2 days ago
Turn on all the lights: why your AI fails without the right data
In our latest episode of Lexicon, we speak with Justin Graham, Director of the Innovation Solutions Center at Barge Design Solutions, and Jake Dein, PE, Technology Solutions Developer.
Together, they demystify a crucial challenge in the age of artificial intelligence: why so many organizations are failing to derive value from AI, and how to address this issue. Their message is clear: before you bring in AI, ensure your data is in order.
Also, subscribe to IE+ for premium insights and exclusive content!
Jake opens the conversation with a powerful metaphor called the "streetlight effect." As he explains, this is a cognitive bias where people tend to look for solutions where it's easiest, rather than where it's most effective.
'It's where we tend to use data that is easily accessible to us, but not necessarily the best to answer engineering questions,' Jake explains.
'We borrowed the term from an old parable about a person looking for their keys at night under a streetlight—not because that's where they dropped them, but because that's where it's easiest to look,' he added
This behavior is widespread across industries, particularly as companies rush to adopt AI without first connecting the dots between their siloed systems.
'AI, to be really effective, needs to leverage all the relevant data, not just what's easiest to find,' he explained.
The episode's title comes from a core idea that Justin and Jake repeatedly return to: 'turning on all the lights' before diving into AI or advanced analytics. In practice, this means integrating and organizing fragmented datasets across an organization to create a single, trustworthy source of insight.
'Creating visibility into your business systems—integrating siloed data from HR, finance, customer service, design specs, project archives—is what turns AI from a gimmick into a true decision-making tool,' Justin says.
He illustrates the point with a vivid example: Barge's relocation project.
'We had a warehouse full of 70 years of project records—paper boxes stacked to the ceiling. No one knew what was in them. Without visibility, without a system, even if the answer is in there, it's useless if you can't find it. That's no streetlights, just darkness.'
One of the podcast's most actionable takeaways is deceptively simple: don't start with the technology. Start with a clearly defined problem.
'We see a lack of clearly defined problems all the time,' Jake explains. 'A problem well stated is a problem half solved.'
He compares poorly scoped AI projects to going to a mechanic and asking for a front-end alignment when your car pulls to the left, without first checking if that's the issue.
'A good mechanic will confirm the real problem before fixing it. Otherwise, you're just wasting time. The same is true with AI,' he added.
The highlight of the conversation is a compelling case study: how Barge helped a water utility save nearly $1 million by modernizing workflows and integrating systems.
'It started with a client goal—improving customer satisfaction. But instead of jumping to customer-facing solutions, we looked internally for inefficiencies and silos,' Justin says.
The key wasn't just technology; it was fixing how information flowed.
'They had great systems: inventory, purchasing, HR, customer service, SCADA. But none of them were talking to each other,' Justin explains. 'So we integrated them, reduced workflow friction, and the impact was huge.'
'That kind of return—10X, 50X—isn't magic. It's the payoff of turning on all the lights,' he explained.
The duo is candid about the less glamorous side of AI readiness, data hygiene.
'Good data hygiene isn't a delay—it's an accelerator,' Justin emphasizes. 'It's often overlooked because it's hard and not very sexy. But it's the foundation.'
Barge even built a simple archive search tool for internal use that ended up saving $350,000 per year.
'That's just from reducing the time people spend looking for information,' he notes. 'It's not flashy, but it's incredibly impactful,' he explained.
Both guests are skeptical of AI implementations done for trendiness rather than utility.
'We've seen plenty of flashy AI solutions that don't solve real problems,' Jake says. 'Without proper integration, they under-deliver—or worse, mislead.'
He shares an example of a health and safety plan that they developed.
'It used to take 8–10 hours to write a good plan. With integrated systems and generative AI, we can now complete a draft in 10–15 minutes. That's the power of doing it right.'
But they stress the human role isn't going away.
'We're big on the 'human in the loop' idea,' Justin says. 'AI should augment decision-making, not replace it. The goal is to free people to do higher-level thinking, not to automate away responsibility.'
As the conversation wraps up, both guests offer practical advice for organizations excited about AI but unsure of where to begin. 'Just take action,' Justin urges. 'Start with a specific point of friction. Solve one real problem. That's how you gain clarity.'
'Ask yourself: how do you know what you think you know about your business? Can you point to the data that backs it up? That question alone reveals gaps you need to address,' Jake explained.
'Every company should be doing that—AI or not. Avoiding the streetlight effect is just good practice,' he concluded.
As for what's next, Jake is optimistic, 'AI's need for integrated data might finally push industries to break down silos. But it will take critical thinking and focus. APIs exist—it's the people and processes that need to catch up.'
Justin agrees, emphasizing balance, 'security and access need to be aligned. That's the art—pulling the right data without exposing everything.'
Justin and Jake don't just preach best practices; they live them, showing how thoughtful data integration leads to real savings, smarter workflows, and better decisions.
'Turn on all the lights,' Jake says in closing. 'You'll build better, more effective solutions.'
