More than 80% of AI projects fail. That’s twice the failure rate of traditional IT projects.
The numbers get worse. In 2025, 42% of companies abandoned most of their AI initiatives—up from just 17% the previous year. The average organization scraps 46% of AI proof-of-concepts before they reach production.
We’ve seen this pattern repeatedly across data warehouse and AI projects. The mistakes are predictable. The consequences are expensive. And somehow, teams keep making the same errors despite knowing better.
Mistake #1: The Planning Problem Nobody Wants to Face
Here’s what surprises us most: the meetings where requirements should be defined often don’t happen at all.
Data engineers build warehouses based on what they “feel is best.” AI teams train models without business direction. Everyone works in silos. They guess what they need to build, get it wrong, then waste time and money redoing everything.
When planning meetings do happen, they’re typically run by tech people, not business people. The conversation starts with “let’s build a data warehouse” instead of “let’s make more money.” These are completely different starting points.
The average $3 million project costs companies $5.87 million when they use poor requirements practices. That’s a $2.24 million premium for skipping the planning phase.
Why Teams Skip the Definition Phase
Planning feels slow and painful. It’s time-consuming. The temptation to rush to build is always there.
We all want to deliver. Leadership applies pressure. Tech teams get impatient. So everyone skips the definition phase even though they know it will cost them later.
But here’s the reality: defects caused by poor requirements consume up to 85% of project rework costs. Fixing an error in the requirements phase is exponentially cheaper than fixing it after you’ve built and tested the product.
Mistake #2: The Language Barrier Between Tech and Business
Tech people use IT jargon: APIs, ETL, MCP, MFA.
Commercial people use business jargon: CAC, CLV, CTA, KPI.
They talk two different languages and struggle to understand each other. But the problem goes deeper than jargon.
Tech teams focus on building things. Business teams focus on outcomes. When a business person says “we need to increase customer lifetime value by 20%,” the tech team often rushes to build before making sure it’s all defined first.
Research shows that most failing data warehouse and BI projects are delayed or go over budget due to poor communication of requirements between management and IT teams. Various studies report a 50 to 60 percent failure rate for data warehouse implementations.
The Solution: Make It Real Before You Build
Creating ideal user journeys or wireframes of what users will see helps make it real for people. You need both perspectives working together to ensure projects deliver commercial benefit while being secure and robust.
More than 50% of data warehouses fail because projects don’t meet user requirements. The most common cause? Companies do a fine job of technically building a data warehouse, but the system doesn’t help solve any of their business needs.
Mistake #3: The ROI Calculation Nobody Wants to Do
Calculating ROI is really hard. That’s why people avoid it.
The investment part is easy. But estimating the return can be incredibly difficult. Some benefits are straightforward: increased revenue and cost savings. Other benefits are harder to quantify: reduced risk or improved data quality.
Teams typically focus on the easy-to-measure stuff and hope the rest is understood. Projects fail due to vague goals or lack of alignment with tangible business outcomes.
A Practical Reframe
Instead of “reduced risk,” estimate the business impact in pounds if you lost your biggest customer. Turn intangible benefits into concrete scenarios.
You can provide calculators to help break it down and make it easier. Clear business objectives are measurable, and this activity is critical since management will have to justify the expenditure once the project is completed.
Mistake #4: Skipping the Technology Audit
You can’t build on a foundation you don’t understand.
When teams skip the technology audit—failing to map out current data sources, data flows, and existing infrastructure—they create cascading failures throughout warehouse architecture and AI model development.
92.7% of executives identify data as the most significant barrier to successful AI implementation. And 99% of AI and ML projects encounter data quality issues.
Nearly half (48%) of the data migrated into businesses’ data warehouses requires cleaning before it becomes useful.
That’s the cost of not knowing what you have before you start building.
Mistake #5: Not Understanding Your Users
More than 50% of data warehouses fail because projects don’t meet user requirements.
The most common cause? Companies do a fine job of technically building a data warehouse, but the system doesn’t help solve any of their business needs.
When you don’t understand users well enough, you build technically sound but practically useless systems.
Research shows that most failing data warehouse and BI projects are delayed or go over budget due to poor communication of requirements between management and IT teams. Various studies report a 50 to 60 percent failure rate for data warehouse implementations.
The solution: spend time with actual users before writing a single line of code. Understand their workflows, pain points, and what success looks like from their perspective.
Mistake #6: Failing to Link Data and AI to Business Strategy
Data and AI initiatives that aren’t connected to the business model or business strategy rarely deliver value.
Projects fail due to vague goals or lack of alignment with tangible business outcomes. Teams build impressive technical solutions that don’t move the needle on what actually matters to the business.
You need to understand what business levers can be pushed or pulled to make the business go faster. Without this connection, you’re just building technology for technology’s sake.
Clear business objectives are measurable, and this activity is critical since management will have to justify the expenditure once the project is completed.
Mistake #7: Not Bringing Tech and Commercial Teams Together
For five straight years, executives report that cultural challenges—not technology challenges—represent the biggest impediment to successful adoption of data initiatives.
When tech people and commercial people don’t work together, you get one of two outcomes: technically secure systems that deliver no commercial benefit, or business ideas that can’t be implemented safely.
You need both perspectives working together to ensure projects deliver commercial benefit while being secure and robust.
The failure rate for analytically immature organizations is around 90%. For mature organizations, it’s approximately 40%. That massive gap comes from breaking down silos and creating genuine collaboration between technical and business teams.
Mistake #8: The Values Misalignment You Don’t See Coming
Many organizations are purpose-led, B Corp certified, and focused on environmental values. But when it comes to AI usage, people use AI frivolously—creating memes without realizing the energy cost.
A single query to an AI-powered chatbot uses up to ten times as much energy as an old-fashioned Google search. Generative AI systems use 33 times more energy to complete a task than traditional software.
The wake-up call usually comes when someone uses AI for something the organization didn’t expect. The business values don’t line up with AI usage.
Define Your AI Philosophy First
Start with your values and philosophies. Create your own rules for what you can use AI for and what you can’t. Do this by team and by department in a personalized way for your business.
Don’t just copy what other people are doing. Align your data usage principles and AI philosophies to your organizational values before deployment.
Mistake #9: Not Defining KPIs and Success Metrics
Not getting specific enough with KPIs is one of the most common planning failures.
If you can’t measure it, you can’t manage it. And if you haven’t defined what success looks like before you start building, you’ll never know if you’ve achieved it.
Studies show 70% of digital transformation projects fail. And 70% of those failures are due to issues with requirements. That means 49% of digital transformation projects fail due to requirements alone.
Get specific about KPIs upfront. Define exactly what metrics will move, by how much, and by when. Make success measurable before you write a single line of code.
Mistake #10: The Maintenance Crisis After Deployment
Deployment is just the beginning, not the end.
Problems come from new people getting involved. When everything wasn’t defined properly at the beginning, systems aren’t maintained in a logical way. New people come on board and change things in unexpected ways.
83% of IT decision makers aren’t fully satisfied with their data warehousing initiatives. And 42% of existing data management and warehousing processes that could be automated are currently being done manually, consuming valuable time, resources, and money.
If you have everything defined well in the beginning and properly documented, new people can come on board without creating infrastructure drift.
The Pattern We Keep Seeing
Studies show 70% of digital transformation projects fail. And 70% of those failures are due to issues with requirements. That means 49% of digital transformation projects fail due to requirements alone.
The pattern is clear across every mistake:
- Not defining requirements well enough
- Not getting specific enough with KPIs
- Not understanding what business levers can be pushed or pulled
- Not doing a technology audit
- Not mapping out current data sources or data flows
- Not understanding users well enough
- Not linking data and AI to the business model or strategy
- Not bringing together tech people and commercial people
- Not calculating ROI upfront
- Not defining data usage principles and AI philosophies
The end result? Wasted time and money. A load of frustrated people.
What Actually Works
The failure rate for analytically immature organizations is around 90%. For mature organizations, it’s approximately 40%. That massive gap comes from defining everything properly at the beginning.
You need detailed requirements defined before you start work. This makes you much more likely to deliver right first time—quicker and cheaper too.
Resist the temptation to rush to build. Planning may feel slow and painful, but skipping it will slow everything down eventually.
For five straight years, executives report that cultural challenges—not technology challenges—represent the biggest impediment to successful adoption of data initiatives.
The technical work is the easy part. Getting the planning, alignment, and principles right is what separates successful projects from the 80% that fail.