Why Most Enterprise AI Projects Struggle to Deliver Results

Many companies pour money into AI projects but see little return. Over half of enterprise CEOs report no clear financial gain from their AI investments. The problem often isn’t the AI itself. It’s that the AI agents don’t understand the business they serve.

Take Absa, one of Africa’s largest banks. In 2022, it took 120 days to deploy a new financial crime detection model. That’s a long wait, especially when threats evolve fast. By 2023, Absa cut deployment time to just 15 days. They kept the same rules, transaction volumes, and compliance needs. What changed was how they integrated AI deeply into their workflows and data.

AI Needs More Than Just Models

AI models alone don’t solve problems. They must fit into existing processes. Bratin Saha, CEO of NTT DATA AIVista, puts it simply: “It’s not just a model, you’re building a system around the model.” He adds, “When you’re deploying AI in the enterprise, you’re not deploying a technology. You are taking a workflow that exists and taking it from point A to point B.”

This means AI must align with business rules, workflows, and compliance. Most AI projects fail because of poor integration and gaps in domain knowledge. Enterprises face complex, regulated workflows like multinational insurance claims. These workflows involve handwritten notes and checkboxes that AI struggles to handle without context.

Specializing AI with customer data and tribal knowledge improves reliability. Wrapping models in enterprise data and guardrails completes the “last mile” of operationalizing AI. Without this, AI agents remain unreliable and disconnected from actual business needs.

Why AI Builders Matter

Most employees use AI tools occasionally. About 50% use AI at least once in a while, and 15% use it daily. But only about 5% qualify as sophisticated AI builders. They create custom tools that embed AI directly into workflows.

At BBVA, a bank known for AI innovation, employees built over 20,000 custom GPTs. Around 4,000 of these are in regular use. This shows how empowering employees to build AI solutions tailored to their work makes a big difference.

Building AI is often invisible. To change this, companies must make AI builds unavoidable, celebrate builders, and track their creations. This helps shift employees’ identities from AI users to builders. That shift unlocks real AI productivity.

The Economics of Enterprise AI

The cost of AI tokens dropped over 90% between 2023 and 2026. Despite this, enterprise AI spending more than doubled. Lower token costs let companies run more AI agents, automate more workflows, and generate more code.

Still, the financial benefits remain uneven. AI decision-making has driven an 8% lift in sales for a global food and beverage company. It also sped up shelf resets by 35%. These wins come from tight integration of AI into business operations, not just flashy models.

By 2027, half of all business decisions will be augmented or automated by AI agents, Gartner predicts. But to get there, companies must focus on governance, domain expertise, and clear ownership of AI outcomes. Without these, AI remains a tool that struggles to deliver.

As Caroline Davis, Chief of Staff at Capital Factory, puts it, “Most of my day is run through Claude at this point.” This shows how AI agents can become indispensable when they know the business and fit seamlessly into daily work.

Stay connected via Google News

Follow us for the latest AI News