Executive Insights on the Current and Future State of Artificial Intelligence

This article is featured in the new DZone Guide to Artificial Intelligence: Automating Decision-Making. Get your free copy for more insightful articles, industry statistics, and more!

To gather insights on the state of artificial intelligence (AI), and all of its subsegments — machine learning (ML), natural language processing (NLP), deep learning (DL), robotic process automation (RPA), regression, etc., we talked to 21 executives who are implementing AI in their own organization and helping others understand how AI can help their business. Specifically, we spoke to:

• Assaf Gad, Vice President and Strategic Partnerships, Audioburst
• Tyler Foxworthy, Chief Scientist, DemandJump
• Patric Palm, CEO, Favro
• Sameer Padhye, CEO, FixStream
• Matthew Tillman, CEO, Haven
• Dipti Borkar, V.P. Product Marketing, Kinetica
• Ted Dunning, Chief Application Architect, MapR
• Jeff Aaron, VP Marketing, Mist Systems
• Ebrahim Safavi, Data Scientist, Mist Systems
• Dominic Wellington, Global IT Evangelist at Moogsoft
• Dr. Nils Lenke, Director, Corporate Research, Nuance Communications
• Mark Gamble, Senior Director of Product Marketing, OpenText
• Sri Ramanathan, Group Vice President of Mobile, Oracle
• Sivan Metzger, CEO and Co-founder, ParallelM
• Nisha Talagala, CTO and Co-founder, ParallelM
• Stuart Feffer, Co-founder and CEO, Reality AI
• Sven Denecken, SVP Head of Product Management, SAP S/4 Hana Cloud
• Steve Sloan, Chief Product Officer, SendGrid
• Simon Crosby, CTO, Swim
• Liran Zvibel, CEO and Co-founder, WekaIO
• Daniel DeMillard, A.I. Architect, zvelo

Key Findings

1. The keys to a successful AI strategy are to have a well-defined business problem to solve and to have a sound data management program in place to ensure you have the data to solve the business problem. Truly understand what you are trying to do. Think about the use case. Understand the problem you are trying to solve and the benefit of doing so. The proper definition of the problem will help you determine the optimal type of AI you should use to provide the solution or answer the question. It all starts with data. ML is an exercise in overcoming variation in data. The amount of data needed depends on the amount of variation. Having a huge number of sources is critical for AI implementation. More accurate and properly formatted data leads to more accurate models.

2. The overarching way companies benefit from AI is by automating business processes and operations. Automation saves time, improves accuracy, and frees up workers to use their brains to solve higher-level problems. This improves the efficacy of your operations and provides a significant return on investment. Data has a transformative ability on efficiency and revenue by finding value in vast amounts of data that humans can not possibly see on the magnitude of a 99 percent reduction in events and a 10X improvement in IT productivity. How many smart decisions could be made if you had the ability to use data to make decisions without fatigue or cost? These are ML candidates. You cannot afford for a human to make decisions that save pennies, but a machine can and the impact on the bottom line can be tremendous.

3. The biggest changes in AI in the past year are that companies are beginning to see real business value from their data and they are able to do so because GPUs have become affordable. The past year has seen many breakthroughs in AI, ML, natural language processing (NLP), and deep learning. Companies are seeing real benefits in how AI can drastically help complex systems adapt, learn, and perform even as the broader dynamics of human communication changes. When done correctly, it can strengthen customer communications, improve security, provide delivery alerts, yield insights into customers, and optimize programmatic advertising and email based on engagement. There's hardware available at a lower cost and there's data available. You're able to buy GPUs for a few pennies on the cloud, and this enables machines to make decisions by themselves based on new sources of signals and more data. There are new ways to handle vast amounts of disparate data at great speed.

4. The technical solutions being used most frequently in AI are TensorFlow, Python, R, and Spark. A data platform is used to bring data in from across the organization and no one is using a single tool. Everyone tries a bunch of tools and then uses the ones that help them fulfill their particular need. TensorFlow has created a big democratization of the ability to use AI.

5. There are a wide number of real-world problems being solved with AI with those in automotive, financial services, and healthcare leading the way. The most frequently mentioned applications were security and compliance. Training models for autonomous vehicles, medical imaging, homeland security, genomic analysis, risk, and fraud detection. A premier autonomous vehicle company is using machine learning to train and refine algorithms for use in self-driving cars. Genomic analysis is being used as an early detection system for bioterrorism. In another case, image data from pathological samples are being scanned to detect early stages of cancer. Several clients in financial services are using applications for fraud detection and loan approval helping to institute an AI practice in the organization so people can focus on operational ML as well as compliance and governance. Some providers have helped banking reduce mortgage approval from 45 days to one day. Replacing repetitive tasks with automation is reducing the amount of time expensive of people doing commodity work.

6. The most common issues preventing companies from realizing the benefits of AI are data, failure to define the problem you are trying to solve, trying to apply AI where it is not the right tool, silos, and insufficient skillsets. Not having a handle on data across silos is a huge problem. When your datahouse isn't in order, you cannot succeed. There is so much data in so many formats it's overwhelming and complex. It becomes a barrier unless you have the right AI platform that can handle the diversity and amount of data. It's important that an organization approaches AI from the starting point of, "What problem do we need to solve?" rather than, "Let's do something with AI." A lot of people attempt to use AI/ML where it's not needed. To get value from ML, you really need to understand what's going on and the problem you need to solve.There is a skill set problem: skills and talent are in short supply. Agencies are no better, so we end up with a blind leading the blind situation. We need more effort around executive education. Arm execs with the right questions to ask. AI solutions providers need to be able to help with execution. Subject matter experts understand datasets and the problem, but don't know how to manage and process data.

7. The biggest opportunities for AI are reduction of operating costs, automation of repeatable processes, and around call centers and improving customer service. AI is going to become ubiquitous as there is opportunity in every industry. AI is already driving business model transformation. Manufacturing and supply chain are huge cost centers where costs will be able to be reduced significantly. Automation is big around cars and trucks. Auto manufacturers don't need to worry about selling more cars, but who they sell to will change. Automation is increasing adoption of ERP workloads in the public cloud leading to faster consumption and reconciliation of data. Humans are being empowered to be more effective in their jobs. Chatbots are revolutionizing customer support. Digital communications have the potential to feel like actual human conversations and as analytics capabilities become more robust, they will be able to deliver valuable insights into consumer preferences, behavior, sentiment, and intent that can be used to deliver personalized experiences throughout the customer lifecycle.

8. The biggest concerns regarding AI today are around hype, ethics, and security. There's a fair amount of hype and excitement. Buzzword fatigue is a concern along with over-inflated expectations. Early experiences may slow down the positive and productive progress we need to make AI seamless.We can build a place with huge inequalities or we can build a place where AI/ML capabilities make the world a better place. That's a choice we have to make by deciding the types of businesses and business practices we want to support. There's a lot of talk around "AI for good," but simply wishing for the best as we develop new technology is not enough. Society needs to rethink how the future will look so we can be prepared for the changes to public safety and the workforce. We need to ensure AI is used to benefit every member of society rather than just the top corporations.There is an arms race going on in security. Hackers are becoming more sophisticated and finding increasingly clever ways to bypass safeguards. AI is essential to solving that problem. There's a lot of potential for these systems to detect fraudulent or suspicious activity based on the ability to evaluate massive amounts of data in real time.

9. To be proficient on AI projects, developers need to understand the fundamentals of data science and be proficient in Python. A developer needs to know the strengths and weaknesses of AI projects, but they do not need to know all of the math and statistics that go into the background. The difference between a developer and a data scientist will grow smaller as there is a democratization of tools and learning making it easier for people to do this work. In the meantime, developers and data scientists need to spend time talking and understanding the problems they are trying to solve and the algorithms that will help the applications solve the problem.

This article is featured in the new DZone Guide to Artificial Intelligence: Automating Decision-Making. Get your free copy for more insightful articles, industry statistics, and more!