• Digital Economy Dispatches
  • Posts
  • Digital Economy Dispatch #192 -- The Battle for Enterprise AI: From Experimentation to Enterprise-Wide Transformation

Digital Economy Dispatch #192 -- The Battle for Enterprise AI: From Experimentation to Enterprise-Wide Transformation

Digital Economy Dispatch #192 -- The Battle for Enterprise AI: From Experimentation to Enterprise-Wide Transformation
14th July 2024

In recent months, discussions about AI have taken an interesting turn. Initial excitement about how quickly AI will drive rapid widescale disruption across business and society has dissipated. Despite significant hype, concerns are now being raised that AI has yet to deliver the anticipated change in business practices and economic benefits that many had expected. While there are notable successes in specific situations like customer service and marketing, widespread AI adoption seems some way off, hindered by integration, security, and privacy challenges. Most organizations are still trapped into experimenting with AI on a limited basis as they ask questions about where AI fits and how it will deliver value to them.

To fully unlock AI's potential, businesses must overcome obstacles to scaling AI from pilot projects to enterprise-wide implementations. To do so, critical questions must be addressed to effectively integrate AI into existing operations and sustain initial successes to achieve substantial, measurable returns.

Reviewing the GenAI Experience

A very useful insight into the issues being faced in scaling AI can be obtained by looking at the current situation with adoption of generative AI tools such as ChatGPT. As observed recently by Ben Evans, widely quoted figures about the millions of people who have experimented with ChatGPT mask a deeper set of issues. Several recent surveys report that the majority did not become regular users, indicating a disparity between initial curiosity and ongoing utility.

Ben believes that there is a variety of reasons for this limited engagement. In particular, he comments that many users viewed ChatGPT as an interesting novelty rather than a necessary tool. Moreover, the best versions of these AI models are often behind paywalls, limiting accessibility. Additionally, steps to change user habits and integrate new types of AI tools into daily life are still in their infancy. Large Language Models (LLMs) like ChatGPT require significant development to transform from a basic technology into a product with practical applications. This is evident in enterprise settings where, despite substantial interest and experimentation, actual deployment remains cautious and varies widely by use case. While useful in fields like coding and marketing, emerging data shows that LLMs have yet to find similar traction in areas like law, HR, education, and government services.

The Shifting Sands of AI

There is no doubt that Ben Evans makes some important points. However, I would go further. I would assert that much of the limited impact of AI can be attributed to a shift in the evolution of enterprise AI as organizations struggle to get to grips with a new phase of enterprise-wide AI deployment.

Initially, AI efforts focused on small-scale, isolated use cases driven by technologists and data scientists. However, as organizations recognize AI's transformative potential, the emphasis is shifting. Tactically, organizations are beginning to identify a limited set of situations in which AI can offer immediate impact and focusing their efforts. Strategically, they are investigating where and how AI can be used in challenging enterprise scenarios that require deep domain expertise, broad business support, and significant enterprise-wide integration into diverse legacy systems supported by common data infrastructure services and standards. However, attempting to meet these needs while sustaining day-to-day operational activities is in many cases overwhelming IT and digital services groups facing mounting work backlogs and staff shortages.

Furthermore, this AI-driven transition is now not led by technical teams seeking emerging technology insights, but spearheaded by business leaders seeking a strategic AI advantage. As business leaders begin to get their heads around the business implications of AI, they are ramping up their demands to see meaningful returns from significant AI investments already made. Future funding depends on demonstrating that AI is having impact where it matters: On the bottom line.

As a consequence, the AI landscape is quickly maturing beyond the experimental phase, where data scientists explored AI's possibilities in isolated organizational pockets. The frontier now lies in integrating AI into core business functions to enhance strategic advantage and operational efficiency. This transition from technology-led, small-scale AI to business-led, large-scale AI is a pivotal moment, defining organizations as either AI leaders or laggards.

The Two Phases of AI Adoption

To understand the challenges ahead, it's essential to differentiate between the two distinct phases of AI adoption. Phase 1 we can refer to as the experimental phase, while phase 2 is focused on enterprise-wide deployment. Having success in the technology investigations of the experimental phase is no guarantee of success in the more business-focused second phase. Their distinct characteristics highlights the gap between these approaches and the challenges facing organizations today as they try to shift gears in the enterprise AI adoption strategies. 

Characteristic

Phase 1: Experimental

Phase 2: Enterprise-wide

Leadership

Technologists and data scientists

Business Leaders, C-suite

Scope

Small-scale, isolated use cases

Large-scale, integrated systems

Data Architecture

Limited data access and quality, project-specific datasets

Robust, centralized data architecture and standards

Focus

Technology-centric, proof of concept, feasibility experiments

Business-centric, strategic implementation, competitive advantage

Success Criteria

Technical performance

Business outcomes and ROI

Funding

Project-based, incremental

Strategic, continuous

Integration

Limited, isolated projects

Deep integration across business functions and systems

Skillset Requirements

Specialized technical skills

Cross-functional collaboration skills

Maturity

Emerging tech, R&D

Established tech, IT service delivery

 The first phase was characterized by experimentation and learning. Data scientists and technologists explored AI's capabilities in specific, isolated areas. While valuable insights were gained, the impact on the overall business was limited.

Now, organizations are entering the second phase, where the goal is to scale AI across the enterprise. This requires a fundamental shift in mindset, strategy, and capabilities. AI is no longer a technology project; it's a business imperative.

The Challenges of Transition

Yet, the transition from Phase 1 to Phase 2 is fraught with challenges. These must be addressed as organizations move from considering their AI adoption efforts as projects aimed at delivering technical insight to well-governed programmes driving business value. Here are some of the most important priorities in making that shift:

  • Data Challenges: Scaling AI requires a robust, accessible, and high-quality data foundation. Organizations must address data silos, inconsistencies, and privacy concerns to build a common core data architecture.

  • Talent Gap: The demand for AI talent far exceeds supply. Organizations need to invest in developing existing talent and attracting new skills.

  • Organizational Change: Integrating AI across the enterprise requires a cultural shift. Employees at all levels need to understand AI's potential and how it impacts their roles.

  • Technology Infrastructure: AI demands significant computational resources and advanced infrastructure. Organizations must invest in modernizing their IT systems.

  • AI Governance and Ethics: As AI becomes more pervasive, the need for robust governance and ethical frameworks increases.

Facing up to Enterprise AI

Digital leaders must navigate substantial challenges to successfully transition to the second phase of AI adoption. From working with a variety of organizations on AI and digital technology adoption, I have found that there are critical principles that form the core of a successful strategy to win the battle for enterprise AI.

The first is to ensure you adopt a business-centric AI strategy. Develop a clear AI strategy aligned with the organization's overall business objectives. Identify high-impact use cases and prioritize AI initiatives accordingly.

Second, ensure you promote a clear data-driven culture across the organization. Foster a data-driven culture where data is seen as a strategic asset. Invest in data quality, governance, and accessibility.

Third, it is essential you invest sufficiently to build and develop your AI talent. Invest in AI talent development programs and create a collaborative environment where data scientists, engineers, and business leaders work together.

As our understanding of AI matures, these 3 principles can form a central core of an organization’s enterprise AI strategy. However, in practice the journey to enterprise AI is complex and demanding. Organizations that successfully navigate this transition will gain a competitive advantage, drive innovation, and create new business models. Make no mistake, the battle for the next phase of enterprise AI is underway.