Why Many HR Gen AI Projects Are Doomed to Fail

A dangerous narrative is taking hold among HR leaders and boards across the country: the promise of effortless, overnight transformation powered by artificial intelligence. This illusion of seamless automation from a technology still in its turbulent adolescence tempts organizations with a future that is as intoxicating as it is unrealistic, threatening to undermine the genuine, lasting value of this powerful new capability.

Promoters of agentic AI paint a dazzling picture of autonomous AI agents seamlessly orchestrating your entire business, delivering returns in under a year and creating exponential value. Yet this vision stands apart from the messy reality of 2025. This level of hype is not just optimistic — it is actively harmful, setting the stage for a potential crash that could undermine the real, tangible benefits of AI for years to come. 

The Flawed Promise of the Agentic Enterprise

Many reports tempt us with a future where “no-code agent builders” allow any business user to create AI workers, forming an “agentic AI mesh”: an interconnected ecosystem of programs autonomously negotiating, planning, and executing complex workflows. Imagine an AI agent in procurement autonomously identifying a supply need, negotiating terms with a vendor’s AI agent, and executing the purchase order without human input. Now, imagine that agent misreading a regional sales forecast and ordering $10 million of the wrong component, or the vendor’s agent exploiting a loophole in your agent’s programming to lock in unfavorable terms. 

The vision of full autonomy dramatically underestimates the immense challenges of reliability, security, and integration. Even state-of-the-art models can be surprisingly fragile and fail unpredictably, according to Stanford University’s AI Index Report. These agents must function within a company’s network of legacy systems — decades-old enterprise resource planning (ERP) software, proprietary databases, and custom-built applications — that were never designed for this kind of interaction.

Granting an AI agent unchecked access in this environment is not a strategic advantage; it is a security risk waiting to happen. The governance structures required to prevent catastrophic errors, malicious exploits, or costly “hallucinations” are massive undertakings the hype often ignores. The promise of an easy, no-code revolution is misleading when the underlying foundation is so complex and the cost of failure so high. 

The Ground-Level Reality: Where AI Shines Today

Should we abandon AI? Absolutely not. We must focus on the powerful, practical tools available now. The true revolution is not in full autonomy, but in powerful augmentation. In my work advising over two dozen organizations on AI integration, the most significant successes have come from pragmatic projects that solve current, specific needs. By targeting repetitive tasks, generative AI can deliver measurable returns without the existential risks of the agentic vision. 

The best results come when employees learn to use today’s generative platforms themselves. Capability transfer — not just solution delivery — means teaching employees to design, test, and maintain AI tools, instead of relying on outside vendors. When staff understand each design choice, learn to refine prompts or workflows, and apply governance directly, the organization keeps expertise in-house. That expertise grows over time, rather than being lost when a contract ends.

In every engagement, the fastest return on investment came after staff saw, step-by-step, how to build a no-code prototype — then repeated the process for another use case. Demonstrating the build process demystifies AI, reduces resistance, and sparks the curiosity that fuels constant improvement. Those steps can be mapped as follows: 1) introductory build, 2) behind-the-curtain demonstration, 3) guided co-creation, 4) independent innovation, and 5) long-term self-sufficiency. When insiders become builders, AI stops being a threat and becomes a valuable tool. 

How HR Teams Have Empowered Their Colleagues with AI

HR teams offer clear examples. One midsize automotive parts manufacturer in Indiana had three HR generalists assembling 40-page onboarding packets for every hire. During a two-week workshop, the most junior generalist — who had never used Visual Studio — built a conversational intake assistant with a drag-and-drop workflow in Power Automate. New employees now upload identification through a secure chat, complete direct deposit details, and e-sign policies in minutes. The bot checks for completeness, creates a personalized first-day agenda, and saves everything to SharePoint. The generalists handle only exceptions — usually an expired driver’s license or address mismatch — rather than spending hours on paperwork. Onboarding time dropped by 68 percent, and the generalist who built the flow now coaches supervisors on creating contractor and intern versions. 

Recruiting showed similar results in a 1,200-bed regional hospital facing nurse shortages. Two nurse educators, frustrated by résumé review delays, built a ranking copilot during a workshop. They customized an off-the-shelf language model using anonymized hiring records and a clinical-skills rubric, then connected it to the applicant tracking system. The copilot highlights licenses, shift preferences, and experience, bringing the top 20 résumés to the front. Educators review the short list, add nuance, and schedule calls the same week applications arrive. Phone-screen time fell from 13 days to four, first-year nurse turnover dropped eight percentage points, and the educators — now skilled tool builders — update their rubric each quarter to reflect insights from exit interviews.

Public-sector HR demonstrates the same approach works in unionized settings. A metro-area government managing 17 labor contracts dealt with high email volume for leave requests. A payroll analyst and union liaison participated in a co-creation sprint, sketched a decision tree in a whiteboard session, and in six weeks launched a bot that extracts key dates from emails, checks tenure in the ERP, and drafts approval letters using the correct contract language. Clerks who once re-entered data now coach employees on policy nuances and flag exceptions for legal review. Processing time per request fell from 25 minutes to six; leave grievances dropped by almost a third, and the city set aside stipends for frontline staff replicating automations elsewhere. 

Building a Pragmatic Path to Value

These case studies reveal that results last when the builders remain on staff. Success spreads by word of mouth — often through lunch-and-learns hosted by employees who developed the tools — leading to a culture in which managers find problems to solve, staff build solutions, and governance guides rather than blocks change. 

Psychology drives continuous improvement: people support the tools they create. This is well established in behavioral science and codified in my methodology. Workers once wary of automation now look for new ways to put it to work. Leaders, in turn, learn the fastest way to reduce risk is to empower the experts most familiar with the workflow. When insiders design and test tools, they find real-world edge cases early, replace guesswork with knowledge, and turn AI into a transparent process.

The lesson is simple: start small, let insiders build, and watch expertise grow. One HR assistant who builds an intake bot can save months of vendor costs, inspire peers to try low-code tools, and help create a culture of continuous improvement. Dependency fades, costs fall, and organizations develop advantages money can’t buy. Across the organizations I’ve advised, replacing vendor-built automations with staff-built solutions reduced project costs by about two thirds and eliminated recurring maintenance fees within a year. 

According to Gartner’s Hype Cycle methodology, after the “Peak of Inflated Expectations” comes the “Trough of Disillusionment.” The current frenzy makes that trough arrive faster. The organizations that will succeed are those focused on staff-built, pragmatic solutions — brick by brick, solving real problems and creating measurable value — while others remain distracted by unrealistic promises.

Dr. Gleb Tsipursky, called the “Office Whisperer” by The New York Times, helps tech-forward leaders replace overpriced vendors with staff-built AI solutions. He is CEO of the future-of-work consultancy Disaster Avoidance Experts and author of eight books, including The Psychology of Generative AI Adoption (2026) and ChatGPT for Leaders and Content Creators (2023).