important, but IT must do more with less. The old model of IT Services Management (ITSM) is broken — it doesn’t scale, doesn’t provide visibility into user experiences, and relies on archaic manual management schemes. AI is the fundamental approach to meet the future requirements for IT as it brings much-needed automation, insight, and scale. But it must be pervasive, and it cannot operate in a vacuum.
- The right approach is to figure business use case — predictive maintenance, fraud detection, and simulations for drug discovery. AI is very broad and a shorthand for smart code. It is a problem-driven approach. Depending on the use case and vertical, it might be different. In financial services, a lot of risk management is driven to the next level. GSK working on drug discovery needs to innovate faster because simulations and new outcomes are more important.
- The champions of the IT department have been struggling through some less-than-ideal work environments. You’ve heard the cliché: an IT professional seated in a dark corner of the office, staring at a screen all day while wrangling thousands of incident tickets. That doesn’t have to be the reality. We change that by using automation so enterprises can evolve the next generation of IT departments. Now, many people might shake their heads at the idea of using AI in IT because they think it will eliminate people from the workforce, but this is not the case at all. We are not seeking to eliminate roles, we’re in the business of elevating them. This means IT professionals are spending less time focused on reducing the noise of thousands of errors and more time on what truly matters. For example, a digital insurer with over 30 million customers, saw a 99 percent event reduction and 10X increase in their IT team’s productivity. That’s less time fixing problems and more time building innovative features and delivering world-class services.
- Working with sensor data for getting machines to report on their condition is called predictive maintenance. Automotive suppliers in which vehicles work autonomously and all components need to figure out when they need servicing. CPG is smarter about own condition reporting to user or self-correcting.
- Mechanically, how they’re doing data is becoming much more available. Insurance companies charging by the mile versus the year based on how far you drive, acceleration, and brakes. If we know this, we can give you a much more tailored policy. It’s a fine decision to offer people. Car manufacturers are getting more information. You can get a monthly health status report on the car: here’s how the car is doing. The car company has to handle much more data. The second how is: how do you identify these things and find the one that really matters? Do a 360-degree review of the business. What’s happening, what critical steps are taking place, and where are they taking shortcuts because they do not have time to make informed decisions? Everyone knows this is just too expensive to even try. How many decisions could be made if you had the ability to use the data to make decisions without the fatigue and cost? These are ML candidates. What could my business do if I could make these decisions considering all of the data at a very low cost? That are worth milli-pennies or less, so you can make a lot of them and save a lot of money — a couple of parts out of $10,000 fraud is in the pennies. You cannot afford a human to take a bit of thought for pennies where machines can. Those are candidates for ML. A new car will hit the brakes if we’re about to hit something. That cuts the rates of serious accidents by more than half.
- When we start to look at EIM through analytics and AI lenses, we see business use cases and how AI provides a distinct advantage. We power essential business processes by being able to trigger automated responses and delivering insights on massive amounts of complex data. We use an amalgamation of different applications of AI. Experience management automates CX to maximize lifetime value: predictive maintenance, hyper-optimized inventory management, combining AI with business process automation with low code app development, RPA, and business process automation. It's able to tie together disparate data to have a broader understanding of what customers and suppliers are doing and harness knowledge to predict or anticipate what they will do. Mitigate risk, fix a machine before it breaks down, and capture a customer before they abandon.
- The benefits from AI are as numerous and varied as the targeted use cases, with more emerging all the time. Companies benefit from increased operational efficiencies and improved customer responsiveness. Many companies use AI to extend the capabilities of their customer service organizations through automated bots that respond to incoming inquiries or systems that route those inquiries with greater accuracy. This is especially important for service organizations in retail, healthcare, banking, and others. Manufacturing companies benefit from streamlined production operations, and IT leverages AI to improve network security and automate resource provisioning. It’s worth mentioning that automation is a key benefit of AI, but automation by itself is not the same as AI.
- Companies can benefit from AI both externally, by delivering better experiences for their customers, and internally, by improving workplace productivity, efficiency, and work conditions.
- AI can drastically help complex systems adapt, learn, and perform even as the broader dynamics of human communication continuously change. When done correctly, it can strengthen customer communications, increase security, deliver alerts and insights to customers, and optimize programmatic and campaign sending based on engagement.
- Most prevalent in financial services deep into the journey have people, ideas, and datasets leveraged well down the journey of creation. The forward-looking are now looking for best practices on the operations side of the house. There is a need to build a management solution for this. Folks are looking for a way to do this.
- Data has a transformative ability with efficiency and revenue how to find value in vast amounts of data. The city of Palo Alto has 200 traffic lights and 4 terabytes per day. This is not relevant to the real problem we solve, consume, and predict, which every intersection will do to route cars and never stop at a red light. AI/ML hidden and used spare compute. Most of the pipelines in the cloud are limited by database-driven architecture and REST. Digital twins live in memory at the edge and learn really fast. Innovation can self-train an intersection with 50 sensors and throw in a DNN as a prediction of what you should see. Compare prediction with reality and help make the model better. Billions of learning cycles in one intersection. A city with a few hundred intersections has a few hundred models. Solving the Palo Alto problem is over $5,000 per month on AWS versus on the edge with existing CPUs is $40. We use an API in Azure to get those insights. Sell access to API who want to make use of the insights. Self-training and the ability to catch own errors on the fly. The crazy thing that comes out of this same software system consumes data that recharges batteries and can predict failure 48 hours in advance. It doesn’t know what a battery is. Present data on the fly in a time series and learn from that. Find insights at great value for a very low cost. You can absolutely nail a lot of cities on Nvidia boards. Idle capacity at the edge is huge. The edge is getting smart fast a lot of CPUs that can drive insights on the fly.
- From an audio perspective, AI has the ability to go and deep dive into a specific story and get different angles from the same trending topics. For example, our voice recognition platform is able to filter a trending topic like Trump and North Korea, identify the relevant keywords, and serve multiple audio clips that talk about a story related to the initial keyword even if the clips do not mention Trump or North Korea. The AI technology is able to understand context, classify, and segment to deliver personalized clips.
- Most customers need to understand the use cases that will benefit from the use of DL for customers and internally. What can and should AI do for them? We’re not going to go back to do a pre-built model. The customer has the data to tweak a model and use it. Everything in AI comes down to a model that’s unique for you. Use our technology and their data to build and deploy models for their use. Something unique and valuable to the end user. DL different from ML all classification problems. Classify into a set of known intents. Know if a customer is happy or upset. In most enterprises, you must learn what these things mean. Internal transformation to improve every process and internal user experience. AI is not the thing you do it’s what you do to make what you do better. COE-driven by Chief Data Officer and data scientists who use TensorFlow. It’s not mainstream yet.
- Anytime an organization shifts from heuristics to rigorous statistics, there are always gains to be had. Guessing introduces many structural inefficiencies into the process. Using data and rigorous statistics (not necessarily AI) is really the most critical step.
Here's who we spoke to:
Opinions expressed by DZone contributors are their own.