Almost every CEO on the planet has been promoting the potential of Geneai and Agent AI for the past two years as a future of efficiency and profitability improvements. Naturally, we are now in the midst of an AI supercycle that drives the biggest technological change that businesses have seen in the last 25 years. But before that happens, business leaders need to find a way to transform these new technologies into concrete results.
Often, it's not happening today. In fact, according to BCG, about 70% of enterprise AI initiatives have failed to achieve value and expand. Reflecting these findings, our recent Enterprise AI survey found that 40% of senior executives say that the AI initiative does not exceed the pilot phase.
That is, the majority of companies using Genai and experimenting with AI agents do so in small pockets and silos within their organizations. There may be special tools and narrow sets of solutions, but few people offer the end-to-end, AI-powered workflows that the world has long dreamed of.
Not potential, but integration issues
What is causing this growing gap between potential and real-world values? The common bond in all of these struggling AI projects is the lack of understanding of the key interactions between data, domain expertise, and the ability to integrate AI into enterprise workflows. You can have the most powerful leading language model (LLM) or the world's most accurate agent agent framework, but if you can't fully integrate technology into each step in a complex process, you won't create value for your end users.
This is especially true in complex and regulated industries such as insurance, banking, finance, and healthcare, where great opportunities for AI to streamline processes and improve the customer experience are often hidden by data restrictions, security challenges, and management obstacles.
For example, let's look at the insurance industry. Every year, the global insurance industry spends around $7 trillion in claims management and underwriting processes, spending around $350 billion. This includes everything from actuarial risk modelling to ensuring policy coverage details match consumer payments. Each step in that series of events involves thousands of data points, handoffs between career staff and customer touchpoints. It is also full of inefficiencies.
In the personal line auto insurance business alone, insurers lose an estimated $30 billion each year due to missing or incorrect errors in underwriting or other errors that occur in the claims process. Of course, this is exactly the type of repetitive, manual, labor-intensive task that is highly data-rich, designed to simplify AI. However, many insurance companies are struggling with AI integration.
This is because even the best off-the-shelf AI models and tools are not designed for this type of professional use case, and few companies have modernised their data estate until all the basic information used in one area of business is available to all other business functions. As a result, efforts to modernize with AI usually encounter challenges with data silos, resulting in results that are not accurate enough to be completely trusted.
AI for workflows with domains, data and AI
There are two solutions. First, you need to embed AI directly into your workflow with a powerful combination of domain expertise, data accessibility, and the right AI technology. Second, seamless orchestration of these elements (domains, data, AI) is required at a speed, allowing organizations to accelerate their value realization and move quickly from experiments to concrete business impacts.
This means starting with expertise rather than technology. Before you can successfully integrate your AI model into your workflow, you need someone with a deep understanding of the domain (the nuances and components of that workflow) and you need to know where to go to get all the data you need to make things work. If you want to automate your insurance workflow, you need to know your insurance in close detail. This is not a place where generalist or technology-specific knowledge replaces experience. The same applies to banking and finance, healthcare, consumer packaged products, and all other industries with their own complex processes.
From there, data is important. Data drives AI. This is an important ingredient, especially if you want to use AI in your workflow. However, in many cases, businesses believe they need to engage in large data migration exercises to get all data into a fully centralized repository or data lake before making it fully utilized for AI applications. it's not. Using modern data ontology and APIs that can connect to multiple applications, you can extract only the data needed for a particular feature, marry other data sets, and create a fully data-driven workflow.
Then, once the right people are placed and the right data is accessible, it's time to start digging up AI use cases. Returning to the insurance claims example, this is the time to start tweaking LLMS, test your data extraction tool to see how well your underwriters can reach faster, more accurate views of the risks of the claims process, reduce data anomalies and lag. This is the stage in which the team discovers that off-the-shelf Genai solutions may not be the most effective tool for scrutinizing claim documents and critical insights on the surface, but custom agent models may be the best solution for claim analysis.
Adjusting AIELED Workflows
From understanding the workflow details, navigating the data needed to integrate AI solutions into AI integration, that complete process is unlocked, so that AI enterprise value is adjusted in a deliberate, highly choreographed way before it is unlocked. The true winner of AI Arms Race is the one that can analyze the problem they are trying to solve first, and then helps Cherry choose the right dataset and AI solution, achieving those results as quickly, inexpensively and accurately as possible.
Approaches to help organizations integrate AI into their workflows include leveraging the domain expertise, data, and AI capabilities they have built over the years. This allows clients to tailor the right horizontal and vertical solutions to help them reach value efficiently, accurately and cost-effective. Continuingly innovating both horizontal and vertical stacks, delivering value to our clients at speed.
Today, the work we do to incorporate AI into the enterprise workflows of major insurance, banking, finance and healthcare payment machine organizations is not only reducing efficiency and costs, but also driving improved customer experience and improved business outcomes. For example, in Healthcare, around $180 billion wasted every year on false bill payments, so by using AI-powered algorithms to find false or false billing instances, it was able to bring $2.2 billion in savings back to the healthcare system. It's much bigger than efficient play. It has reached the core of the complex problems that have been challenging the industry for decades. And we're just starting out.
In learn more Visit www.exlservice.com to learn how EXL incorporates AI into some of the world's leading business workflows.
About the author:
Rohit Kapoor is chairman and CEO of EXL, a global data and AI company.
