--- type: podcast-episode title: "How does a startup actually become AI-native?" url: https://thelastfounder.com/episodes/tMxZKi2qAOM/ video: https://www.youtube.com/watch?v=tMxZKi2qAOM duration: "45 min" participants: "Daniel Maman / Corgi.ai" topics: ["AI adoption", "Building companies"] language: en content: editorial notes, not a transcript --- # How does a startup actually become AI-native? Daniel Maman / Corgi.ai Adding a coding assistant to R&D is a start, not an organizational transformation. Daniel Maman, co-founder of Corgi.ai, joins Ron Gross to examine the less obvious work of becoming AI-native: changing products, internal processes, customer delivery, and the way a founder delegates. ## AI adoption does not move at the same speed everywhere Daniel separates three problems that are often treated as one: putting AI into a product, using it to build that product, and using it across the company. R&D has an advantage. Developers already work in environments where outputs can be tested, and coding tools give them a relatively clear starting point. Other teams face a different challenge. Sales, finance, marketing, and customer success do not automatically become AI-native because engineers write code faster. Their users may not be technical, their workflows are less standardized, and the useful result is often harder to define. Daniel describes Corgi's work as helping companies bridge the gap between what people need, what they want, and what they can comfortably use. ## When agents become customers The conversation also looks outward. A business may need to serve software agents as well as people. That changes product interfaces and how a company is discovered, evaluated, and bought from. An interface that works for a person browsing a website may not be the right interface for an agent acting on someone's behalf. Daniel connects this shift to his experience with automated activity on the web. The important distinction is no longer simply human versus bot: some automated actors can now be legitimate customers. Founders need to consider what an agent can understand and do, without assuming every automated request should be welcomed. ## The boundary between software and services AI makes it easier to deliver work that once required a person using software. That blurs the line between a product and a service. Daniel and Ron discuss tailored workflows, productized services, and forward-deployed engineers who work close to customers to understand how a system should fit into their operations. The point is not that every company should become a consultancy. It is that delivering an outcome may require learning the customer's process, not merely handing them a tool. The business model and the implementation work need to match what the customer is actually buying. ## Building together, rather than alone Ron brings the discussion back to the Builders community: people learning by shipping, working on their own projects, and helping one another. They consider multiplayer agents and collaborative workflows, where AI supports a group rather than a single isolated user. Daniel sees a reason for these communities beyond access to tools. People want to close knowledge gaps, compare experiences, and have others alongside them while they build. Collaboration still has value even when producing a first version becomes easier. ## Can the founder hand the company to an AI CEO? Ron describes the appeal of participating more like a board member: setting direction without carrying every operating task. Daniel draws a distinction between giving an agent a narrow assignment and asking it to own an abstract business outcome. The more senior the role, the more judgment is needed to decide which work matters. His assessment in the conversation is that AI can take on well-defined tasks, but the founder still needs to orchestrate the work and decide what will move the business forward. A system that can execute instructions is not necessarily a system that can own the company. ## Learn the principles, not just the latest tool Daniel's practical advice is to understand the model and the conventions around it, rather than treating a particular agent framework as the source of all the capability. For technical builders, creating a small agent can make that distinction concrete. The closing thread is about staying grounded: try the tools, notice what they can actually do, and avoid confusing a compelling demonstration with a working business. The episode leaves the founder with a broader question than which coding tool to choose: what has to change across the company for AI to create useful results? ## Topics in the conversation 00:19 Meet Daniel and Corgi 06:20 AI in the product, R&D, and the rest of the company 15:12 Products, services, and customer delivery 20:34 Multiplayer agents and building together 29:00 Founderless companies and the AI CEO question 37:42 Models, tools, and learning by doing Source: https://www.youtube.com/watch?v=tMxZKi2qAOM Guest profile: https://thelastfounder.com/guests/daniel-maman/ Edited notes, not a verbatim transcript. [Join the WhatsApp community](https://chat.whatsapp.com/DgKpG63438y8N7D3pHJ57t) ## Links - Guest profile: https://thelastfounder.com/guests/daniel-maman/index.md - Watch on YouTube: https://www.youtube.com/watch?v=tMxZKi2qAOM - HTML page: https://thelastfounder.com/episodes/tMxZKi2qAOM/ - All episodes (JSON): https://thelastfounder.com/episodes.json - Site index: https://thelastfounder.com/index.md