As companies move forward from the radical business changes forced upon them by the COVID-19 pandemic, enterprise AI use cases are evolving: While many of the AI use cases for business will continue to focus on automating manual business processes or compliance rules, others will address the uncertainty caused by the global health crisis, said Andrew Schwarz, Ph.D., professor in the Department of Entrepreneurship & Information Systems at Louisiana State University.
New AI use cases will be seen in retail, healthcare, higher education transportation and even waste management, Schwartz said. In the near term, he said he expects AI to have its biggest impact on improving customer-facing applications, manufacturing operations and supply chain management.
Here are 8 emerging AI use cases.
1. Rethinking AI models to navigate uncertainty in pandemic era
As vast sectors of the economy are quickly reshaped by the coronavirus pandemic, AI can help companies understand how they need to pivot to stay relevant and profitable.
"In the current COVID-19 emergency, the most significant use cases are going to center on scenario planning, hypothesis testing and assumption testing, " said Arijit Sengupta, founder & CEO of Aible, an AI platform. "We'll be in uncertain economic times for the foreseeable future, so businesses need to shift from static ways of thinking about the world to approaches that focus on how we adjust to change."
AI can help to improve the ability to sense the changes and respond quickly. For example, in a sales optimization use case, a business might at first decide to make massive cuts across the board in resourcing in response to the downturn. But that could very well be the wrong decision. AI could help do complex scenario planning to figure out what the optimal resource reallocations are. These recommendations could be pushed out to the sales team, who in turn can help update the model with on-the-ground observations.
"For any use case, getting end-user feedback is crucial in uncertain times, because end users will know things that haven't surfaced in the data yet," Sengupta said.
This will require a fundamental rethink in the way enterprises approach AI. Instead of spending a lot of effort to get one predictive model perfect, enterprises will have to create a portfolio of models that covers a wide range of potential business realities.
2. Digital twins will be used across the business
Digital twins create a simulation that is updated in response to new data. The technology has typically been used for managing hard assets like heavy machinery.
Anand Rao, global AI lead at PwC, expects to see new families of digital twins applied to various aspects of the business, including supply chain and e-commerce operations. Different roles within the company can then do scenario planning, anticipate failures and tune performance using different flavors of digital twins that simulate consumers, companies and other entities.
"The need for strategic and operational modeling is especially critical in this COVID-19-affected world," he said.
Unfortunately, there are currently few people who specialize in this field, and it requires a completely different skillset from machine learning. He expects to this to drive an urgent need to upskill staff with systems thinking training.
3. Document intelligence replaces OCR
Optical character recognition merely captured paper documents into a more machine-readable format. Nigel Duffy, global AI leader at EY, expects to see a greater adoption of "document intelligence," which uses AI for reading and interpreting these documents. This capability has matured rapidly in recent years and is seeing increasing adoption.
The legislative and regulatory responses to COVID-19 are making this need more acute. Large enterprises are reviewing their contractual obligations in order to address new requirements related to things like loan forbearance, payment deferrals and early termination. This work is happening under accelerated timelines and with staff who are increasingly impacted by working from home and by illness.
"AI technologies are making this kind of rapid review more feasible and improving the quality of the results," Duffy said.
The key to success with these technologies is to view them as accelerators rather than as replacements for people. They are most effective when there is still a human in the loop reviewing and correcting information that the AI may have missed and helping train the AI for future edge cases.
4. Supply chain AI and machine learning cry out for more investment
AI and analytics enable supply chain leaders to gather and analyze voluminous amounts of data that spans customers, suppliers and market conditions. This is something humans can't do in a timely manner, if at all, said Sam Pearson, principal and U.S. supply chain analytics leader at Deloitte Consulting LLP. But automating this task require significant investment in bringing the right data and tools together.
The Deloitte 2019 Supply Chain and Digital Analytics Survey found that many companies underinvest in technologies like supply chain AI and machine learning. While 76% of respondents said developing digital and analytics capabilities was critical to their supply chain strategy, only 44% invest at least $5 million annually to develop these capabilities.
"To realize its promise, organizations must devote resources and time to ensure that AI technologies receive current and accurate data," Pearson said.
In addition, many only consider AI in helping them reduce costs, boost efficiency and meet other small-scale goals. Pearson also recommends enterprises also consider bigger picture opportunities that can include differentiating customer experience, real-time supplier collaboration, manufacturing equipment optimization and customer connectivity and engagement.
5. AI in marketing becomes critical to keeping customers
Marketing orchestration involves coordinating interactions across the customer journey. The key decisions in this relate to the messaging, execution and delivery strategies.
"AI can help make the right messaging, execution and delivery decisions to optimize for business objectives such as revenue and retention with marketing constraints such as budgets and touchpoints," said, Ramesh Hariharan, CTO at LatentView, a digital analytics consultancy.
For example, AI can help accurately predict the right channel, message and time that customers are most likely to engage with based on historical data and then drive this through the appropriate channels.
Some of the challenges in doing this well include identity resolution (crafting a unified identity of a consumer fragmented across devices and data sources); data integration; attribution (identifying the marketing message that prompted the customer to act); and integration of all that into the orchestration workflow. For example, a B2C company may have customer data residing in different silos. CPG industry players may not have a lot of addressable consumer data, while others such as retailers may have data with incomplete attributes.
"Orchestration is a complex problem, so we recommend starting small and building capabilities over time," Hariharan said.
6. RegTech will be turbo-charged by AI
AI is helping to streamline various aspect of regulatory technology, or RegTech, said PwC's Rao.
Many AI tools are being designed to improve specific back-office functions such as finding and pulling requirements from regulatory documents and comparing them to existing controls or policies. This greatly aids the compliance departments of organizations and can increase employee productivity.
Document extraction involves extracting fields (invoice numbers, pricing, etc.) from documents such as invoices and purchase orders as well as extracting segments of documents and comparing them semantically against known policies that are worded differently. The technology can then help summarize the information in documents.
"This is an active area for many financial services companies given the quantity of regulations the RegTech industry handles," Rao said. It's also valuable for any industry that works with numerous semi-standard documents such as invoices and leases that have been too complex to process with RPA, he added.
There are challenges to doing this well, so businesses need to start with a bite-size problem and then scale out.
"Attempting to develop a perfect solution from the get-go sets unreasonably high expectations and dooms a project to failure in the long run," Rao said. He recommends picking a narrow subset that makes sense, has a lot of data (like previously processed invoices or processed regulatory documents) and that has the organization's support.
7. The intelligent contact center gets smarter by necessity
The AI-enabled automation of the customer contact center is one of the AI use cases that is expected to grow, as businesses scramble to find new ways to enhance customer engagement and provide more personalized service.
"The intelligent contact center will take on increasing importance over the coming year as businesses adapt to the changing circumstances around COVID-19," said Dawn Anderson, Accenture's Customer, Sales & Service global practice leader.
"With work-from-home mandates and higher call volumes, AI is playing a more important role than ever," Anderson said. This infusion of AI into the contact center includes deploying virtual assistants to interact directly with customers and using virtual assistants to augment how contact center agents respond to inquiries. There has been an uptick in interest for COVID-19 specific AI libraries for understanding how to respond to general questions about the virus and how it affects different aspects of a business or government agency.
For example, Accenture has been setting up virtual agents for state unemployment agencies that are now fielding an unprecedented number of calls. This approach will help provide essential information to citizens, while freeing up human agents to focus on incoming claims.
8. Automating more business processes in tandem with RPA
While new AI use cases are emerging, AI is likely to have the greatest immediate impact in combination with a sister technology that has taken off in the enterprise: robotic process automation (RPA).
RPA software is used to automate a wide range of rote, rule-based digital tasks, from onboarding employees to accounts receivable to verifying e-signatures. When combined with machine learning and other AI technologies, RPA's ability to automate a variety of manual, repetitive and time-consuming tasks will mean businesses can automate more -- and broader portions of -- enterprise jobs.
"RPA is becoming more prevalent in organizations because of its ability to offload employees' low-value tasks onto a software bot so the human can focus on more thought-provoking aspects of their role," said Kapil Kalokhe, senior director of business advisory services at Saggezza, a global IT consultancy.