Scaling AI proving to be tough, but more organizations moving beyond pilots: Study

Artificial Intelligence offers the chance to change the way organizations do business, from delivering operational efficiency to massively increasing the understanding of customers and markets. As a result, organizations across sectors are engaged in an AI arms race, with investments pouring into AI start-ups across industries. Today, over half (53%) of organizations have now moved beyond AI pilots, a marked increase from 36% three years back (see Figure 1), according to Capgemini Research Institute?s report, The AI Powered Enterprise: Unlocking the potential of AI at scale.

Figure 1: Percentage of organizations that moved beyond AI pilots/pocs increased to 53%

Source: Capgemini Research Institute, State of AI survey, March?April 2020, N=954 organizations implementing AI; State of AI survey,

June 2017, N=993 organizations implementing AI.

However, while progress has been made in moving beyond pilots to deployment, scaling those AI deployments across the enterprise has proven to be tough. Only 13% have rolled out multiple AI applications across numerous teams. Some key challenges faced by organizations in achieving scaled deployment are:

  1. Lack of mid- to senior-level AI talent (selected by 70% of respondents)
  2. Lack of change management processes (65%)
  3. Lack of strong governance models for achieving scale (63%)

Furthermore, 78% of AI-at-scale leaders continue to progress on their AI initiatives at the same pace as before COVID-19, while another 21% have increased the pace of their deployment. This is in stark contrast to the ?struggling organizations?: 43% of whom have pulled their investments while another 16% have suspended all AI initiatives due to high business uncertainties related to COVID-19 (see Figure 2).

Figure 2: Almost all AI-at-scale leaders are progressing as planned or even faster on their AI deployments

Source: Capgemini Research Institute, State of AI survey, March?April 2020, N=120 AI-at-scale leaders, N=690 struggling organizations.

The report reveals that the successful implementation of AI at scale delivers tangible benefits on the top line, with 79% of AI-at-scale leaders seeing more than a 25% increase in sales of traditional products and services. In addition, 62% of the AI-at-scale leaders saw at least a 25% decrease in the number of customer complaints, and 71% witnessed at least a 25% reduction in security threats.

Sector view: Life sciences and Retail continue to lead the way in AI adoption; Financial Services and Utilities lag

In terms of the top five sectors leading AI adoption, life sciences and retail organizations are far ahead of others making up 27% and 21% of the AI-at-scale leaders respectively; followed by automotive and consumer products with 17% each, and then telecommunications (14%) (see Figure 3).

Figure 3: Life sciences and retail lead the scaling race

Source: Capgemini Research Institute, State of AI survey, March?April 2020, N=120 AI-at-scale leaders, N=954 organizations

implementing AI.

The figure above indicates only 38% of life sciences organizations have either suspended or pulled investments because of COVID-19, compared to the insurance (66%), banking (64%) and utilities (64%) sectors. This reflects the importance of e-Health in today?s context, where virtual assistants, contact tracing apps and chatbots are proliferating as organizations, like the World Health Organization, launch AI-based tools to gather as well as provide information during the ongoing pandemic.

Trusted, quality data is essential for scaling AI

AI-at-scale leaders rank ?improving data quality? as the number one approach that helps them generate more benefits from their AI systems. A strong data governance ensures that the AI teams have the right quality of data and improves the trust placed in data among the executives. Establishing the required technology platforms, such as a hybrid cloud architecture and democratizing the data access, serve as core building blocks for scaling AI.

Hiring dedicated AI leads is key to supporting the AI goals of an organization

Capgemini?s research shows that 70% of organizations find a lack of mid to senior-level talent a major challenge for scaling AI. Over half of AI-at-scale leaders (58%) have appointed an AI Head/Lead/Chief AI officer who can provide development teams with a vision, establish guidelines around prioritization of use cases, ethics and security, while harmonizing the use of platforms and tools for AI development. Organizations also need to focus on a wide range of skillsets for scaling AI applications, beyond pure AI technical skills, to include business analysts and change management specialists. However, there is currently a significant gap between demand and supply in important disciplines like machine learning or data visualization. Training and upskilling are therefore critical to address these gaps and ensure that these skillsets can be kept in-house.

Ethical AI interactions play a vital role in creating consumer satisfaction and trust

Regardless of the strong consumer and regulatory focus on ethical AI, Capgemini found that many organizations are not actively addressing issues like the need to have an empowered ethics team. The report found that less than one-third of struggling organizations (29% compared with 90% of AI-at-scale leaders) agree they have a detailed knowledge of how and why their AI systems produce the output they do (see Figure 4). This is important for business executives to be able to trust organizational AI systems. At the same time, it is impossible to establish consumer trust if the customer-facing employees lack trust in the models or data organizations use.

Figure 4: Fewer than one in two organizations have a strong focus on ethics

Source: Capgemini Research Institute, State of AI survey, March?April 2020, N=120 AI-at-scale leaders, N=690 struggling organizations, N=954 organizations implementing AI.

?In light of the recent COVID-19 crisis, while organizations are looking at data and AI to bring resilience to their operations, there is an even stronger need for connections between tactical and strategic business objectives and implementation in order to achieve scale,? says Anne-Laure Thieullent, Artificial Intelligence and Analytics Group Offer Leader at Capgemini. ?Our research highlights that the most successful organizations combine efforts to rationalize and modernize their data landscape and data governance processes, focus on bringing new agile tools from partners ecosystems as well as approaches like DataOps and MLOps (machine learning ops) to develop and deploy AI solutions, nurture teams from diverse backgrounds, and set up balanced operating models.?

As per the report, organizations need to focus on four principles to successfully scale AI:

1. Empower: Build strong foundations providing easy access to trusted, quality data through the right data and AI platforms and tools along with agile practices

2. Operationalize: Deploy AI through the right operating model, prioritize initiatives and ensure well-balanced governance while embedding ethics. A large majority of organizations that are successful in wide-scale deployment have one thing in common ? a strong AI governance and change management. Figure 5 shows that there are five areas that are key to embedding AI in operations.

Figure 5: Operationalize AI: Key Initiatives

Source: Capgemini Research Institute Analysis

3. Nurture: Build diverse talent and collaboration with ecosystems and partners

4. Monitor and amplify: Continuously monitor model accuracy and performance to deliver and amplify business outcomes

Share on

Leave a Reply

Your email address will not be published. Required fields are marked *