Awesome Chat: Why AI Works in Demos but Fails in Real Life with Victor Coimbra of Artefact
Why AI Works in Demos but Fails in Real Life
AwesomeChat with Victor Coimbra of Artefact
Artificial intelligence can write emails, generate graphics, summarize meetings, analyze data, and even perform tasks on our behalf. But there's a big difference between watching an impressive demonstration and building a system that actually improves how a business operates.
On this edition of AwesomeChat, Michael Sorg welcomes Victor Coimbra, Partner and CTO Americas at Artefact, for a conversation about why AI doesn't always deliver the return businesses expect—and what needs to change.
The AI Productivity Paradox
Many people already believe AI is making their work better. But what happens when companies look at their financial statements?
Victor shares figures drawn from Artefact's conversations with major businesses: 83% of employees surveyed or consulted reported that AI helped their work, while only 3% of CFOs reported seeing an impact on their company's profit and loss.
Those figures, as presented by Victor, illustrate the central challenge facing businesses today: people may feel more productive without that improvement necessarily translating into measurable financial results.
The problem isn't always that the AI doesn't work. Sometimes businesses haven't changed their operations enough to take advantage of it.
Stop Thinking About AI as Just a Productivity Tool
Much of today's AI conversation focuses on individual efficiency: writing an email faster, preparing a report, or summarizing a document.
Victor argues that businesses need to look beyond those incremental savings.
He discusses how AI can become part of the core business process, helping organizations rethink how they serve customers, evaluate opportunities, and potentially generate new revenue.
He also points out another overlooked concern: AI isn't necessarily becoming less expensive overall.
While prices per token may decline, increasingly sophisticated reasoning systems can perform far more processing to complete tasks. Organizations may find themselves consuming significantly more AI resources even while individual operations become cheaper.
AI Agents Change the Relationship
The conversation turns toward one of the biggest developments in artificial intelligence: AI agents.
Traditional chatbots generally respond to instructions and work through problems alongside the user. Agents can be given a goal and then perform multiple steps to accomplish it.
That changes the relationship from collaboration to delegation.
Victor describes the emergence of hybrid organizations where human workers and AI systems contribute different kinds of capabilities.
But autonomous AI introduces new responsibilities. When agents have access to company systems, passwords, files, or sensitive information, businesses must decide what those agents are allowed to do and how their actions will be supervised.
As Sorg points out, canceling an AI subscription doesn't undo the damage if an automated system has already made a serious mistake.
Your Biggest AI Asset Might Be Inside Your Head
Perhaps the most useful part of the conversation focuses on the difference between data and context.
A business may have decades of invoices, spreadsheets, client records, and reports. But those documents don't necessarily explain why certain decisions get made.
Why does an experienced employee know to call one particular person when a problem occurs? Why does a video producer immediately recognize a technical issue that a newer operator might miss? How does a team decide which exceptions matter?
Much of that knowledge exists only through experience.
Victor argues that these undocumented details are a major obstacle to effective AI deployment.
For Sorg, who has spent decades producing live events and video projects, the challenge is familiar: How do you transfer 30 years of experience and troubleshooting into something another person—or an AI system—can use?
Victor's recommendation is straightforward: start capturing the context behind decisions.
Document the mistakes, the workarounds, the reasoning, and the processes. These records can become increasingly useful when building automated workflows.
Will AI Take Our Jobs?
The future of employment remains one of the largest concerns surrounding artificial intelligence.
Victor draws comparisons to previous industrial revolutions, arguing that major technology shifts have historically introduced new industries and opportunities alongside changes to existing work.
He expects humans to spend more time on creativity, judgment, and planning, while AI takes on more analytical and data-intensive tasks.
It's an optimistic perspective, but one that recognizes the difficulty of changing how people work and how organizations are structured.
The conversation also explores entrepreneurship and the possibility that AI could make it easier for people to build businesses that previously required far more resources.
AI Adoption Around the World
With Artefact operating internationally and Victor joining the conversation from Brazil, Sorg also asks how different regions approach artificial intelligence.
Victor describes the United States as particularly focused on innovation and moving quickly, while Europe places greater emphasis on regulation, responsibility, and long-term societal impact.
He suggests those approaches can complement each other as AI develops.
The discussion provides a preview of the Adopt AI Summit, coming to the Grand Palais in Paris on December 3–4, 2026, where business leaders, technologists, and policymakers will explore practical AI adoption.
One Thing You Can Do Right Now
Victor leaves AwesomeCast listeners with a simple piece of advice:
Capture your context.
Start documenting how you make decisions, how you approach problems, and what you've learned from experience.
That knowledge could become the foundation for the next generation of AI-powered tools in your business.
Whether you're running a major enterprise, managing a small production company, or simply looking for ways to make AI more useful in your own work, the message is the same: successful automation starts with understanding how the work actually gets done.
Watch or listen to the full conversation on AwesomeChat, hosted by Michael Sorg.
AwesomeCast: https://awesomecast.com/
Guest: Victor Coimbra, Partner & CTO Americas at Artefact
https://www.artefact.com/
Adopt AI Summit — December 3–4, 2026
https://adoptai.artefact.com/
Support independent tech conversations from Pittsburgh:
https://www.patreon.com/awesomecast
