7 minute read 2 Jul 2018
Business people having a meeting in office

Five potential advantages of integrated AI strategy

By EY Global

Ernst & Young Global Ltd.

7 minute read 2 Jul 2018

As the opportunities — and risks — of AI emerge, defining an AI strategy is important, but maybe less important than the process itself.

In assessing where, how and when to deploy AI, and its various technologies and processes, an organization may well reveal broader and important opportunities to drive valuable top- and bottom-line improvements.

Here’s a recent example of the potential impact of the AI strategy process: a credit union organization set out to define an AI strategy to compete with larger banks that have the advantage of scale. The heads of finance, marketing, product, risk and technology came together to explore the problem; they determined that the key to competitive advantage and protecting market share lay in getting to know their customers better. With that goal in mind, they identified that machine-learning-based automation could help them achieve a better understanding of customer behavior and preferences, more cost effectively than conventional tools for doing so. The process resulted in an AI solution, but along the way, the collaboration revealed important insights that may shape more than just AI activities.

Five potential advantages of integrated AI strategy

In working through the strategy process with several EY clients, I have observed five areas which can benefit from an AI strategy that is both developed and implemented across the organization.

1. Cost reductions

For organizations seeking to be more cost-effective than their competitors, the business case for automation can be compelling. 

For example, robotic process automation (RPA) projects are designed to automate simple activities involving data processing, reducing the cost and time of human hours dedicated to these tasks and often improving accuracy. Think of RPA as a form of “non-invasive” system integration that can scrape data off screens and manipulate it without having to change the systems in which the data is stored. For example, a company may have multiple product websites where customers can place orders. If these websites do not share a backend system to aggregate customer orders, the time, cost and risk of error of manually collecting order information can be significant.

RPA software can be programmed to ”scrape” the order information from various sources and integrate it into fulfillment, inventory control or customer systems. It could even send order status e-mails to the customer. This kind of automation can deliver cost savings of between 5% and 25% depending on the business function.

For all its advantages in driving cost efficiencies, RPA does have limitations. RPA tools must be configured to read data from specific places and require defined sets of rules within which they will operate. If the data is unstructured (e.g. free text) or an organization is dealing with judgement-based activities, another type of technology called “cognitive computing” may be a better option. Cognitive computing software uses machine learning techniques to turn unstructured data into structured data — for example, by labelling text to give it structure — and to replicate the decisions of experts.

One step further, “Intelligent Automation” combines RPA and cognitive technologies to automate business processes from end to end. Some companies have already experienced meaningful cost takeouts using this powerful combination; for example, Japan’s Fukoku Mutual Life Insurance has used Intelligent Automation to replace 34 employees, representing an annual cost saving of US $1.65 million.

2. Improved decisions

AI can help to mitigate the impact of information overload and capacity constraints on the quality of decisions. For example, a mining company deployed machine learning algorithms to automatically flag policy compliance issues on thousands of free-text purchase orders. In another case, an energy company’s senior engineers trained a cognitive computing system to give advice to junior engineers, based on a knowledge base consisting of thousands of technical specifications.

3. Better customer experiences

One of the most popular uses of AI is to improve customer and employee experiences. This is especially important in a time when customers are demanding better, faster and more customized service and response options. Chatbots allow people to contact an organization or internal department without downloading another application, calling a service center, visiting a website or writing an email. They enable people to interact with services on their own terms and can be configured to aggregate multiple services into one user interface with consistent brand language and tone.

As an example, at a Canadian telecommunications company, chatbots helped to increase customer satisfaction by 65%.1

4. Improved data quality

An emerging application of AI is in improving data quality. According to a recent Data and Analytics Impact Index from Forbes and EY, most executives believe their organizations lack data of high quality that is easy to access. An insurer seeking an overall profile of its customers across business lines, for example, discovered that its biggest constraint was determining whether a person in one system was the same person in another. That’s because each system used a different customer identifier. Today’s powerful analytics solutions can replicate the cognitive processes of humans by combining AI technologies, third-party data and domain-specific knowledge models. Together these technologies can fill in missing fields, fix errors and link records across systems.

Another related application of growing importance, particularly in light of regulatory obligations such as the European Union’s new General Data Privacy Regulation (GDPR), is AI-driven improvements to the security and privacy of personally identifiable information (PII). An emerging category of software product known as “metadata discovery” employs machine learning and rule-based techniques to automatically map and track PII in data sets to manage who can access them, and even flag which regulatory obligations apply.

5. Cybersecurity

With both the sophistication and cost of cyberattacks increasing exponentially, a promising opportunity for AI is in detecting and classifying threats and attacks. 

Recently EY worked with a financial services client to deploy machine learning algorithms to detect anomalous traffic in its network, which would be difficult for humans to identify. The bank’s cyber security platform generates automated alerts for investigation by the bank’s Security Operations Center by processing billions of network flow records per minute.

Combining human and AI for better outcomes

In defining an AI strategy, organizations need to think carefully about where to use AI to its greatest impact. Sometimes automation could make customers’ experiences worse, not better; a human touchpoint can make all the difference. Consider the employee who goes above and beyond to help a customer by working around a standard process or using information that was gleaned through conversation, rather than a computer system.

AI cannot do this because it cannot generate new knowledge. Cognitive computing software needs a knowledge base and training from experts. And any business process that relies on information that people keep in their heads will be harder to automate.

Perhaps the biggest question then is whether replacing people with AI is “right.” This is especially relevant to those organizations still questioning whether replacing people with machines is consistent with their overall purpose. Of course, being able to compete with others that adopt AI more extensively is another key factor to consider.

That does not mean that AI is a zero-sum game for an organization’s workforce.

A recent survey by Richo Group in Europe2 found that 98% of employees are positive about the potential of new technology to empower them to work in smarter ways.

Almost two-thirds believe that automation would enable them to be more productive and more than half think that AI would have a positive impact on their role.

Coherent AI strategy demands a cross-functional process

As company leaders work through the issues surrounding AI, it is important to facilitate constructive conversations on the topic. This can be difficult because AI programs often involve organizational redesign. Many AI strategy projects, for example, have combined outsourcing and technology decision frameworks to facilitate alignment across multiple groups of stakeholders. These activities involved consulting with different groups to:

  • Define the leading theme for AI in the organization (e.g., cost takeout)
  • Identify the target business functions
  • Map and benchmark the performance of business processes
  • Define treatment strategies (e.g., full, partial or no automation)
  • Evaluate and recommend coherent data and technology investments

The team, whether internal or part of a consulting partner, will need to bring these frameworks together with objectivity and practical experience in delivering AI programs.

As companies grapple to map their path forward in the fast-evolving AI landscape, they will need to bring stakeholders together to connect purpose and business objectives to the technologies and processes best suited to achieve them. Those organizations able to align these factors are more likely to experience the benefits not only of coherent AI strategy but also of the insights revealed within the strategic process along the way. These insights are likely to help them to make smarter decisions around developing specific AI solutions, and emerge stronger to thrive in today’s Transformation Age.


If a company is a coalition of people brought together to do something better, cheaper and/or faster, what happens when Artificial Intelligence (AI) can replicate what employees do? What will this mean for an organization’s future?

About this article

By EY Global

Ernst & Young Global Ltd.