Many teams want AI to take over tasks, but Carlos Almeida thinks that automation comes later. Before you hand over the work, you need to understand it, control it, and measure it.

Carlos is the CEO and co-founder of Optave, where they’re building AI copilots and autopilots for customer support. In this conversation, he shares why the real work isn’t in generating responses, but in building the systems around them: the guardrails, escalation triggers, and insights that make automation trustworthy.

We talk about what makes AI useful in day-to-day operations, why trust and visibility are more important than speed, and how focusing on metrics and manager tooling led Optave to better outcomes than competitors chasing 90% automation.

🔑 What You’ll Learn in This Episode

📌 For AI Product Builders & Founders:

• ✅ Why automation only works after you’ve earned trust
• ✅ What makes user interfaces more important than model capabilities
• ✅ How to structure real-time feedback loops for humans and machines

📌 For Business & Ops Leaders:

• ✅ How to spot empty AI promises (and what to ask instead)
• ✅ Why control systems will be the next differentiator
• ✅ How to measure AI success without hiding behind accuracy

🔗 Resources & Links

• 🤝 Connect with Carlos on LinkedIn
• 🤖 Learn more about Optave
• 📩 Subscribe to the Artificial Insights newsletter for episode summaries

🚀 Found the episode valuable? Share it with someone building trust into AI systems

In this short, Mike Kirkup shares his experiences being on the frontlines of different tech revolutions. This AI revolution? It feels different.

Why?

Because everyone's working on it. Everyone's using it (or it's getting there). And it's forcing business leaders to question entire business functions and how they do things.

🔗 Resources & Links

• 🎧 Listen to the full interview with Mike: https://rss.com/podcasts/manaryhaus/2061508/
• 📩 Subscribe to the Artificial Insights newsletter for new episodes and reflections: https://manary.haus/podcast/#haus

🚀 Found this useful? Share it with someone who’s trying to think clearly about AI’s future.

At EnPowered, Mike Kirkup ended up running the financing arm, manually collecting and validating documents... until he couldn’t take it anymore.

The internal tooling he built became Arlo, a product now used by lenders to automate and augment document workflows with AI.

In this episode, Daniel and Mike talk about what makes using AI different from old school automation. Yes, Arlo automates mundane tasks, but it also asks follow-up questions and flags mismatches humans miss.

Mike explains how Arlo became goal-oriented rather than rules-based. That shift not only made the AI smarter, it pushed users to rethink what they were actually trying to accomplish and how they were doing it.

Listen in to learn about how to truly be AI-first when designing solutions, building trust in AI despite its unpredictability, and how a simple question—“Why do you even ask for that document?”—can change how a whole industry works.

🔑 What You’ll Learn in This Episode:

• ✅ What it's like to demo AI that asks unpredictable questions
• ✅ The courage it takes to give AI more room to operate
• ✅ How AI nudges companies to reexamine outdated workflows
• ✅ Why domain expertise is critical to building useful AI tools
• ✅ Why you shouldn't get a PhD to break into AI

🔗 Resources & Links:

• 🤝 Connect with Mike on LinkedIn
• 📓 Learn more about Arlo
• 📩 Subscribe to the Artificial Insights newsletter for summaries and takeaways

🚀 Enjoyed this episode? Leave a review and share it with someone building in AI!

Why do we keep getting AI wrong?

It’s not because it’s too intelligent. It’s because we’re bad at extrapolating what intelligence makes possible.

In this bonus short, Daniel shares a conversation with Chyngyz Dzhumanazarov, co-founder and CEO of Kodif, who draws a striking parallel: when electricity was invented, no one predicted it would lead to television, or that actors would build careers from it.

People worried about job loss in coal.

They missed the entertainment industry entirely.

The same thing is happening with AI. We’re focused on what it might replace, not what it might enable.

🔑 What You’ll Hear in This Episode

• ✅ What electricity, television, and the entertainment industry can teach us about new technologies
• ✅ Why we overestimate disruption and underestimate emergence
• ✅ A reminder that mystery isn’t the same as incomprehensibility

🔗 Resources & Links

• 🎧 Listen to the full interview with Chyngyz: https://rss.com/podcasts/manaryhaus/2041601/
• 📩 Subscribe to the Artificial Insights newsletter for new episodes and reflections: https://manary.haus/podcast/#haus

🚀 Found this useful? Share it with someone who’s trying to think clearly about AI’s future.

Will AI make human support teams obsolete? Kodif's answer: No, but humans will be managing and teaching AI agents. Kodif is building systems where AI agents resolve tickets, write help docs, and surface insights, all under the guidance of real people.

In this episode, Daniel sits down with Chyngyz Dzhumanazarov, CEO and Co-Founder of Kodif, and the first person from Kyrgyzstan to earn an MBA from Stanford. Chyngyz shares how Kodif augments instead of replaces human support, and how his non-technical perspective helped him see opportunities others overlooked.

They talk about what it means to build AI-native operations, how to test for real user pain, and the shift from software that people use to tools that people teach.

🔑 What You’ll Learn in This Episode

📌 For Founders & Product Builders:

• ✅ Why automation only scales when local teams can shape it
• ✅ How curiosity (and not knowing what’s “impossible”) opens up new ideas
• ✅ What agentic AI looks like in real-world support workflows

📌 For Business Leaders Adopting AI:

• ✅ How AI can augment (not replace) support roles
• ✅ Why your "non-technical" team can manage AI agents
• ✅ Why understanding process variation is key to deploying AI across markets

🔗 Resources & Links

• 🤝 Connect with Chyngyz on LinkedIn
• 🤖 Learn more about Kodif.ai
• 📩 Subscribe to the Artificial Insights newsletter for summaries and takeaways

🚀 Enjoyed this episode? Leave a review and share it with someone building in AI or ops!

When Tair Asim helped build Sage, an AI-powered networking tool for builders, the team had the tech working... but building something usable is not the same as building something people will use.

In this clip, Tair reflects on what made Sage successful inside the buildspace community, and what he learned about trust, traction, and what early users need.

He shares how the product changed from a long onboarding process that only worked because early users believed in the brand, to something more approachable as the product matured.

It’s a rare inside look at the early-stage moments most people skip over when telling their "how we built this" story.

💡What you'll hear in this episode:

• Why Sage worked best when users already knew and trusted the team
• How the buildspace brand helped users stick around through friction
• What it felt like to build something people showed up for even when it wasn’t polished
• How starting with a narrow, familiar niche helped the team grow beyond it

What happens when you combine brand, community, and AI, not just to automate tasks, but to create real human connection?

In this episode, Daniel sits down with Tair Asim, creator of Sage, the AI-powered networking tool that helped thousands of builders connect through natural language, and Head of Product at Sync, where he’s now leading vertical integration in AI-powered video editing.

Tair shares lessons from both worlds: why Sage worked because of the community behind it (not just the tech), how vertical integration creates defensibility in a fast-moving AI market, and why focusing narrowly on a niche often leads to the biggest breakthroughs.

🔑 What You’ll Learn in This Episode

📌 For Founders & Product Builders:

• ✅ How brand, community, and AI combined to make Sage stand out
• ✅ Why vertical integration matters when foundational models evolve fast
• ✅ How to spot the difference between a feature and a defensible business

📌 For Business Leaders Adopting AI:

• ✅ What spontaneous, unstructured AI interactions can unlock in your product
• ✅ How to prioritize quality inputs to get high-value AI outputs
• ✅ Why narrowing your niche can be the best way to scale later

🔗 Resources & Links

• 👨‍💻 Connect with Tair on X: @tair
• 🎥 Check out Sync’s latest product launches: sync.so
• 📩 Subscribe to the Artificial Insights newsletter for episode recaps and key takeaways

🚀 Love the show? Leave a review and share it with someone building in AI!

When's the right time to partner with sales and marketing? Alex's journey was to build and validate a prototype with a small group of customers, then bring in distribution experts. If you can build it, you can validate it! But what got you here might not get you there.

Are you more comfortable asking your colleagues a question, or ChatGPT? GovAI found that people preferred using AI over connecting with each other to ask questions. They liked the ease, speed, and anonymity compared to asking each other.

In this episode, Daniel sat down with Alex Millar, co-founder of GovAI and former co-founder of Bonfire, which he scaled and sold for $108M after working with hundreds of public sector teams. Now he’s building AI tools for a sector known for strict requirements and slow adoption—by focusing on usage first, and letting that usage show the path to revenue.

🔑 What You’ll Learn in This Episode

📌 For Founders & Product Builders:

• ✅ Why starting with free usage—before charging—can unlock adoption in high-trust environments
• ✅ How failed queries can reveal what to build next, and which features your users are asking for without saying it
• ✅ What it looks like to serve a niche so well that pricing, onboarding, and training become part of the product

📌 For Business Leaders Adopting AI:

• ✅ Why a flat-fee pricing model works better than per-seat licenses in the public sector
• ✅ Why you should put humans in the loop as a resource the AI can recommend, not just an override for edge cases
• ✅ How to roll out AI tools with training and change management—not just logins

🔗 Resources & Links

• 👨🏻‍🦲 Connect with Alex!
• 🌐 Learn more about GovAI
• 📩 Subscribe to the Artificial Insights newsletter for episode summaries and founder takeaways

🚀 Love the show? Leave a review and share it with someone building in AI!

When deciding what product to pursue and build a business around, it turns out you can't just ask ChatGPT for advice. Join Daniel and Bijan in this bonus short as they discuss how Merchkit got to where it is now, and what it really means to iterate and innovate.

Key takeaways:

• ✅ Start with a large market: the more potential customers, the more ideas you can try before running out of people to try them with.
• ✅ Generative AI will disrupt a lot, so pick the biggest problems that excite you. You'll spend a lot of time working on them
• ✅ To find something worth building, you need to dig into the specific workflows and day-to-day of your customer and keep asking, "why didn't that work for you?"
• ✅ Knowing what the boundaries of what AI are and what it can and can't do will allow you to create the bridge for that workflow gap for your customer.