Your GenAI project will fail without basic project management principles

Your GenAI project will fail without basic project management principles
Image by wirestock on Freepik

Introduction

In early 2024, Air Canada learned an expensive lesson about deploying AI without proper testing or oversight. The airline’s chatbot confidently told a passenger he could book a full-price ticket to his grandmother’s funeral and apply for a bereavement discount within 90 days. When Jake Moffatt tried to claim the promised discount, Air Canada denied it. A tribunal later ruled the airline liable for the chatbot’s misinformation, ordering them to pay CA$812.02 in damages. The airline hadn’t taken “reasonable care to ensure its chatbot was accurate,” essentially skipping the most basic project management step: testing before deployment.

Air Canada isn’t alone:

According to a August 2025 MIT report, 95% of generative AI pilots at companies yield no measurable business return.
Gartner predicts that at least 30% of GenAI projects will be abandoned after proof of concept by the end of 2025, citing poor data quality, inadequate risk controls, escalating costs, and unclear business value.
Harvard Business Review reports that some estimates place the AI project failure rate as high as 80%, nearly double the rate of corporate IT project failures from a decade ago.

The common thread? These failures rarely stem from inadequate technology.

They fail because organizations skip fundamental project management principles.

Despite the hype around GenAI being “revolutionary,” the basic project management disciplines of clear objectives, stakeholder buy-in, planning, testing, budget control, and change management still determine whether projects succeed or collapse.

The unchanged fundamentals

GenAI offers new capabilities. It can tackle problems that were completely unsolvable just three years ago, from natural language understanding to content generation at scale. But here’s what hasn’t changed: how you implement any technology follows the same patterns that have worked since humans started doing projects.

The ancient Egyptians demonstrated exceptional project management when building the pyramids. According to research on ancient project management, they used division of labor, hierarchical supervision, phased planning, and quality control so precise that the variation between the longest and shortest sides of the pyramid’s base is less than eight inches.

The Romans contributed standardization and infrastructure project methodologies that project management authorities compare to current best practices. These weren’t primitive guesses; they were sophisticated management principles refined over centuries.

The same fundamentals apply to GenAI projects

PM Principle	Why It Matters
Clear objectives and success metrics	Projects need defined, measurable goals, not vague objectives like “implement AI” or “become AI-powered”
Stakeholder identification and buy-in	Every affected party, from end users to compliance teams, must understand and support the initiative
Budget control and resource allocation	Projects require funding not just for initial purchase, but for integration, training, maintenance, and iteration
Risk assessment and mitigation planning	Potential failures must be identified before they become actual disasters
Testing and quality assurance	No system should reach production without thorough validation under realistic conditions
Change management and training	People need preparation, support, and clear communication about how new tools will affect their work

These principles aren’t bureaucratic obstacles. They’re the accumulated wisdom of thousands of successful projects across every industry and across many centuries.

Where GenAI projects go wrong (some real examples)

When organizations skip basic project management for GenAI implementations, the failures follow predictable patterns:

What Happened	PM Principle Ignored	The Predictable Result
Company implements ChatGPT licenses across the organization without defining specific problems to solve or success metrics.	No clear objectives	Employees experiment randomly, no one tracks value, ROI remains unmeasurable, executives question the investment within months.
IT department selects and deploys an AI tool without consulting the end users who will actually use it daily.	No stakeholder buy-in	Tool sits unused because it doesn’t fit real workflows, investment wasted, team resentment builds, shadow IT solutions emerge.
Company goes straight from vendor demo to full company-wide deployment without piloting with a small team first.	No testing phase	Critical errors and limitations surface in production, employees lose trust in the tool, damage control consumes more resources than proper testing would have.
Project scope expands from “chatbot for customer service” to “AI-powered transformation of all customer interactions” without gate reviews.	No budget control	Project runs over budget, timelines extend, executives lose patience and cancel the initiative before anything ships.
New AI tool appears in employees’ software stack on Monday morning with a two-paragraph email and no training.	No change management	Confusion, resistance, anxiety about job security, employees create workarounds, adoption rates remain below 20%, project declared a failure.

These aren’t hypothetical. Consider New York City’s Microsoft-powered MyCity chatbot, which launched without adequate testing in 2024. The AI gave entrepreneurs dangerously incorrect advice, including telling users they could fire employees for reporting harassment or keep customer tips for the business, both violations of city and federal law. The failure revealed poor data quality and insufficient governance,
the exact issues that standard project management testing phases would have caught.

Or McDonald’s AI drive-through pilot, which IBM developed and deployed to over 100 US locations. Viral TikTok videos showed the system repeatedly adding unwanted items, with one customer’s order escalating to 260 Chicken McNuggets despite repeated requests to stop. McDonald’s ended the partnership, but not before significant brand damage and customer frustration.
A proper pilot program with controlled testing and clear success criteria would have identified these issues before wide deployment.

The NTT Data research found that 70-85% of GenAI deployment efforts fail to meet their desired ROI, primarily due to human factors: lack of trust, insufficient training, and poor change management. The MIT report echoed this, finding that failure was “less about what’s wrong with today’s AI models and more about what’s wrong with how companies are trying to use them.”
Technical capability isn’t the bottleneck; project management discipline is.

The false belief that “AI is different”

Many leaders genuinely believe that GenAI is so revolutionary that traditional project management will slow innovation and competitive advantage. This belief is dangerous because it contains a kernel of truth wrapped in a false conclusion.

What’s true: The technology capabilities are genuinely new. GenAI can perform tasks that were impossible three years ago. The pace of AI advancement is unprecedented.

What’s false: That revolutionary technology requires abandoning proven implementation principles.

Most projects are still “change projects”

You’re introducing new tools to people with existing workflows, mental models, and concerns. You’re allocating budget and resources. You’re expecting measurable business outcomes. These dynamics don’t change just because the underlying technology is novel.

Historical parallels

When companies first adopted computers in the 1960s and 1970s, the technology was revolutionary, yet research on early management practices shows that successful implementations followed traditional project planning principles. The same pattern repeated with email adoption in the 1990s and cloud computing in the 2010s. Revolutionary technology, unchanged implementation fundamentals.

“But doesn’t project management slow things down?”

Of course, proper planning adds initial overhead. You’ll spend days or weeks on stakeholder interviews, pilots, and testing rather than immediate deployment. However, HBR research on AI project success found that projects following structured management approaches had significantly higher success rates. The MIT study noted that the small percentage of AI pilots that succeeded (5%, based on the 95% failure rate) did so because organizations took time to understand workflows, train users, and integrate tools properly rather than rushing to production.

Failed projects feed “The AI bubble will burst”

Failed projects are the ultimate innovation killer. A GenAI initiative that collapses after six months and $2 million doesn’t just waste resources; it makes your organization skeptical of future AI investments, damages champion credibility, and potentially sets you back years. The “fast” approach that skips planning becomes the slow approach when you factor in recovery time from failure.

The principles that matter most for GenAI

While all project management fundamentals apply to GenAI, five principles prove especially critical:

1. Define clear, measurable objectives

Why it matters for GenAI: Vague goals like “implement AI” or “leverage GenAI” provide no basis for tool selection, success measurement, or project scope control. GenAI’s versatility makes focus essential.

How to do it: Start with a specific business problem and quantifiable success metric. Not “improve customer service with AI” but “reduce customer service response time from 2 hours to 30 minutes while maintaining 90%+ satisfaction scores.” This clarity guides every subsequent decision: which tool to select, how to measure success, when to expand beyond the pilot.

Practical tip: Write your objective as a before/after statement with numbers. If you can’t articulate the quantified improvement you expect, you’re not ready to select technology.

2. Identify and engage stakeholders early

Why it matters for GenAI: AI implementations typically affect multiple departments (IT, legal, compliance, end users, managers) with different concerns (security, privacy, job impact, workflow changes). Skipping any stakeholder group creates landmines.

How to do it: Map every group that will be affected by or needs to approve the project before selecting a tool. Interview representatives from each group to understand their priorities, concerns, and requirements. The PwC pharmaceutical client case study demonstrated this by using GenAI to analyze stakeholder interview transcripts, ensuring all perspectives informed the SAP transformation process.

Practical tip: Create a simple stakeholder matrix listing each group, their primary concern (e.g., “compliance team worries about data privacy”), and how your project addresses it. Share this document with all stakeholders to demonstrate that their input shaped the project.

3. Start small with pilot testing

Why it matters for GenAI: AI behavior can be unpredictable, and integration challenges often emerge only in real-world use. Full organization deployment before validation multiplies risk exponentially.

How to do it: Select one team, department, or process for initial deployment. Run for 30-60 days collecting both quantitative metrics (usage rates, time saved, error rates) and qualitative feedback (user satisfaction, workflow friction, unexpected issues). Use insights to refine before wider roll-out.

Practical tip: Choose your pilot team carefully. You want early adopters who’ll engage honestly but represent typical users, not just tech enthusiasts who’ll overlook issues other employees will find critical.

4. Budget for the full life-cycle

Why it matters for GenAI: Gartner’s research found that escalating costs are a primary reason for project abandonment. Many organizations budget only for initial licensing, then face surprise expenses for integration, training, and maintenance.

How to do it: When creating your budget, include:

Initial tool licensing or development
Integration with existing systems (often 2-3x the initial cost)
User training and support materials
Ongoing maintenance and updates
Iteration based on feedback
Personnel time for project management and change leadership

The PwC case study showed that GenAI itself can help with budget management by rapidly assessing change request impacts on project scope and resources, but only if you’ve built that capability into your plan.

Practical tip: Apply the 1-3-1 rule as a rough estimate: if the tool costs $1, expect $3 for integration and support, and another $1 annual ongoing costs.

5. Plan for change management

Why it matters for GenAI: NTT Data found that human factors, lack of trust, fear about job security, and insufficient training cause most GenAI failures, not technical limitations.

How to do it: Build a communication and training plan before deployment:

Explain clearly why you’re implementing GenAI and what problems it solves
Address job security concerns directly and honestly
Provide hands-on training tailored to actual workflows
Identify and empower champions within each user group
Create easy access to support when users have questions
Celebrate early wins and share success stories

Practical tip: Allocate at least 20% of your project timeline specifically to change management activities. If you’re spending 80% on technology selection and deployment and 20% on the rest, your ratios are wrong.

These five principles align with standard methodologies you already know. PMBOK emphasizes stakeholder management and scope control. Agile prioritizes starting small and iterating based on feedback. The terminology might vary, but the fundamentals remain constant across frameworks.

What good looks like

When organizations apply proper project management to GenAI implementations, the contrast with failure stories is striking. Consider a global pharmaceutical company that integrated GenAI into their SAP S/4HANA transformation following structured project principles.

The PwC case study describes how the company applied GenAI across their “end-to-end process harmonisation and technical system build journey,” delivering value “throughout the end-to-end lifecycle, from preparation and planning to discovery, design, realisation, testing, go-live and post-implementation support.”

Rather than assuming GenAI would solve everything, they identified four specific use cases where automation would provide clear value:

Streamlining input collection: Automated workshop summaries and stakeholder interview transcripts
Dynamic change request management: GenAI-powered assessment of change request impacts on scope and resources
SOP updates and compliance: Automated updates to standard operating procedures for regulatory compliance
Automated BPMN generation: Creating business process diagrams from stakeholder transcripts

By focusing on specific, measurable applications rather than vague “AI transformation” goals, the company achieved tangible results: streamlined processes, accelerated project timelines, and most critically, boosted stakeholder confidence in the SAP system itself.

The results demonstrate what following PM fundamentals delivers: increased efficiency, reduced manual effort, accelerated timelines, and most critically, enhanced stakeholder confidence. The project succeeded not because the AI was more sophisticated, but because the implementation was more disciplined.

This pattern repeats across successful implementations. Google Cloud’s real-world GenAI case studies show companies like YDUQS (Brazilian education) achieving 90% success rates and 4-second response times by automating specific processes with clear metrics, not by pursuing vague “AI transformation” goals. The difference isn’t the technology; it’s the rigor applied to implementation.

Conclusion

GenAI projects fail at alarming rates not because the technology is immature, but because organizations abandon the project management principles that have ensured success for thousands of years. From the pyramids of ancient Egypt to modern enterprise transformations, the fundamentals remain unchanged: define clear objectives, engage stakeholders, test before full deployment, control budgets, and manage change deliberately.

We all already know these principles. The challenge isn’t learning new project management skills; it’s remembering to apply them when caught up in AI hype and competitive pressure to “do something with GenAI.” The 95% of pilots that fail aren’t run by incompetent teams; they’re led by capable professionals who convinced themselves that this time, with this technology, the fundamentals didn’t apply.

Before your next GenAI initiative, run through this basic checklist:

Are objectives clearly defined and measurable? Can you articulate the specific problem and quantified improvement target?
Have all stakeholders been identified and consulted? Have you mapped every affected group and understood their concerns?
Is there a testing plan before full deployment? Have you identified a pilot team and defined success criteria?
Has budget been allocated for the complete life-cycle? Have you included integration, training, support, and iteration costs?
Is change management built into the timeline? Have you planned communication, training, and support activities?

If you can’t answer yes to all five, pause and address the gaps. Taking an extra two weeks for planning beats spending six months recovering from a preventable failure.

Consider Eanis as one example of this focused approach. Rather than promising to “transform your organization with AI,” it solves one specific, measurable problem: reducing the time employees and managers spend on repeat questions.

Clear objectives (time saved, coverage metrics),
defined stakeholders (managers and employees),
controlled scope (starts with one role), and
built-in feedback (weekly digest showing what’s working).
The principles that built the pyramids, applied to modern GenAI implementations.

Ready to implement GenAI with proper project management fundamentals? Start with one clearly defined problem, involve the right stakeholders from day one, and test rigorously before expansion. The technology is revolutionary; your implementation approach should be proven and disciplined.

Learn more about focused GenAI implementation at Eanis or contact us for a consultation on applying project management principles to your AI initiatives.

References

Learn more about Eanis - See how it works - Schedule a call