The daily grind of business operations hasn't changed much. Data still bounces between Slack, Gmail, Notion, and whatever other tools a company has cobbled together. Automation promised relief years ago, but existing platforms demanded either precious engineering hours or forced teams into opaque workflows they couldn't peek inside, let alone modify.
Enter Bubble Lab, fresh from Y Combinator's Winter 2026 batch, with a different wager: What if the AI assistant just lived in Slack, where ops teams already spend their days?
The two-person startup launched Pearl late last month—an AI that transforms conversational requests into working automations across a company's entire software stack. The pitch? "Prompt once, automate forever." Type a task in plain English through Slack, and Pearl builds the workflow. Yesterday's support tickets need summarizing and emailing to your team lead? Done. Specific Slack threads should automatically spawn Jira tickets? Pearl handles it.
According to both the company's homepage and Y Combinator's official LinkedIn amplification on February 26-27, more than 4,000 business professionals are already using the platform. That's a notable clip for a startup that appears to have launched just months ago.
Why Slack? Why Not Another Dashboard?
Traditional automation platforms ask users to open yet another tab, navigate to a dashboard, configure triggers and actions. Bubble Lab made a different choice—Pearl operates entirely within Slack channels and direct messages.
The company frames this as "in the flow of work" execution. Ops teams trigger automations without the cognitive tax of context-switching. "Deploy Pearl in one click," the website states, before adding a pointed clarification: this is "not another chatbot" but a system that "takes action across your stack."
The platform already connects to more than 25 tools—Gmail, Google Calendar, Drive, Sheets, GitHub, Jira, Notion, Postgres, Stripe, Airtable, Telegram. A demo dated January 27 showcasing a Slack-plus-Gmail workflow suggests the company had preview builds circulating ahead of the official YC announcement.
Under the hood, things get more technical. A December 15 blog post detailing a "Pearl Upgrade" reveals the architecture: the AI assistant now runs code-based sandbox execution for gathering context. It scrapes documentation, lists database schemas, inspects Google Sheets, then generates finalized TypeScript workflows. Each one requires user approval and runs in a secure sandbox with full execution logs.
Perhaps more importantly—and here's where Bubble Lab diverges sharply from competitors—the company compiles workflows to production-ready TypeScript code and releases its core platform under an Apache 2.0 license on GitHub.
For ops teams wary of vendor lock-in, or compliance departments that need to audit every line of code touching company data, that's not a small detail.
The Economics of Deterministic Workflows
Bubble Lab claims that workflows promoted from one-off Pearl requests into automated runs operate roughly 10 times faster and 90 percent cheaper than repeated ad-hoc agent executions. A "99.9% reliable" badge appears on the site, though the methodology behind that percentage remains unclear.
The platform also imports existing workflows from n8n, a popular open-source automation tool—suggesting Bubble Lab sees migration paths from established players as part of its growth strategy. Whether n8n users will actually jump ship is another question entirely.
Observability features include detailed execution tracing, logs, token usage tracking, performance metrics. Table stakes for enterprise ops teams, certainly, but not always guaranteed in newer AI-native tools where speed to market sometimes trumps operational rigor.
Security Claims in a Scrutinized Market
On its website, Bubble Lab asserts SOC 2 Type 1 certification and CASA Tier 2 verification. CASA (Cloud Application Security Assessment) Tier 2 represents "lab tested/verified" status within Google's App Defense Alliance ecosystem—relevant for applications integrating with Google data scopes. No public trust center link is available yet, though many vendors share SOC reports under NDA rather than publicly.
The timing of these security claims is worth noting. TechRadar reported critical vulnerabilities in n8n across multiple versions in January and February, urging immediate upgrades. For open-source workflow platforms competing on transparency and self-hosting capabilities, these incidents create both risk and opportunity. Bubble Lab's pitch on code auditability lands differently when competitors are patching holes.
Two Founders, 4,000 Users, Months to Build
Bubble Lab was founded in 2025 by Selina Li (CEO) and Zach Zhong. Both are Cornell Tech graduates who previously co-founded gymii.ai—Li completed her undergraduate degree at UPenn in 2023 before pursuing a Cornell Tech MEng in 2025; Zhong holds a Cornell Tech CS degree from 2025 and a UCSD CS degree from 2024.
According to a founder LinkedIn post, the pair met at Cornell Tech and built gymii.ai before automation pain points led them to pivot. They're working with Diana Hu as their primary YC partner.
Customer logos on the homepage include Dyna Robotics, Corgi, Hattrick Ventures, InsForge, and PeakMojo. Jason Ma from Dyna Robotics posted on LinkedIn approximately three days before launch, stating his team uses "Pearl daily" and noting "internal traction among our ops team."
In a mid-January or early-February LinkedIn post, Li noted that "two months later" after starting, the company had reached "3,000+ business professionals" through design partnerships with ops teams. By the official YC launch amplification on February 26-27, that figure had climbed to over 4,000.
How much of that represents active, daily users versus beta signups remains to be seen.
Pricing That Competes, Maybe
Bubble Lab's tiered pricing starts with a free Starter plan: 100 successful workflow executions per month, one active workflow, five Pearl requests daily. The Pro tier runs $29.99 monthly (6,000 executions, 10 active workflows, 15 Pearl requests per day). Scale costs $99.99 monthly (50,000 executions, 25 active workflows, 100 Pearl requests per day). Enterprise pricing is custom and includes role-based permissions, SLAs, and on-premises deployment options.
The company only charges for successful executions and bills managed integration usage through monthly credits with a five percent markup—a model that sounds reasonable until you're troubleshooting failed workflows and wondering if debugging counts toward your limit.
A Crowded Field Gets More Crowded
The product enters a market where every major automation platform is racing toward AI-native workflows.
Slack itself rolled out Workflow Builder AI enhancements in August 2024, with additional features throughout 2025. Salesforce's Agentforce now operates inside Slack, offering employee agents and templates. Zapier made significant changes to its Agents product in May 2025, with admin controls updated in December and beta agent versioning discussions appearing in January 2026. Make announced AI Agents in an April 14, 2025 press release and shipped next-generation agents in late 2025. n8n launched its own AI Workflow Builder, documented in current help resources.
Bubble Lab's pitch centers on differentiation through Slack-first execution, TypeScript code ownership, open-source architecture, and the ability to promote experimental agent runs into deterministic workflows.
Whether that's enough to compete against platforms with mature app ecosystems and multi-year head starts depends on how quickly ops teams adopt—and how deeply security and observability features resonate with technical buyers. Third-party aggregators list roughly $500,000 raised, likely representing a YC-stage pre-seed round, though this remains unconfirmed pending Y Combinator Demo Day on March 24.
For a two-person team claiming 4,000 users within months of launch, the trajectory suggests either strong product-market fit among early adopters or aggressive beta signup campaigns.
Either way, the ops automation space just got another competitor betting that AI agents belong inside team communication platforms, not siloed in standalone dashboards. The founders are betting that giving users code they can own and workflows they can inspect matters more than polished enterprise sales decks.
Time will tell if ops teams agree.
