The difference between slapping a chatbot onto your existing product and building a truly AI-first company comes down to one fundamental question: When AI models get better, are you happy or sad? If you're happy, you're positioned for explosive growth. If you're sad, you're in trouble.
This insight came from a powerful panel discussion at HubSpot for Startups' annual AiSummit in San Francisco, featuring founders who are designing AI into their company DNA from day one. The conversation revealed how today's most innovative startups are fundamentally rethinking everything from hiring practices to organizational structures to product development.
The panel brought together Siqi Chen, CEO and founder of Runway Financial; Kashish Gupta of Hightouch; Varun Anand of Clay; and Elias Torres from Agency. Moderated by Adam Coccari from HubSpot Ventures, these founders shared hard-won lessons about what it really means to build an AI-native company.
Rethinking Company Structure: Why Traditional Scaling Is Broken
Elias Torres opened with a bold declaration that set the tone for the entire discussion. After building his fifth company and experiencing the challenges of managing 800 employees at Drift, he's taking a radically different approach with Agency.
"I just don't want to ever do that again," Torres explained. "I don't want to be dealing with people, their problems, promotions. It's just a mess. What I want to do is solve for my customer and build the right product."
His solution? Build a billion-dollar revenue company with just 100 people instead of 1,000. At Agency, Torres has collapsed traditional roles into just three key hires: a lawyer, an MBA, and a head of sales. These three people handle everything from finances and payroll to customer support and fundraising.
The approach isn't just theoretical - it's working in practice:
- Radical role consolidation: Traditional functions like HR, business development, and customer support are distributed across the small team
- Direct customer access: All customer communication flows through Torres' email, with full transparency across the company
- Universal authority: Any team member can respond to customers using Torres' authority, eliminating bottlenecks
- Public accountability: Daily company-wide meetings where issues are addressed openly, similar to NVIDIA's Jensen Huang's management style
The Rise of the GTM Engineer: Collapsing Sales Roles
Clay has pioneered one of the most innovative role transformations in the startup world: the GTM (Go-To-Market) Engineer. Varun Anand explained how this single role has replaced traditional sales development representatives, account executives, and sales engineers.
"We've collapsed salespeople, sales engineers, SDRs into just one role," Anand said. "They're very commercially minded, have the soft skills needed in sales, and are also very technical and product-minded."
What makes this approach revolutionary:
- Non-traditional backgrounds: Team members come from diverse backgrounds - founders, artists, biomedical engineers, bridge builders
- No commission structure: The entire team works without variable compensation, creating a collaborative rather than competitive culture
- Technical proficiency: Everyone uses Clay and other technical products to automate pipeline generation
- Curiosity-driven hiring: Common traits include curiosity, commercial mindset, scrappiness, and resourcefulness
The results speak for themselves. Clay's sales team is frequently praised by industry veterans as "the highest IQ sales team" and "the most collaborative sales team" they've ever encountered.
Hiring Former Founders: The Patient Approach to Building Elite Teams
Runway Financial has taken an equally bold approach to talent acquisition, focusing heavily on hiring former founders and senior executives willing to work as individual contributors.
Siqi Chen shared their methodology:
- Extreme patience: Runway hired a principal recruiter from Uber as one of their first employees and waited months to find the right people
- Unattainable standards: Their recruiter noted that their hiring standards exceeded even Uber's requirements
- Senior ICs: They hire VP-level engineers and executives who prefer building over managing
- Role fluidity: People transition between roles as needed - their VP of Finance started in customer success
"Anyone who's spent time in senior management knows that job sucks," Chen noted. "The best people just want to build, and we say you can have a lot of impact, you just have to build stuff."
Empowering Non-Traditional Roles with AI Tools
One of the most striking examples of AI enabling new capabilities came from Clay, where a sales team member shipped 6,000 lines of code in a single week, including implementing dark mode for their product.
This transformation is enabled by giving everyone in the company access to tools like Cursor, allowing non-engineers to build and ship features. As Anand explained, "The people who want to build now can build even if they're not an engineer."
Similar examples emerged across the panel:
- Hightouch: A solutions engineer used Replit to build a fully customized demo for a large airline client in just one day
- Agency: Team members handle customer communications directly through the founder's email with full authority
- Runway: Customer success managers transition to finance roles, and engineers take on customer-facing responsibilities
AI Product Strategy: Beyond the Chatbot Trap
All four founders agreed on one critical point: adding a chatbot to existing software isn't AI-first product development. As Kashish Gupta from Hightouch put it, "It's strapping on a chat box into your existing UI. I see a lot of SaaS tools just add a little chat box, and it's the AI feature. I think that's pretty not customer centric."
Instead, truly AI-native products require fundamental rethinking:
- Context awareness: Products that understand your company and situation without extensive setup
- Proactive intelligence: Systems that take action without being prompted
- Embedded decision-making: AI that makes decisions at the level of granularity that humans used to handle
- Explainable AI: Combining large language models with traditional machine learning for compliance and transparency
Siqi Chen offered a compelling example from ChatGPT: "The really first AI native product design feature wasn't the chat part. The thing that was cool is you have this conversation with the agent, but on the sidebar, you didn't have to name the chat. It just named it."
Organizational Design for the AI Era
The panel revealed several innovative approaches to organizing AI-first companies:
People and Brand Integration: Clay combined their people leader role with brand design, ensuring external and internal brand consistency.
Finance Team AI Policy: Hightouch instituted a rule that their finance team cannot block any AI software purchases, enabling rapid experimentation.
Hackathon Partnerships: Clay's external collaboration strategy involves bringing customer companies into their office for day-long building sessions, creating FOMO and accelerating adoption.
Public Decision Making: Agency's transparent communication model where all customer interactions are visible company-wide, enabling anyone to respond with founder authority.
Learning from AI Implementation Mistakes
The founders were candid about their failures and learning experiences:
Clay's Evolution: Started requiring JavaScript for complex workflows, then successfully used GPT-3 API to convert natural language to code in early 2023, but failed when trying to extend this to auto-filling company data - a capability that works well today.
Runway's Patience: Repeatedly tried AI features that didn't work 2-3 years ago, like having AI output JSON reliably, but kept experimenting as models improved.
Agency's Chatbot Lesson: Built a comprehensive customer health chatbot that could answer any question and provide recommendations, only to realize that expecting humans to regularly engage with chat interfaces doesn't work at scale.
The Future of AI-First Companies
The panel concluded with optimism about transformation possibilities. When asked if non-AI-first companies can become AI-native, the consensus was overwhelmingly positive.
Chen's test remains the gold standard: "When the models get better, are you happy or sad? If you're sad, you're not an AI first company. If you're happy, things are gonna be great for you."
Key strategies for transformation include:
- Identify high-agency employees: Find the 5-10 people in your company who think beyond incremental improvements
- Enable talent transition: Be prepared for talent exodus followed by talent influx as roles fundamentally change
- Embrace role fluidity: Stop distinguishing between engineers, PMs, and tech leads - focus on who can deliver customer-facing value
- Maintain experimentation culture: Keep trying AI implementations even when they fail initially
The message from these AI-first founders is clear: the companies that thrive in the next decade won't be those that add AI features to existing products, but those that rebuild their entire operational DNA around AI capabilities. The transformation isn't just possible - it's essential for competitive survival.
~~~~~
AI Disclaimer: AI helped us summarize the key points from these videos, and our editorial team reviewed everything to make sure it's clear and correct.
