You’ve got a high-stakes call to make. Your company needs an AI strategy, but which way do you go? All-in on a comprehensive refounding—rebuilding everything around AI-native architecture—or the safer bet of adding AI features to what you’ve already got?
Get this wrong and you’re either months deep into unnecessary transformation or you’re watching AI-native competitors build self-improving systems you can’t match by sticking features onto your legacy setup.
This guide is part of our comprehensive understanding the refounding trend, where we explore how companies navigate AI-driven business model transformation. Companies like Airtable, Handshake, and Opendoor are announcing refounding initiatives, not incremental feature roadmaps. Meanwhile other companies are doing just fine with measured approaches.
So in this article we’re giving you a four-factor decision framework: technical feasibility, business model economics, organisational readiness, and competitive positioning. You’ll get assessment criteria, board negotiation guidance, and risk mitigation strategies to make the right call based on what your company can actually do.
What Is Startup Refounding and How Does It Differ From Pivoting?
Refounding is when you rebuild your business around AI-native architecture while keeping your market position. It’s not the same as pivoting.
The difference matters. Pivots are about course correction after you’ve worked out an approach isn’t cutting it. Refounding is proactive transformation to grab new opportunities without admitting your previous strategies failed. When Airtable’s CEO Howie Liu announced their refounding, he was clear: “This is not about changing direction after getting something wrong”.
Refounding changes everything at once—business model, technical architecture, organisational culture, and value proposition.
Pivots usually tackle one dimension—product, market, or business model—while keeping the rest steady. You pivot your product-market fit or your revenue model. You refound your entire company.
The timing’s different too. Pivots happen reactively when your current path fails. Refounding happens proactively to capture emerging opportunities. Investors view these announcements as necessary adaptations to technological disruption rather than distress signals.
Think about the historical examples. Salesforce defeating Siebel wasn’t a pivot—it was a new player unburdened by legacy models. ServiceNow dominating legacy ITSM vendors followed the same pattern. Refounding is about recapturing that new entrant advantage while keeping your market position and customer relationships.
Why Are Companies Like Airtable, Handshake, and Opendoor Refounding Instead of Adding Features?
These companies get it. AI-native competitors build self-improving data flywheels that create advantages you can’t match by tacking features onto legacy architectures.
Airtable kicked this off in June, declaring they’d treat AI adoption as a foundational company reset rather than incremental feature development. They didn’t just add an AI assistant—they repositioned as an “AI-native app platform”.
Handshake’s refounding shows you the clearest financial picture. Their AI division grew from 15 to 150 employees within months and pulled in $100 million in annualised revenue in just eight months. They’re now at $200 million combined ARR and projecting “high hundreds of millions” for 2026.
CEO Garrett Lord said it plainly: “Winners and losers are being defined right now”. Without aggressive AI investment, Handshake becomes “merely okay,” stuck in incremental improvements generating modest quarterly gains—a pattern that leads to corporate deceleration.
Handshake rolled out dramatic cultural shifts, including mandatory five-day office weeks with expectations for employees to work “with a pace and number of hours that is meaningful and will help us hit goals.”
The threat is architectural. AI-native startups create self-reinforcing loops where usage generates data, improved data refines AI, better AI attracts more usage, and more usage generates better data. You can’t bolt this onto existing systems built around different assumptions.
What Are the Four Key Factors for Evaluating Refounding vs Incremental AI?
Your optimal strategy hangs on four factors: technical feasibility of agentic AI (can your team build agentic AI), business model economics of AI-first SaaS (can you make outcomes-based pricing work), organisational readiness for transformation (can you drive cultural transformation), and competitive positioning (how urgent are AI-native threats).
Technical feasibility isn’t about whether you can add machine learning features. It’s about whether you can build agentic AI architecture—systems that perceive environments, make decisions, act independently, and adapt strategies without constant human oversight. It’s about whether you can create data flywheel patterns where collected data continuously refines AI models.
Before we get into the details, here’s what you need to know about Bret Taylor’s three-layer AI market framework. Taylor identifies three AI market layers: Frontier Models (capital-intensive foundation models like GPT-4), AI Tools Market (infrastructure enabling AI), and Applied AI (agent-based solutions for specific job functions).
Applied AI is the most exciting tier—this is where specialised vertical agents solve specific industry problems and deliver measurable outcomes rather than productivity enhancements. Taylor argues what were SaaS applications in 2010 will be agent companies in 2030. Your refounding decision depends on which layer you compete in, with Applied AI companies having the clearest refounding imperative.
Business model economics asks if you can move to outcomes-based pricing where customers pay for results delivered, not software access. Traditional SaaS enjoys margins of 80-90%, but AI-first companies typically operate at 50-65% gross margin because of inference and infrastructure costs. Can you capture enough value from outcomes to justify those costs?
Organisational readiness looks at your ability to drive cultural transformation. Can you get the board on board? Can you keep key employees during comprehensive change? Can you sustain transformation effort before seeing results?
Competitive positioning examines the urgency. AI-native startups reach $1 million revenue in 11.5 months versus 15 months for traditional SaaS firms. Are AI-native startups threatening your market now, or have you got runway for an incremental approach?
Be honest with yourself on each factor. High technical feasibility and viable outcomes economics combined with AI-native competitive threats and organisational readiness point to refounding. Low scores on technical feasibility or economics with limited competitive threats suggest incremental approaches. For a complete overview of all aspects of the refounding overview, see our comprehensive guide.
How Do I Assess Technical Feasibility for AI-Native Architecture?
You need to work out if your team can build agentic AI systems with autonomous decision-making, data flywheel patterns for continuous improvement, and enterprise data architecture for model refinement.
Agentic AI is characterised by autonomy, proactivity, and goal-oriented behaviour. These systems perceive their surroundings, make decisions, act independently to achieve objectives, and adapt strategies based on new information.
The data flywheel creates a self-improving loop: production data refines models, better models generate better outputs, and better outputs create more valuable training data for continued improvement.
Can you create these systems? Work out the gap between your current architecture and AI-native foundation. How much technical debt is in the way? Do you have AI/ML expertise or mainly traditional software engineers?
Build capabilities in logical order: modern data platforms, then robust metadata and treating data as a product, then mapping business process metadata, then agentic automation. Skipping steps doesn’t work.
Start with human-in-the-loop systems that let agents propose actions while keeping humans in control of final decisions. Focus on high-impact, narrow-scope use cases where human involvement is expensive and decisions are repetitive but data-rich.
The build versus licence decision matters. Which bits need internal development versus licensing frontier models? Companies building Applied AI solutions should focus on solving customer problems rather than developing frontier models.
When Does Business Model Economics Favour Refounding Over Incremental Features?
Refounding becomes economically viable when you can move to outcomes-based pricing where customers pay for results delivered. This lets you capture higher value than per-seat subscription models despite increased infrastructure costs.
Traditional B2B SaaS enjoys margins of 80-90%, but AI-first companies typically operate at 50-65% gross margin. The cost drivers are real. Model development costs millions in GPU compute. Each user action triggers computationally intensive inference operations.
GitHub Copilot initially lost approximately $20 per user monthly when charging $10/month, as compute costs ran $20-80 per user.
Outcomes-based pricing changes the equation. Bret Taylor’s company Sierra demonstrates the model: customers pay a pre-negotiated rate when an AI agent autonomously resolves a customer issue. If escalation to a human is necessary, there’s no charge.
The market is heading this way. Gartner forecasts 40% of enterprise SaaS will include outcome-based elements by this year, up from 15% a few years ago.
The closer you get to solving a problem for a company, the more successful your business will be. Applied AI companies building agent-based solutions for specific job functions have clearer paths to outcomes-based revenue.
Can you quantify customer outcomes AI delivers? Are customers willing to pay for results versus software access? Can you absorb revenue model risk?
AI-driven SaaS will likely mature toward 60-70% gross margins—lower than legacy software but sustainable with proper cost management and value-aligned pricing.
How Do I Evaluate Organisational Readiness for Refounding Transformation?
Technical capability is necessary but it’s not enough for successful refounding. You need to work out if you can drive cultural transformation including renewed startup intensity, stakeholder alignment through board negotiations, and employee retention during comprehensive change.
Refounding means changing work location policies, pace expectations, decision-making speed, and risk tolerance.
Change-seeking cultures actively hunt for innovation rather than passively responding to disruption. They encourage workers to test ideas and take calculated risks, establish safe spaces for learning from failures, and use feedback mechanisms to distribute insights.
Can you get the board on board? Handshake’s board member Mamoon Hamid from Kleiner Perkins initially responded with surprise, but the board ultimately backed the direction. You’ll need to quantify competitive threats, model outcomes-based pricing economics, show technical capability gaps with incremental approaches, and present phased roadmaps to reduce perceived risk.
You need a compelling vision of what the aspirational culture will offer employees. Leaders must communicate a change narrative that creates shared understanding of the past, reasons for transformation, and a compelling vision for the future.
The way to change culture is to change how people behave. Create cross-functional teams. Hold blameless postmortems. By removing fear, you enable teams to surface problems and solve them more effectively.
When people see positive outcomes from initial changes, they become more open to further change—creating a virtuous cycle.
Which key employees might push back or leave during transformation? How long can you keep transformation going before seeing results? Does leadership have experience driving comprehensive organisational transformation?
Even with strong organisational readiness, competitive dynamics may force your hand on timing.
What Competitive Signals Indicate Urgent Need for Refounding vs Incremental Approach?
AI-native startups entering your market with data flywheel advantages, customers demanding autonomous AI solutions rather than enhanced workflows, and declining competitive differentiation from your current product all point to urgent refounding need.
The battle lines are clearest in vertical software. AI-native startups push deeper into industry-specific workflows—automating insurance claims, legal briefings, revenue cycle management. Traditional SaaS players face a stark choice: evolve or become obsolete.
Legacy companies must evolve from workflow-centric architectures to data-centric ones where data serves as the central product. AI-native companies empower generalist employees who adapt quickly across functions. Traditional siloed specialist teams move too slowly.
What are customers asking for? Autonomous problem-solving or feature enhancements? Outcomes or software access? This tells you whether incremental features will satisfy demand or whether customers expect fundamentally different capabilities.
By prioritising architecture over tools, enterprises can build knowledge engines—platforms that iterate, refine, and compound advantage with every interaction.
Is your differentiation dropping as competitors add similar features incrementally? How fast are AI capabilities advancing in your industry?
Are you protecting market share or capturing new opportunities? Defensive positioning with incremental features works when competitive threats are distant. Offensive positioning requires refounding when AI-native competitors are gaining traction.
If AI-native competitors are already in your market building data flywheel advantages, the incremental approach may be a temporary holding pattern rather than a viable long-term strategy.
How Do I Structure a Phased Refounding Approach to Reduce Risk?
Roll out refounding through parallel development tracks where AI-native architecture runs alongside legacy systems, with clear milestones, reversible decision points, and gradual customer migration. Avoid big-bang transformation failures.
A strategic, phased rollout builds internal confidence, allows for continuous learning, and minimises disruption. Follow this structure: Phase 1 focuses on low-risk, high-impact internal automation. Phase 2 addresses core value-chain enhancements. Phase 3 targets strategic differentiation and new business models.
Run AI-native systems alongside legacy systems for customer continuity. Define clear success criteria at each phase before committing to the next stage. Structure early phases so you can pivot back to an incremental approach if refounding proves unviable.
Customer migration needs planning. Will you gradually transition users or use full cutover strategies for different customer segments? How do you fund refounding without starving existing product development?
The strangler pattern gradually replaces parts of the system but, using a facade, those outside your system notice no difference. The aim is reducing migration risk by doing it a small bit at a time.
For data migration, use gradual approaches: sync data between old and new systems, migrate users in batches, validate data consistency continuously, and roll back individual users if issues arise.
Phase 1 examples include internal IT support automation: triage tickets, answer common queries, reset passwords, and autonomously resolve known issues. Phase 2 examples include finance operations and supply chain management.
Phase 3 focus: agents can power entirely new services, optimise complex decision-making at scale, and create dynamic, personalised customer journeys impossible with traditional systems.
Split teams between maintenance and rebuild. Feature freeze the old system except for emergency fixes. Put quality assurance focus on the new system.
Design a multi-year roadmap that balances quick wins with long-term strategic transformation. Communicate each phase’s success to build internal champions and secure ongoing investment.
FAQ
How long does typical startup refounding transformation take?
Comprehensive refounding usually takes 12-24 months from decision to initial AI-native product launch, with another 12-18 months for full customer migration and organisational transformation. Phased approaches let you capture value earlier. As detailed in our case studies with specific metrics, Handshake’s AI division generated $100 million in annualised revenue in just eight months, showing that aggressive execution can deliver results faster than typical timelines suggest.
Can I refound my company without returning to the office like Handshake did?
Return-to-office mandates are one cultural transformation approach, not a universal requirement. Some companies successfully refound with distributed teams by emphasising other mechanisms for startup intensity renewal—increased decision speed, clearer accountability, tighter alignment.
What if my board resists refounding and prefers incremental AI features?
Use the four-factor assessment framework to build a data-driven business case. Quantify competitive threats: AI-native startups reach $1 million revenue in 11.5 months versus 15 months for traditional SaaS firms. Reference Handshake and Airtable precedents to show how other companies secured board support. Model outcomes-based pricing economics versus subscription limitations. Present a phased roadmap to reduce perceived risk.
Should I hire a Chief AI Officer or restructure existing leadership for refounding?
Leadership structure depends on existing technical depth. Companies with strong engineering leadership often embed AI ownership in the CTO role. Those with weaker technical capabilities benefit from dedicated Chief AI Officers, particularly during transformation phases.
How do I know if my technical debt is too high for refounding to be viable?
Do an architecture assessment. If core systems need a complete rebuild to support agentic AI and data flywheel patterns, and rebuild timeline exceeds your competitive window, an incremental approach may be more viable as a starting point. For detailed guidance on data-centric architecture requirements, see our technical deep-dive. Build capabilities in logical order: modern data platforms first, then metadata and data as product, then business process metadata, then agentic automation.
What happens if I choose incremental AI features and competitors refound?
The incremental approach creates path dependency. AI-native competitors build data flywheel advantages that compound over time, making future refounding progressively harder. Production data refines models, better models generate better outputs, and better outputs create more valuable training data. If competitive assessment shows AI-native threats, the incremental approach may be a temporary holding pattern.
Can SaaS companies successfully transition to outcomes-based pricing?
Transition viability depends on your ability to quantify customer outcomes AI delivers, customers’ willingness to pay for results versus software access, and your capability to absorb revenue model risk. For a comprehensive outcomes-based pricing analysis, see our detailed economic breakdown. Sierra demonstrates this approach, charging only when AI agents autonomously resolve issues. Gartner forecasts 40% of enterprise SaaS will include outcome-based elements by this year. Applied AI companies building agent-based solutions have clearer outcomes paths.
How do I communicate refounding to employees without causing attrition?
Frame refounding as a growth opportunity, not a failure response. Give them a clear vision of the AI-native future state. Involve key technical leaders in strategy development. Offer learning and development opportunities in AI. Be open about cultural changes. For detailed guidance on cultural change requirements, see our comprehensive playbook. When people see positive outcomes from initial changes, they become more open to further change.
Should I refound if we’re already planning to IPO or get acquired?
Refounding usually extends exit timeline because of transformation complexity. If exit is imminent (12-18 months), incremental AI features may better serve near-term valuation. If exit is longer-term (24+ months), refounding may position the company for higher valuation in an AI-transformed market.
What’s Bret Taylor’s three-layer AI market framework and why does it matter for refounding?
Bret Taylor identifies three AI market layers: Frontier Models (foundation models like GPT-4), AI Tools Market (infrastructure enabling AI), and Applied AI (agent-based solutions for specific job functions). Applied AI is the most exciting tier, where specialised vertical agents solve specific industry problems and deliver measurable outcomes. Taylor argues what were SaaS applications in 2010 will be agent companies in 2030. Your refounding decision depends on which layer you compete in, with Applied AI companies having the clearest refounding imperative.
