The Top 5 AI Mistakes Businesses Make and How to Avoid Them

The Core Argument

Guided by host Sanjay Saini, Professional Scrum Trainer Florin Manolescu warned that the industry is experiencing a massive disconnect between delivery speed and actual business value. Caught up in the hype, organizations are treating AI simply as a feature factory, churning out outputs without measuring if they actually improve the customer's life. Florin argued that businesses are repeating the exact same "framework-washing" mistakes they made during the Agile boom 25 years ago—adopting the technology just because competitors are doing it, rather than utilizing it to solve specific customer problems through empirical experiments.

“I feel that people think AI will resolve all our delivery problems, like Agile was the promise back then. History is repeating.” Florin Manolescu (07:50)

Detailed Answers

1. Could you give a brief intro about yourself and your background?

Florin Manolescu is a veteran IT leader based in Bucharest, Romania, with nearly 30 years of industry experience. He serves as a Project Manager and a Professional Scrum Trainer with Scrum.org. Additionally, he leads an active Agile community in Romania (Agile Minds) and works as a certified training partner with Sense & Respond Learning, focusing heavily on shifting product teams from an output-driven mindset to an outcome-driven one.

2. Why did you choose the topic of AI output versus customer outcome?

Florin observed a massive hype cycle where developers and businesses are pushing AI features faster than ever before. However, almost no one is pausing to ask: "What's in it for the customer?" He chose this topic because creating an AI feature (an output) is useless if it doesn't result in a measurable positive change in customer behavior (an outcome), such as better user satisfaction or higher success rates.

3. Are we repeating the same mistakes with AI that we made with Agile?

Yes. Florin noted that just as companies adopted Agile 25 years ago under the false assumption it would magically fix all delivery bottlenecks, they are doing the exact same thing with AI. He warned that organizations are mandate-adopting AI because "the competitors are doing it," which is the worst possible motivation. As a result, teams are spending more time fixing AI hallucinations and poorly generated code than they are actually innovating.

4. How should we measure customer outcomes when using AI?

Florin advised sticking to the basics. The KPIs for measuring success shouldn't change just because AI is involved. Organizations must ask: "Did using AI to build this feature genuinely change the customer's life for the better?" If there is no measurable difference in retention, revenue per customer, or user satisfaction, then the AI implementation was a waste of expensive processing tokens.

5. What are some successful use cases of AI improving customer outcomes?

Florin highlighted IKEA's highly successful AI implementation. Instead of using AI customer support bots to simply lay off their human support staff, IKEA retrained those human workers to become high-level design consultants. The AI handled the repetitive support queries, while the newly upskilled humans provided complex room design consulting, identifying a massive new revenue stream that generated $1 billion in a single year.

6. Is reducing headcount the only ROI organizations are tracking with AI?

Unfortunately, yes for many companies. Florin and Sanjay discussed how reducing headcount is the absolute easiest way for a business to show immediate financial ROI from an AI tool. Rather than doing the hard work of strategizing how AI can create new value or optimize existing workflows, many organizations take the lazy route of equating AI success entirely with payroll reduction.

7. What tools do you recommend for Scrum Masters and Product Owners?

Florin is highly pragmatic about tool adoption. He frequently uses ChatGPT and Gemini for standard queries, and Perplexity for deep, accurate research. For collaborative team learning and synthesis, he highly recommends Google's NotebookLM. For his development teams and Business Analysts, Claude is the primary ecosystem. However, he stressed that the tool itself doesn't matter; what matters is using the approved internal tools with heavy critical thinking.

8. How should an organization start its AI transformation journey correctly?

Florin outlined a strict empirical approach. First, explicitly define the business problem you are trying to solve. Second, establish a baseline of where you currently stand before introducing any AI. Third, pick clear outcome metrics. Fourth, run small, isolated experiments and learn fast. The biggest mistake companies make is skipping the "define the problem" phase entirely and rushing directly into large-scale, thoughtless delivery.

Don't build AI features just because your competitors are. Learn how to define genuine business problems, establish baselines, and measure true customer outcomes in our Artificial Intelligence training program.