Disrupting the Status Quo With AI-Enabled Content Services

It should come as no surprise to you that digital transformation (DX) and artificial intelligence (AI) are two of the hottest topics in strategy planning across almost every industry. There is the hope that DX and AI will drive a new future of work with a promise to improve business operations, employee empowerment and customer engagement. It is also the fear of falling behind their competition which has enterprises betting on AI to bring the advantages they need to survive and thrive. According to IDC, executives are investing in AI systems to the tune of $24 billion in 2018 with a significant continued annual growth of 37.3% to spend more than $77 billion by 2022. While AI is important to a DX strategy, many companies are looking for a few short-term wins that will extract valuable, accurate and timely insights from the data it has available today and address the flood of data knocking on its door. I believe that to achieve this goal, organizations must first consider the significant shift in the data that fuels AI systems and seek the opportunity to recalibrate archaic processes to better leverage a future of machine-led activities.

It’s all about the data

As our world becomes increasingly digitized and connected, the amount of data generated is often too much to handle. The irony is that it is this same data that fuels the training of advanced machine learning (ML) and delivers to us the insight and recommendations we seek. In the past, all the data in the enterprise was kept in well-defined files and databases. It served the needs of the applications for which it was designed. It was slow to change in format and structure. The digital transformation era has brought about a new class of unformatted data that arrives from Slack queries or drone videos, real-time sensors, mobile data, and the Internet of Things. The volume is enormous — IDC estimates the global data sphere will grow to 163 zettabytes (i.e., a trillion gigabytes) by 2025 — 10 times the 16.1 ZB of data generated in 2016. This enormous shift offers significant potential for AI to assist but comes with a set of challenges that have never been seen before.

We’ve been talking about deriving structured data from unstructured content for more than 20 years. And, Alan Turing gave us the foundation for AI almost 70 years ago. So, what has changed to make the idea of an enriched knowledge base a reality today?  I think it is a continuum that benefits from years of advanced computing coupled with the virtually unlimited capacity available in the cloud that makes storing and processing this vast amount of data possible. Agile content services that take advantage of a cloud microservices architecture provides opportunities for developers to create reusable micro-apps that connect, process and analyze content from any source to any edge device. Cloud content services that are ready-to-use, like Microsoft Azure Computer Vision or Amazon Rekognition, can be connected to your library of data and immediately start analyzing its contents — without you needing to obtain a degree in data science or machine algorithms.

But those applications work on a defined set of digital media. To apply ML to the vast amounts of business documents, files, and other unstructured content, you must first establish a taxonomy and tag the data that is to be used to train the AI system. Then it is just a matter of continuous training and testing until you are satisfied with the machine’s level of accuracy (confidence level). There are many approaches that can assist you to classify, aggregate, standardize, and label content for AI learning. For example, some offerings include OCR and intelligent content analytics that can discover insights from contracts, and others leverage capture solutions and robotic process automation (RPA) for document capture of invoices or mortgage applications. In some cases, organizations are implementing a componentized content management system or storage system with structured data schemas for AI-ready content and processing.

Aim for disruption with new AI initiatives

Now that we have the content AI-ready, where do we point our magic wand and innovate our business to create sustainable competitive advantage? One area of DX for which AI can have a positive impact is in content-centric processes that are easily automated or need modernization across industries that have an abundance of multi-structured data to handle. IDC found that banking, retail, manufacturing and healthcare have been the early adopters of AI systems, accounting for more than 20% of the spend to create business applications with intelligent capabilities. They have applied AI to content-centric processes like new account opening, negotiated loan agreements and employee onboarding. 

As you look across the organization for the right scenario to apply AI, consider how AI can help advance your company’s strategic business goals whether it is making human capital more efficient, operations more scalable, or driving higher customer lifetime value. Identify business objectives that AI can help achieve faster and with better accuracy such as adopting an agile methodology, improved product lifecycle maintenance, creating smarter cities and facilities, to achieving a 360-degree view of the customer. Be bold and use AI projects to modernize technology to the cloud, revise company policies and rearchitect archaic processes that were invented when paper was the only medium of record. Whether it is more practical applications like fraud detection or predictive maintenance of machinery, or it is more complex and emotional AI understanding in customer service chatbots and health diagnostics, AI should align with corporate KPIs.

As companies move from experimentation to implementation to deployment, they encounter roadblocks to AI adoption. One of the biggest hurdles is trust in the recommendations produced by the AI system, trust in the source data that fuels it, and trust that the AI system is unbiased, and protects the governance and data privacy regulations in the region. Building trust means feeding the machine system lots of data and providing transparency to the process. As a consumer, I easily trust recommendations from Netflix or Amazon, yet I struggle to trust a chatbot on a website to answer my question. Instead, I ask for an “agent” and hope they can help. Why do we trust a person with limited visibility to our situation over a machine? Because they offer the potential of empathy and relatability? Can a machine find similar historical situations to mine faster than an agent? Yes. Perform a more widespread data search and collection? Yes. Relate to our emotional status with natural language? Getting there. By 2022, IDC predicts affective computing (emotion AI) will include vision and voice technologies and see an increase of 25% in real-world applications. Maybe we are closer than we think.

As I look across the business landscape, I see a myriad of business processes that need a kick in the pants. AI and ML are not a silver bullet to an organization mired in data chaos and haphazard policies. But AI can help learn from past experiences and find solutions for new ones if you give it a chance. For customer service, this means moving away from rules-based rigid systems and embracing a more intelligent understanding of the problems the customer is facing and adopting the medium for which they want to communicate. Move away from paper or even structured web forms to a more intelligent guided interview that interacts with the end user on a mobile device and intelligently augments the discrete data with outside sources that enrich the experience. Once the AI system is trained, it has an ability to work with incomplete and ambiguous information to present a solution or recommendation with a level of confidence the user can choose to accept or refine with further interaction. AI-enabled content services can reach beyond customer service to dozens of use cases in acquisition, onboarding and operations with industry-specific flavors. As we begin to trust the AI system to perform the task at 99.999% confidence, like OCR of postal mail or mobile check cashing, we can move from human-led to machine-led tasks. It will take time to get there, but the easy wins are within reach if you are willing to disrupt the status quo.

Marci Maddox is Research Manager for IDC's Enterprise Content Strategies program, responsible for content workflow and content technologies research. Marci's core research coverage includes the evolution of enterprise content, customer communications, content sharing and collaboration, e-signature, forms and capture solutions. Leveraging 15 years in content and process applications, Marci also analyzes the impact that new technology entrants like AI and RPA have on the way organizations create, process, and deliver content to various destinations.