Is artificial intelligence finally starting to take on a recognizable shape, after its history of being sometimes overhyped, often misunderstood, and always intriguing (but in a tantalizingly futuristic sort of way)?
Perhaps so, judging by a new report from KPMG. To better understand the state of AI development, the firm this year conducted deep interviews with executives at 30 of the world’s largest companies, as well as three of largest providers of AI solutions and services. Out of this work, eight trends came into view.
Rapid shift from experimental to applied technology. “Less than three years ago, many leaders in large organizations were asking themselves if AI could support their productivity and growth objectives, and many were just beginning to pilot and test AI applications,” KPMG said.
Since then, AI has quickly moved from a “technology to watch” to a “technology to deploy,” according to the report. Now, many enterprises are looking to expand beyond pilots to drive organization-wide benefits. Half of the companies interviewed said they expect within three years to be using AI and its subset, machine learning, at scale.
Automation, AI, analytics, and low-code development platforms are converging. Deploying these technologies in an integrated manner “can support virtuous feedback loops: each supports the other, and together they go hand in hand with the scale and industrialization of AI,” the report says.
Is some organizations, this is driven from the top, KPMG observes. But it also can emerge from existing teams: “For example, many large companies have well-established global business service teams. We’ve heard that many of these teams on their own see the value of bringing AI to their existing automation activities.”
The report adds that the GBS teams are often well-positioned to harness these technologies and drive new additional value. That’s because they already have extensive capabilities around enterprise services, deep understanding of the “piping” of different functional areas, and extensive data repositories.
Enterprise demand is growing. A consistent message that KPMG heard in its interviews was that companies are aiming to vault over the functional-level bar and into other areas of the business.
The five companies in the sample with the most mature AI capabilities have an average of 375 full-time employees working on AI. On average, they’re spending an estimated $75 million annually on AI talent.
According to the interviews, high-priority areas for AI initiatives over the next two or three years include:
New organizational capabilities are critical. Organizational capital consists of human capital, values and norms, knowledge and expertise, and business processes and practices, the report notes. “AI needs all of these elements to create value for the enterprise,” it says.
In addition to creating top-down approaches, such as creating SVP-driven steering committees, some companies are pursuing more decentralized approaches, where functions or business lines are expected to lead initiatives, according to the report.
Half of the companies KPMG interviewed said their CIO has a leading role for overall AI strategy. Additionally, 40% of them said a senior line-of-business leader is taking a leading role. And almost all of them either have an establish center of excellence for AI deployment or are in the process formulating a CoE strategy.
“Generally, there are multiple technology and line-of-business stakeholders with responsibility for AI strategy and deployment,” KPMG says.
Internal governance is emerging as a key focus. “We believe large enterprise companies need to create and enforce formal governance policies, processes, and controls around AI technologies, service delivery models, and third-party providers,” KPMG says.
According to the report, governance around AI includes:
Still, only a quarter of the companies interviewed have begun to build out true enterprise-wide AI governance. “The leaders are not leaving the success of AI up to unwritten rules and are treating governance as an enterprise-wide imperative that is key for scale,” the report says.
The need to control AI. As the report notes, successful AI programs rely on the ability to control algorithms that may become more powerful and opaque over time with machine learning.
“It can become difficult to determine why an AI-driven system made a specific decision,” it says. “Where in the millions of lines of code was the decision made? The obvious risk is that continuous-learning systems can produce unintended or biased results.”
KPMG estimates that among the studied companies, only about a quarter are investing heavily in developing control frameworks and methods to drive greater trust and transparency.
Some companies are using new tooling and dashboarding to automate and visualize design, evaluation, and ongoing monitoring, according to the report. “These tools and capabilities are key to identifying and monitoring for unintended AI outcomes, misuse of data, and potential algorithm bias,” it says.
The rise of AI-as-a-Service. According to KPMG, while large companies will always need to build many foundational capabilities in-house, they will increasingly be able to also tap into an AIaaS market.
Some companies are opting for managed services for a specific AI application. And for some industries, third parties are offering end-to-end exception handling for finance departments and AI-enabled contract interpretation services for las firms.
“We believe that many enterprises will access a growing number of AI microservices, which deliver AI as a collection of small, independently deployable services around specific business requirements,” the report says.
But while these trends give companies more options for accessing AI, they do not replace a comprehensive, enterprise-wide AI strategy, according to KPMG: “Companies need to build extensive internal capabilities to even start to take advantage AIaaS options.”
AI could shift the competitive landscape. Almost all the interviewed executives see AI as playing a role in creating new winners and losers.
“A key question now is who will move up the AI activation learning curve the fastest … and at what level of investment,” the report says. “Many executives are unsure of the optimal way to divide spending between AI priorities such as organizational capital, technology, and data.”
Still, companies investing in AI report achieving an average 15% productivity improvements for the projects they’re undertaking, according to KPMG.