The C-suite is abuzz with the transformative promise of artificial intelligence. Boardrooms echo with talk of generative models, autonomous agents, and a future where operational friction is a relic of the past. Companies are pouring unprecedented capital into AI initiatives, with global spending projected to surpass $300 billion by 2026, according to IDC. It’s a technological arms race where falling behind feels like an existential threat. But beneath the shimmering surface of this hype cycle lies a grim and costly reality: the vast majority of these projects are failing to deliver on their promise.

This isn’t a secret whispered in Silicon Valley corridors; it’s a well-documented crisis. According to a widely cited MIT study, a jaw-dropping 95% of AI projects fail to produce a meaningful return on investment. This disconnect between ambition and reality is creating what many are calling an “AI mirage,” where the promised oasis of efficiency remains perpetually out of reach. It’s a challenge that Muralidhar Krishnaprasad, President & CTO of Engineering at Salesforce, confronts daily. He argues that the tech industry and its customers need a serious reality check. The prevailing mindset of simply acquiring a large language model (LLM) and pointing it at a business problem is a recipe for disaster. “You can’t just take an LLM and throw it at a problem,” Krishnaprasad stated emphatically during a discussion at Dreamforce, cutting through the industry noise with a dose of engineering pragmatism. The real work, he contends, is far less glamorous but infinitely more critical.

The Great AI Reckoning: Moving from Hype to Tangible ROI

The widespread failure of enterprise AI isn’t due to a single, easily identifiable flaw. Instead, it’s a complex cascade of issues rooted in data, strategy, and culture. The allure of a turnkey AI solution is powerful, but it ignores the foundational work required to make these sophisticated systems function effectively within the intricate ecosystem of a modern business. It’s the digital equivalent of buying a Formula 1 engine and expecting it to work flawlessly in a family sedan without changing the chassis, fuel lines, or transmission.

The Anatomy of a High-Tech Failure

Understanding why these projects stumble is the first step toward building ones that succeed. The reasons are often interconnected, creating a web of challenges that can ensnare even the most well-intentioned initiatives. The core issue, as Krishnaprasad and other industry veterans point out, is that AI is not a plug-and-play technology; it’s a capability that must be meticulously woven into the fabric of an organization.

The Data Dilemma: Garbage In, Gospel Out?

The most common point of failure is data. The old adage “garbage in, garbage out” is amplified a thousand-fold in the age of AI. An LLM trained on messy, incomplete, or siloed data will produce unreliable, irrelevant, and sometimes dangerously incorrect outputs. Many organizations, in their rush to deploy AI, skip the painstaking process of data cleansing, unification, and governance. They fail to realize that AI models are only as good as the information they are fed. A sales prediction AI, for example, cannot provide accurate forecasts if it’s drawing from a CRM with duplicate contacts, outdated opportunity stages, and inconsistent data entry from the sales team. The model will confidently produce nonsense, leading to misallocated resources and a loss of faith in the technology itself. This foundational data strategy is the unglamorous, back-breaking work that ultimately separates successful AI deployments from expensive failures.

The Integration Nightmare: A Symphony of Silos

Another critical hurdle is integration. A powerful AI tool that exists in a vacuum is little more than a novelty. For an AI agent to be truly useful, it must have deep, real-time access to the constellation of systems that run the business: the CRM, the ERP, the marketing automation platform, the customer support database, and more. Without this seamless connectivity, the AI cannot gain the necessary context to perform its tasks. Imagine a customer service bot that can answer general questions but can’t access a customer’s order history or previous support tickets. Its utility plummets, frustrating customers and creating more work for human agents who have to clean up the mess. Salesforce’s entire strategy hinges on solving this, aiming for a unified platform where AI has a complete, 360-degree view of the business. As Krishnaprasad explained, Salesforce is evolving far beyond its roots. “Previously we were always just relegated to saying you are only doing sales, or service marketing – but now we are going beyond that, we’re managing customers, we’re managing employees, and we’re going to be managing operations…and most excitingly, we’re managing agents as well! All managed by our unified platform.”

Salesforce’s Blueprint: From ‘.com’ Pioneer to AI Pragmatist

Salesforce sees a historical parallel between the current AI frenzy and the dot-com boom of the late 1990s. Back then, every company was scrambling to get online, but many didn’t know how to translate a web presence into real business value. Salesforce’s success was built on simplifying that complexity, making the power of the internet accessible and useful for the enterprise through the Software-as-a-Service (SaaS) model. They are now applying that same pragmatic philosophy to artificial intelligence. “For a SaaS company like Salesforce, we’re making AI useful, just like we did making the dotcom useful for the enterprise,” Krishnaprasad noted, framing their mission. “We believe this is the best way to get AI working for the enterprise.”

“Eating Your Own Dog Food”: The ‘Customer Zero’ Philosophy

To build trust and demonstrate tangible value, Salesforce has adopted a rigorous “Customer Zero” policy. This means before any AI tool is released to the public, it is first deployed, tested, and refined internally across their own global operations. This isn’t just a quality assurance measure; it’s a core part of their product development, allowing them to experience the same challenges and opportunities their customers will face. It’s a philosophy of being “on the journey” with their clients, not just selling them a product.

A prime example is the development of their own AI-powered support agents. Krishnaprasad highlighted how his team built and launched sophisticated agents on the Salesforce help site in just a matter of weeks. The results speak for themselves. In their first year, these AI agents have handled more than 1.8 million customer conversations. This isn’t just an impressive number; it represents a fundamental shift in their support model. By automating the resolution of common queries and issues, these agents have liberated their human support specialists to focus on more complex, high-stakes cases that require nuanced problem-solving and empathy. This is a real-world, quantifiable success story that stands in stark contrast to the 95% failure rate plaguing the industry.

The Power of a Unified View

The secret sauce behind successes like the support agent project is the unified platform. By having all customer data, service history, and knowledge articles residing within a single, coherent ecosystem (the Einstein 1 Platform), the AI has the rich context it needs to be genuinely helpful. It can understand a customer’s history, anticipate their needs, and provide accurate, personalized solutions. This integrated approach is Salesforce’s core argument against the chaotic, multi-vendor, best-of-breed strategy that leaves data fragmented and AI initiatives starved of the context they need to thrive. The goal is to create a single source of truth that powers not just human employees, but an entire workforce of AI agents operating across every facet of the business.

The Human Element: Augmentation Over Annihilation

No conversation about AI is complete without addressing the profound fear of job displacement. The narrative of intelligent machines making human workers obsolete is a powerful and unsettling one. Krishnaprasad confronts this head-on, arguing that this fear, while understandable, is historically unfounded. He draws parallels to previous technological revolutions that sparked similar anxieties.

“This is a constant fear we all have as humans,” he acknowledged. “When the dotcom came we all feared for our jobs, when Tesla introduced auto driving, we all wondered what would happen – but the reality is – humans are the best at adapting. Over millennia, we have adapted to so many different things.” He likens the advent of AI to the introduction of electricity, a force that didn’t simply replace candle-makers but fundamentally reshaped society, enabling countless new industries and roles that were previously unimaginable. “AI is going to do the same, and we shouldn’t be fearing it, because it’s just going to make us better.”

From Drudgery to Discovery: Reimagining the Future of Work

The optimistic vision for AI in the enterprise is not one of replacement, but of augmentation. It’s about automating the mundane, the repetitive, and the soul-crushing parts of our jobs, thereby freeing up human intellect and creativity for higher-value work. Krishnaprasad points to the world of software development as a perfect example. He cites findings that suggest the average developer spends a staggering 40% of their time simply maintaining existing, often legacy, code. This is a massive drain on innovative potential.

The Liberated Developer and the Empowered Professional

Imagine a world where AI agents can handle the bulk of that code maintenance—patching security vulnerabilities, refactoring old codebases, and writing unit tests. “This is where, if AI can really help solve a whole bunch of issues there, imagine what new things we can create in just 20 years,” Krishnaprasad mused. That reclaimed 40% of a developer’s time can be reinvested into research and development, into designing novel user experiences, and into solving entirely new classes of problems. The same principle applies across the organization. A sales representative can be freed from manual data entry and guided by an AI that analyzes buying signals to recommend the next best action. A marketing manager can use AI to generate dozens of campaign variations in minutes, spending their time on strategy and creative direction rather than tedious production work.

The Adaptation Imperative for the Modern Workforce

This transition requires a profound shift in skills and mindset. The jobs of tomorrow will demand a different kind of expertise: the ability to collaborate with AI, to ask the right questions (prompt engineering), to interpret AI-driven insights, and to provide the critical thinking and ethical oversight that machines lack. The focus will move from rote execution to strategic direction. Krishnaprasad’s ultimate vision is one of symbiosis, where AI handles the drudgery and humans focus on the grand challenges. “We have so many drugs left to discover, so many processes we can make better,” he said. “This acceleration is happening because of technology, and with AI, it’s just going to go even faster…we will be able to leave out the drudgeries of maintenance, instead focusing on innovation, new ways of interacting, new ways of helping our human race together.”

The message from one of the world’s largest enterprise software companies is clear: AI is not a magic bullet. It is a powerful, complex tool that demands respect, preparation, and a thoughtful, human-centric strategy. For the 95% of companies whose projects are failing, the lesson is not to abandon AI, but to abandon the hype. Success will belong to those who are willing to do the hard, foundational work of preparing their data, integrating their systems, and, most importantly, empowering their people to partner with this transformative technology to build a more innovative and efficient future.

—

Discover why 95% of enterprise AI projects fail and how Salesforce advocates for a pragmatic, work-intensive approach to achieve real ROI. Learn the secrets to successful AI implementation, moving beyond the hype to build a true human-AI partnership.

AI Implementation, Enterprise AI, Salesforce

Source: https://www.techradar.com