The Patentable Edge: How AI Might Reinvent Startup Defensibility
When Feltsense’s AI agents replicated an entire Y Combinator Winter 2026 (YC W26) demo day batch in under 24 hours, it sent a jolt through the startup ecosystem. What once took months of founder hustle - ideation, product development, go-to-market - was now being compressed into hours by autonomous software. The implications are profound: if AI can clone your startup overnight, what’s left to defend?
In a world where software is increasingly commoditised, the question of defensibility becomes existential. Traditional moats like speed, brand, and network effects still matter - but they’re no longer sufficient. So where should founders look for protection?
One answer might lie in the realm of intellectual property. Specifically: patents.
Patents, by definition, grant a time limited monopoly on an invention – also known as a government granted defensible moat.
But here’s the twist: what if AI could help identify patentable inventions too?
Using AI to identify patentable inventions for startup defensibility: A thought experiment
To explore this, I ran a thought experiment. I took the same startups from the YC W26 cohort and asked an AI agent: what is a plausible patentable invention at the heart of each company’s product? The goal wasn’t to predict actual patent filings, but to see whether an AI perspective could surface technical differentiators that might form the basis of a patentable moat.
Scroll through the table below to see each startup, a short description, and a speculative (but realistic) patentable invention that could be derived from their core idea.
24 YC W26 Startups and Their Hypothetical Patentable Inventions
Startup |
Description |
Speculative Patentable Invention |
AI-native recruiting platform. Uses AI to automate sourcing and screening of candidates, promising 2× better hires 4× faster for startups. |
Adaptive Interview Pipeline - A computer implemented interview system that dynamically adjusts question content and evaluation criteria in real time based on candidate performance data and cultural fit metrics. |
|
User engagement automation tool. Analyses user behaviour and feedback to predict churn or intent, then triggers personalised in-app messages and emails to improve retention. |
Real-time Intent Prediction Engine - A user engagement optimisation method that merges qualitative user feedback with quantitative usage data to predict user behaviour and trigger personalised content delivery in real time. |
|
Marketing suite for the “agentic web”. Helps companies create content that makes them more visible and attractive to AI agents (like ChatGPT plugins or AI search). |
AI-Optimised Content Generator – A content generation method for automatically producing web pages optimised for indexing by AI-driven search algorithms and virtual assistants, by analysing trending query patterns and structuring content for machine readability. |
|
E-commerce “return prevention” platform. Uses AI-driven SMS outreach to customers post-purchase, aiming to solve issues (sizing, installation, etc.) and prevent returns before they happen. |
Proactive Return Mitigation System – An e-commerce return mitigation system that predicts likely product returns by analysing post-purchase customer signals and automatically initiates targeted interventions to pre-empt returns. |
|
AI-native mortgage origination. Automates home loan processing by collecting documents and checking them for compliance, using an AI assistant to streamline paperwork. |
Automated Mortgage Document Validator – An automated loan document verification method using multiple AI agents that extract data from borrower documents and cross-validate the data across said documents in real time as part of the mortgage origination process. |
|
Voice AI payments platform. Enables secure credit card payments via voice (e.g. during a phone call or through a voice assistant) with an API to handle compliance (PCI) and fraud checks. |
Voice Payment Authentication Protocol – A voice based payment authentication system using voice biometrics and contextual processing to verify user identity and authorise credit card transactions over voice channels. |
|
Continuous accounting reconciliation. Connects to payment processors, bank APIs, and ledgers to auto-match transactions, eliminating month-end crunch via a live, rolling close. |
Autonomous Ledger Reconciliation Engine – A continuous ledger reconciliation system combining rule based transaction matching and machine learning anomaly detection to automatically reconcile high-volume financial transactions in real time. |
|
Full stack AI accounting. Offers businesses “real-time, audit ready books” via an AI bookkeeper called Bea, supervised by human accountants. |
Live Bookkeeping Co-Pilot – A real time bookkeeping system that integrates with multiple financial data sources to automatically categorise and record transactions, employing anomaly detection to maintain continuously updated, audit ready financial records. |
|
AI estimating software for contractors. Automatically reads building plans or project specs and produces detailed construction cost estimates and proposals in minutes. |
Automated Construction Cost Estimator – A computer implemented method for generating construction project cost estimates by analysing digital building plans and specifications, extracting material and labour requirements via image/text processing, and mapping them to a cost database. |
|
AI teammates for industrial distributors. Integrates with legacy ERP systems to automate routine tasks like quoting prices and entering purchase orders from emails/pdfs. |
Instant Quote Parsing Algorithm – An automated purchase order processing system comprising an OCR module to digitise unstructured order documents and a domain specific language model to interpret line items and generate corresponding price quotes or order entries. |
|
AI scheduling assistant. Learns your work habits and preferences to schedule meetings on your behalf, handling the back-and-forth of finding times. |
Personalised Calendar Coordinator – A scheduling coordination system that learns individual user preferences and constraints from calendar data, and automatically negotiates and finalises meeting times across multiple participants’ calendars. |
|
Agentic workspace messaging. Puts AI “copilots” into team chat systems to monitor conversations and proactively assist (e.g. summarising discussions, retrieving info, automating tasks across SaaS tools). |
Collaborative Chat Agent Framework – A collaborative messaging platform integrating multiple AI agents that monitor team conversations and orchestrate task execution (information retrieval, summarisation, scheduling) in real time within the chat environment. |
|
“24/7 data analyst” for e-commerce. Integrates 600+ marketing, sales, and product data sources to provide instant analysis and plain-English insights for D2C brands. |
Automated Multi-Source Analytics Pipeline – An automated analytics pipeline that integrates heterogeneous data sources, normalises and aggregates the data, and applies machine learning to generate narrative business insights in natural language. |
|
AI voice agent for dental practices. A virtual receptionist that answers calls 24/7, books or cancels appointments, and handles patient queries by talking to the practice’s database. |
Healthcare Call Automation Agent – An autonomous voice driven assistant for medical practices that uses speech recognition and natural language processing to interpret patient phone calls and interact with practice management systems to schedule appointments and provide requested information. |
|
AI-assisted operations for home care agencies. Provides human virtual assistants supercharged with AI tools to manage caregiver scheduling, hiring, and client intake for care providers. |
Intelligent Care Scheduling System – A scheduling optimisation system for home care agencies that matches caregivers to clients by evaluating multiple constraints (caregiver skills, client needs, locations, availability) and dynamically re-optimises schedules when changes or cancellations occur. |
|
Insurance pre-authorisation automation. Uses AI to draft and submit prior-auth requests and appeals for medical procedures or medications, aiming to increase approval rates for clinics. |
AI Medical Pre-Auth Engine – A prior authorisation document generation system that automatically extracts relevant patient and treatment data, applies insurer-specific rules and terminology, and composes a complete pre-authorisation request ready for submission. |
|
AI workers for home services and trades businesses. Automates the back office so field crews can focus on the actual work (handles quoting, estimating, procurement, insurance coordination, scheduling, follow-ups). |
Distributed Agent Orchestration System - A field service automation platform that coordinates multiple AI agents operating across heterogeneous software environments, enabling dynamic task allocation, real-time synchronisation with both legacy and modern systems, and adaptive workflow management in response to operational changes such as job updates, customer interactions, and third-party system inputs. |
What this means
The above table is a thought experiment.
The inventions suggested are speculative and illustrative, not actual patent filings and certainly not legal advice. But they highlight an important idea: many startups have the seeds of patentable innovation in their core technology.
While Feltsense’s original article argues that defensibility often arises from the “messiness” of a domain, domain complexity is not a requirement for patentability!
Even in more straightforward or operationally clean domains, if an AI system introduces a new method for solving a technical challenge - that innovation may be just as patentable as a breakthrough in biotech.
And ultimately, a granted patent is a formal and enforceable form of AI defensibility: a government-backed monopoly right to exclude others from using your invention for up to 20 years.
The role of AI in patent discovery
Could AI tools help founders identify patentable aspects of their product? Possibly.
The same way AI can generate ideas or write code, it might comb through a startup’s architecture or data flows and highlight components that are novel technical innovations. In this way, AI acts as a kind of R&D assistant - not inventing on its own, but helping human teams spot the inventive concepts that they could protect.
(Of course, any invention still needs to meet the legal criteria of subject matter eligibility, novelty and non-obviousness. That’s where human expertise and patent counsel remain crucial.)
If you’re using AI this way, make sure you’re using a secured, private tool and don’t paste sensitive or confidential information into a public, web-based LLM where it could be stored, shared, or used for training.
What parts of your technology are truly novel? Consider protecting them
If you’re a founder or investor, it may be time to add patent strategy to your defensibility toolkit.
In an era of AI replication risk, relying solely on moving fast or building network effects might not be enough.
Even if you’re “just” automating an administrative workflow with AI, ask yourself: Is there a technical innovation here that’s patent-worthy? If yes, securing a patent can give you a legal monopoly on that innovation - a moat that no fast follower can cross without infringing.
FB Rice can help turn AI insights into actual patent protection. Reach out to our team of experts to explore.
Related IP News & Insights
