Avoiding AI failure: How to drive real business value with AI

95% of AI projects fail, according to research from MIT. As organizations pour resources into AI initiatives, the gap between expectation and realized value continues to widen, leaving IT leaders searching for answers.

There are many reasons why AI projects don’t succeed, but one of the primary culprits lies in the data, which often resides in silos scattered across enterprise environments. Without unified access to comprehensive, quality data, even the most sophisticated AI models will produce irrelevant or inaccurate results.

“Data needs to be accurate, clean, and managed under clear governance,” explains Rusty Searle, interim CIO at Elastic. “You can push all the data into an AI tool, but if the quality isn’t there, the output won’t tell the right story. Poor or siloed data will always show up in the results.”

Organizations that skip this crucial step will find themselves with AI systems that generate impressive demos but fail to deliver consistent business value in production environments. As Jay Shah, senior director of enterprise applications at Elastic, puts it: “Garbage in, garbage out. It’s paramount to have fast access to quality data. It directly impacts the contextuality, reliability, and accuracy of the response output.”

Focus on solving business problems, not cool technologies

Successful AI implementation demands a shift from technology-first to problem-first thinking. Rather than asking “What can AI do?” leaders should start with “What specific business problems need to be solved?” This approach ensures that every AI initiative connects directly to measurable outcomes and genuine user needs.

“Every AI initiative should connect back to a defined problem with measurable outcomes,” emphasizes Searle. “We shouldn’t chase ‘cool tech’ or the latest version just for its own sake without clarity on the use case.”

AI cannot succeed as an IT-only initiative. When business partners — from HR to legal teams — actively participate in defining requirements and maintaining data quality, not only do AI initiatives advance an organization’s strategic goals, but users are much more likely to use the AI tools. After all, it doesn’t matter how impressive AI’s capabilities are if no one uses them.

Elastic experienced this firsthand in the development and deployment of ElasticGPT, its generative AI (GenAI) employee assistant. The goal of the assistant was to provide employees with a GenAI chatbot that could quickly find information they need — like onboarding checklists, laptop upgrade request forms, and company policy documentation — across multiple proprietary data sources. The ultimate aim was to increase employee productivity, improve the reliability of the source data, and create a scalable foundation for future AI initiatives.

Designate an AI champion

“Successful AI implementations involve identifying a single-threaded leader to drive your organization’s vision forward. Pair this with a dedicated group of individuals set on a clear goal and measurable outcomes. This core team will navigate and orchestrate the initiative across business lines to drive the right priority,” said Shah. “Our AI outcomes accelerated exponentially when we identified a lead for our use cases. We have since expanded into a small center of excellence focused on AI.”

The role extends beyond technical oversight. The AI project lead must be able to work not only with various teams within IT — navigating infrastructure challenges, API configurations, compliance requirements, and data quality evaluation — but also with business stakeholders to ensure that the implementation stays on track to meet their needs.

Meeting users where they are

The most successful AI applications integrate seamlessly into existing workflows rather than requiring users to adopt new tools or processes. When AI functionality appears where users already work — in their communication platforms, existing applications, or familiar interfaces — adoption happens organically. This embedded approach transforms AI from an additional burden into an everyday tool that increases productivity.

“It’s critical to embed AI into existing workflows and ensure people are trained and comfortable using it. Otherwise, it becomes shelfware,” Searle explains. “A good example is ElasticGPT in Slack, since it’s embedded where people already work every day, adoption happens almost by default. It enhances the experience instead of forcing users to change habits.”

Change management also plays a crucial role. As AI technology continues to evolve rapidly, organizations must continuously educate users about AI developments to foster a culture of experimentation and growth.

Recover when things go off track

Given the complexity and novelty of AI technologies, it would be unrealistic to expect deployments to go smoothly. At first, Elastic faced challenges at the beginning of its AI adoption journey.

“Early on, we tried to create assistants and attach them to our systems, but we quickly realized the need for better orchestration to ensure that AI interactions could flow securely, contextually, and in alignment with our existing tools,” said Searle. “To address this, we adopted Langchain, a library that helps orchestrate AI integrations, allowing us to build a more structured framework that naturally inherits the access controls and context of our native systems.”

Real-world success with a GenAI employee assistant

Elastic’s internal GenAI assistant, ElasticGPT, demonstrates these principles in action. Built on Elastic’s Search AI Platform with retrieval augmented generation (RAG), the tool connects data sources across the organization, making information accessible through familiar channels. The results speak for themselves: 63 hours saved per employee annually with a two-month payback period and a 98% user satisfaction rate.

The success of ElasticGPT stemmed from several key decisions.

• First, the HR and IT teams started with genuine user needs, addressing common employee issues such as “It is difficult to find the information I’m looking for” and “The process to upgrade my laptop is too complex.”
• Second, Elastic invested heavily in data quality and accessibility, cleaning up its wiki, support data, and strategy playbooks during development.
• Third, they provided multiple access points for employees — the company intranet and Slack — to integrate AI into existing workflows so the use of AI felt seamless.

Start driving business value with AI

Building successful GenAI apps demands strong leadership, high-quality data, and AI integrated into daily operations. It also requires moving beyond isolated software-as-a- service tools to a comprehensive platform that grows as your business grows. 

Follow Elastic’s 8 steps to build a scalable GenAI app to deliver measurable results.