The business value gap: Why AI adoption alone isn't enough

Organizations treated adoption as the fastest way to AI value. Maturity increased, but value has not kept pace. So now what?

When GenAI was still new, leadership conversations often started with a simple question: "Should we be doing more with AI?” For most organizations, the answer was yes. They focused on rolling out tools for employees and building AI awareness and literacy. Headlines carried different versions of the same message: “Adopt AI or risk being left behind.”

The expectation was that using AI would guarantee a major competitive advantage for companies and improve the bottom line. The more people used AI, the more value it would bring to individuals, teams and organizations. The winners had the most tools and the most pilots. Individuals could be more productive, even outside their traditional areas of expertise. Work that once took weeks could now take hours or even minutes.

If AI adoption was the measure of success, what happens when adoption is high but the value remains unclear?

“AI adoption is high. Every function has its own AI pilots, but no one can agree on which are the ones worth scaling, and now the board is asking us to justify our spend.”
- Business Leader

How the leadership conversation has changed

While adoption was the initial focus, we are increasingly seeing that organizations have reached a point where AI ambitions are outpacing foundational readiness. This is resulting in an ever-widening gap between adoption and measurable business value.

In our CGI 2026 Voice of our Clients research, we found:

• High AI adoption rates: 86% of organizations are advancing agentic AI adoption, with 62% embedding it into core business and operational processes.
• Major ROI gaps: Only 51% quantify AI outcomes, while results from digital strategies are plateauing and business and IT model agility remains low.

Research from other organizations supports these findings. For example, in Glean’s Work AI Index, 75% of workers say AI has made them more productive, with automation saving roughly 11 hours of work per week. Yet only 13% say it has meaningfully improved their organization’s performance.¹ This gap is difficult for leaders to ignore.

“There is an expectation that AI will enable a productivity boost, but there are major challenges in understanding how to best apply it in our day-to-day business.”
- Senior Leader, Energy & Utilities

Boardroom and leadership discussions around AI increasingly center on a different question: "Are we getting value from it?" Answering this requires different measurements and metrics, different skillsets and mindsets and a wider group of roles and stakeholders.

A different set of challenges

With previous technologies, adoption was often the barrier, and it took much longer than we’ve seen with GenAI. What is not new, but is becoming increasingly important, is the integration challenge. To deliver value beyond individual productivity gains and pilots in controlled environments, AI must operate within complex, mission-critical systems developed over decades.

Enterprise value is created only when technologies integrate fully into real-world operations. As AI continues to advance, such as with agentic AI or in combination with technologies like quantum computing, that complexity only compounds.

Where we’re seeing success

Organizational and foundational readiness, enabled through digital reengineering, are emerging as key enablers of digital strategy success, often outweighing frontier AI ambition alone. This is where operational and process excellence become critical. Leading companies are focusing on execution discipline, value attribution and closing the gap between ambition and returns.

And increasingly, we’re seeing the strongest results when there is a clearly defined business problem, integrated into real workflows and measured against tangible outcomes. Here are some examples.

1. Eliminated a backlog of 60,000 claims with AI integration

A global manufacturing company accumulated more than 60,000 client claims due to an ineffective claims-handling process, lack of tools and difficulty using historical data effectively. We worked with the client to integrate AI directly into the claims-handling process within its enterprise resource planning (ERP) system, enabling the company to clear its backlog. Pre-work for new claims is now automated with custom AI agents that address specific processes and scenarios, leveraging historical claims data. This has freed employee time for other value-creating tasks.

Lesson: Define value and apply AI to a measurable business process

We see greater success where clients incorporate measurement into the workflow before deployment, instead of bolting on a dashboard afterward.

2. Strengthened AI governance with transparency and control

A major German financial institution successfully embedded AI governance into its banking workflows by:

• Implementing risk-based classification systems
• Standardizing evaluation templates
• Applying AI-specific testing procedures
• Integrating continuous operational monitoring

This supported alignment with regulatory requirements and international standards like ISO/IEC 42001. This transformed AI governance from a compliance checkbox into a practical framework that established clear accountability, managed risk throughout the AI system lifecycle and created the transparency and control needed to build trust and scale AI beyond pilots.

Lesson: Build in responsible use and security

We see success where governance is not treated as compliance alone, but when responsible use and security are built into processes as guardrails.

3. Redesigned workflow saves 3,000+ hours annually

ECH, a leading provider of independent living and aged care services in South Australia, had care partners spending up to 60 minutes per client manually compiling care plans, with inconsistent formatting and frequent omissions of key information. The goal was to reduce administrative burden so care partners could focus more on clients. We redesigned the workflow with:

• A GenAI-powered care plan generation system using Amazon Bedrock and Claude 3.5 Sonnet v2
• Amazon Transcribe to support audio transcription of wellbeing interviews
• Custom prompt engineering to create an empathetic AI care partner persona
• A secure AWS landing zone tailored to healthcare compliance requirements

The new system produced comprehensive, standardized care plans in 36 seconds, improving accuracy from 35.3% to 92.5%, and saving over 3,000 hours across 2,300 clients annually. This shifted valuable time from administration to client support.

Lesson: Reengineer before automating

Clients see the greatest success when they reengineer workflows and integrate change management, rather than adding AI to an existing, imperfect process.

Are your assumptions about AI stalling your progress?

With AI and other frontier technologies advancing faster than they’re being integrated, assumptions that accelerated early adoption may now be slowing progress. Here are some common themes from our client conversations.

1. “I should conduct multiple AI pilots across different areas of my business to identify the best use cases.”

One-off AI pilots can keep technology risk and investment relatively low. They can also build initial momentum, but for leaders, the danger is that they often create the illusion of progress.

Solution: Start with the problem and let technology follow the business.

2. “My AI projects are not delivering value because I don’t have the right tools and technical environment in place.”

When AI projects are not going as expected, the instinct may be to stop the project, return to the previous way of working or try a different set of tools and technologies. While technical expertise and the right environment are critical, the underlying issue may lie elsewhere.

Solution: Understand the human side of the work and develop trust in the systems and processes supporting it before assuming technology is the problem.

3. “I should wait for better data or more favorable conditions before scaling AI across my organization.”

Better data and stronger foundations can improve outcomes, but waiting for perfect conditions can delay progress.

Solution: Define clear outcomes and business value first. AI can also help improve the data along the way.

From adoption to value

AI adoption was the starting point, not the finish line. Real value comes when AI connects with the right business problems and integrates into how work gets done. It also needs to be measured against outcomes that leaders can see and trust. Organizations that make this shift will turn AI from promising technology into lasting business value.

 

Source:

1. Rebecca Hinds et al., “Botsitting, botshitting, and the hidden human labor of AI at work,” Work AI Index 2026, Work AI Institute, Glean, accessed August 26, 2026, https://www.glean.com/work-ai-institute/reports/work-ai-index.

Back to top