Eight in Ten AI Projects Fail for the Same Reason · Laura M Gonzalez, Zeop

What we talked about.

In this episode of the Sunny Ray Show, host Sunny Ray talks with Laura M Gonzalez of Zeop about why roughly eight in ten AI projects fail. Gonzalez argues the cause is rarely the technology and almost always the people and processes around it. She explains how Zeop builds an operational twin of a company so that the CTO, CFO and CEO can finally speak the same language about what a project is for. Drawing on experience across five countries, C-level operating roles and venture investing, she describes how automating chaos only amplifies it, why data is a symptom of broken processes, and how one client had 364 documented steps against 3,200 real ones. She shares a logistics case where a procurement bottleneck, the stick in the wheel, sat in plain sight, describes what clients receive after six weeks, and explains how her first customers found her. She also says which companies should wait before adopting this kind of solution.

Laura M Gonzalez of Zeop says AI projects fail on people and process, not algorithms, and automating chaos only amplifies it.

The questions, and the answers.

Eight in 10 AI projects fail for the same reason. What is the reason, and does that number hold up?

It holds up. We think about the technology and not about the people and the processes being used. This is especially true for traditional companies that were not built AI and data first, and that carry knowledge and years of stacked decisions.

What problem is Zeop solving?

Communication is simple in a room of five people, but it gets complicated at a hundred. As a company grows, the knowledge, tools and workflows spread out, and people, operations, technology and data become siloed. At Zeop, we generate an operational twin of the company so leaders can see the full picture.

What does translating between the engineer and the chief financial officer sound like in a room?

Each area speaks its own language. The CTO asks whether we can build it, the CFO asks where the money is, and the CEO asks how it changes the way of work. What they lack is what it is for. We help them say: we will build this, it cuts X hours, and that moves margin by X percent.

What did the companies where your $50 million in investments worked have in common?

They had an amazing team, but the key was focusing on results rather than the next fundraising round. They were convinced results would bring the clients, the learning and the money. For them, the money followed the results, not the other way around.

What did the failures look like up close, and what question did you keep asking?

I kept asking what this is for and who is in charge of it. Data is just a symptom. If the production band of processes is broken or dusty, your data comes out broken too. You can automate chaos and only amplify it. Sometimes you just need to fix the process first.

How does a company end up with 364 documented steps but 3,200 real ones?

It is normal, because organizations are made of humans, systems and data. We quantified the black box at 9.1 times more interactions than what is officially documented. People download from the ERP and CRM, rework it in Excel, and pass it on, and every person changes it again.

How did nobody see the 3 million euros sitting in procurement at the logistics company?

It is like a painting at home: after three weeks you stop seeing it. They focused on trucks and warehouses, the core business. Bottlenecks are tied together, and procurement was a support area, not client facing. We traced the relationships and found the stick in the wheel.

Which companies should wait before adopting Zeop?

A small company that can already trace its steps, operations and data will not get much value from us. Also, anyone searching for a magic pill is not approaching it the right way. Adopting without really understanding your operation is not a good moment to start.