Investment Memo · Angel Round · YC SAFE · 2026
Vision
AI that learns the way people do — free from repetitive digital labor
$100K
Raising · Standard YC SAFE
$1T+
TAM
Glue work on the internet
$120–150B
SAM
Browser-native teams
$1–3B
SOM
High-freq automatable today
None
Valuation Cap
Investor-friendly entry
10%
SAFE Discount
Future priced round
$65K
Already Raised
65% funded · $35K left
Problem
AI Agents Require a 300-Page Manual.
Prompting fails operators — they skip steps, assume context, contradict themselves
Copilot problem — the capability exists, but the interface fails non-technical users
Glue work stays manual — CRM updates, data copying, form filling remain entirely manual
Core Insight: Instead of humans prompting agents, agents should prompt humans.

The Right Team
ZG
Zainab Ghadiyali
Co-Founder & CEO
Meta Ads Growth · Airbnb · Canva
Built 7-figure ARR company in under a year
GP
Giselli Panontini
Co-Founder & CTO
Youngest VP Engineering at Microsoft
Led Microsoft's Core Copilot infrastructure
10+ yrs neural networks & transformers
Partnership forged during a high-stakes AI security incident — international cybercrime ring. Paid zero ransom. Evidence before a federal judge. Build fast, rigorous, resilient.

Investment Ask
$100K
Strategic Angel Round · Standard YC SAFE
None
Val. Cap
10%
Discount
Raised: $65KTarget: $100K
Microsoft × Meta × Airbnb × Canva pedigree
Observation-based = proprietary data moat
Inference costs ↓ 90% · browser reliability ↑ 70–90%
88% companies use AI · none can configure agents
How It Works
The Show-and-Clarify Loop
01 Install 02 Demonstrate 03 Clarify IP 04 Compile IP 05 Execute 06 Supervise
Agent learns by watching a single demonstration — not by reading a 300-page manual
Asks clarifying questions to fill gaps — no need to anticipate every condition upfront
Compiles into a structured instruction set (IP) models can execute and refine over time
Runs autonomously across all browser-based apps — Sheets, HubSpot, Notion, Slack
Goal: 100,000 self-trained agents. Proprietary instruction graph builds a compounding data moat.
Market Opportunity
$1T+ Market for Glue Work
TAM
$1T+
20–30% of knowledge worker time
SAM
$120–150B
Browser-native teams
SOM
$1–3B
Automatable today
Why Now
Foundation models — inference costs ↓ 90%, specialise on dozens of examples
Browser automation — reliability jumped from ~40% to 70–90%
88% of companies use AI — but SMBs still can't prompt or configure agents
Competition
Two Axes · Four Quadrants · One Winner
How agents learn (Prompt vs. Observation) × Where they learn (Single vs. Cross-domain)
Q1 · Prompt × Cross-Domain
Zapier, UiPath, LangChain — assume users can encode complexity. They can't.
Q2 · Prompt × Single-Domain
AI SDRs, Receptionists — fail when logic spans multiple systems.
Q3 · Observation × Single-Domain
Macros, RPA recorders — features, not platforms.
Q4 · Observation × Cross-Domain ← STACKBIRDS
The ONLY platform enabling self-trained, cross-domain agents the way people actually learn.
GTM Strategy
Land 1 Workflow → Win the Org.
ICP
Ops/Mktg/Sales teams · 10–200 employees · heavy browser glue work
Wedge Target ARR
~$100K
within first months
1 Workflow
Role Automation
Multi-Agent Teams
Enterprise AI Workforce
Template flywheel: 1 workflow → templates → users → creators → more automation
Integrations: Scribe, Tango, Loom — millions already record workflows here
Business Model
Monetise Labor, Not Seats.
Agents priced by workflow runs — no config fees, no per-seat licensing.
Free
$0
1 agent · 1 workflow · 1 run/day
Usage ★
~$5
per workflow/month · scales with labor performed
Pro
$1,000
per month · unlimited workflows · multiple daily runs
Aligned with labor replaced — pay only when agents execute real work
Usage-led expansion — workflow → role → team value compounds naturally
Customer Validation
Early Customers Validate the Problem.
"Having an Office Manager Agent handle scheduling and invoicing would save me 10+ hours a week. I just want to focus on teaching Pilates."
Casey — Everybody Pilates Studio
"You are basically replacing outsourcing."
Patrick — Activus Capital (owns & operates 6 businesses)
"Our entire Customer Support team has tons of operations heavy repeatable workflows we could use help with."
Suhas — Pyjama HR
"I love that it learns by observation. I was a teacher — this is how I taught my students. So intuitive."
Liz — Liz Miller Communications
For AI labor to become mainstream, agents must:
Learn by watching humans Operate in existing apps Require no prompting Be supervised Work reliably Accessible to non-technical operators
Stackbirds is the
first to deliver all of this.