The sales director had watched it happen four times in one quarter. A promising enterprise deal would reach the final stages, procurement signed off, legal happy. Then someone in operations would ask if the platform could export compliance reports in a particular Excel format their auditors required. Engineering would log the request. The deal would stall.
By the time someone greenlit building that export tool—six months and seven figures in lost pipeline later—those prospects had signed with competitors.
This is the gap that Gigacatalyst, a San Francisco startup in Y Combinator's Spring 2026 batch, says it can close. Not by building features faster, but by letting the people who lose those deals—sales reps, customer success managers, sometimes even the customers themselves—build what they need.
The pitch sounds almost reckless at first: hand non-technical users an AI tool that generates custom features inside your SaaS product. But after talking to founder Namanyay Goel and reviewing data from early deployments, a more nuanced picture emerges. One customer unblocked over $1 million in stalled pipeline within six weeks, according to company-reported data.
Building Features Without the Backlog
What Gigacatalyst has built is an embedded AI customization layer that sits inside a host SaaS application like a native feature. A sales rep facing objections about missing functionality can prompt it in plain language—"create a dashboard showing overdue tasks by region"—and the system generates a working app that runs sandboxed within the platform.
The distinction from existing low-code tools matters here. This isn't Zapier or Retool operating as a separate service customers need to learn. The builder inherits the host application's authentication, design system, and permissions structure. To end users, it appears as though the SaaS vendor shipped a new capability.
"After 2000+ daily users, 900+ apps built, and 70% 30-day retention, today we're opening a public demo," Goel wrote in a launch post on Hacker News on May 13, 2026—an oversight that hints at either the chaotic pace of a YC batch or a startup still finding its footing with external communications.
The technical architecture relies on what Goel calls "agentic API discovery." The system maps the host product's capabilities, then generates code that passes through validation layers—static checks, runtime analysis, an LLM acting as judge—before anything executes. Everything runs sandboxed, with most generated apps limited to read-only operations. Write functions exist but operate under tight constraints.
Setup reportedly takes two days for white-glove installation, full deployment within two weeks.
The Math on Feature Bloat
Gigacatalyst's core argument addresses a well-documented tension in B2B software: customers want customization, but building every requested feature leads to bloat. Research from Pendo published in July 2024 found that just 6.4% of features in a typical product generate 80% of usage. The rest? Dead weight in the codebase.
Yet ignoring feature requests carries its own cost. The company frames its product as an escape hatch from that binary trap—build everything or build nothing. Instead, let individual customers or account teams create the micro-features they need without burdening the core roadmap.
Whether that vision holds up at scale remains to be seen. The company reports five live customers serving 2,000 daily users, with 70% retention at 30 days. One client, Scalio, hit 500 daily active users within a month and called the embedded builder "the highest retention part of our app."
Perhaps more telling: Ryan Chan, CEO of UpKeep—a computerized maintenance management platform—offered a testimonial noting that "over 1,000 UpKeep customers use it daily." UpKeep recently launched "Studio," an in-product AI builder with capabilities that align closely with Gigacatalyst's offering, though neither company has explicitly confirmed the relationship. The timing of UpKeep's February announcement and Gigacatalyst's presence in the current YC batch suggests the partnership predated the accelerator stint.
Market Territory and Open Questions
The competitive landscape gets crowded quickly. Retool offers AI-generated apps and embedded customer portals. Superblocks emphasizes governed enterprise AI applications. Platforms like Workato Embedded, Tray Embedded, and Zapier's white-label offering let SaaS products surface integration marketplaces, though they focus on connecting external services rather than generating net-new UI.
Goel positions his product against Lovable, a standalone AI app builder, with a key distinction: "Think Lovable, but on your platform." The embedded nature—appearing as a native capability with no separate login or learning curve—is the entire premise.
Pricing remains opaque. The company says cost "depends on expected usage" and that "most implementations pay for themselves within the first four weeks," which is either confident or wishful depending on how you read it.
There's a small puzzle in the company data. Y Combinator's profile lists four team members. LinkedIn shows a range of 11 to 50 employees. The discrepancy likely reflects contractors, advisors, or the messy reality of startup headcount during rapid scaling, though it's worth noting.
The company incorporated as Giga Next Inc. in October 2025, months before entering the accelerator. Goel, who's been coding since age 13, claims his AI-focused writing reached over 4 million views in 2025—though that reference to a future year raises questions about either the draft's currency or the founder's optimism about content virality.
A public demo launched in mid-May at app.gigacatalyst.com/try for SaaS teams curious enough to test the concept themselves.
What It Means if This Works
If Gigacatalyst's approach proves out, it suggests something larger than a feature-building workaround. The real shift would be in who controls product roadmaps. Not entirely, of course—core features still require engineering rigor. But at the edges, where deals stall over small workflow mismatches, the power balance might tilt toward the people closest to customer pain.
That's either liberating or terrifying, depending on where you sit. For sales teams tired of losing winnable deals, probably the former. For engineering leaders already managing technical debt, maybe the latter.
The next few quarters will tell which instinct proves correct.
