57% augmentation and 43% automation
Anthropic Economic Index found AI usage skews toward augmentation over automation. Real usage today is mostly humans using AI, not AI replacing humans.
View sourceABOUT AGOSEC
Human in the loop. Always.
We build AI tools for real people. We keep people in control. We call this HITL. Human in the loop.
AI gives intelligence at our fingertips. It levels the playing field for learning. It helps people digest hard info fast. We believe this can raise society level thinking in a big way.
AI should create better conversations. AI should not replace conversation. We are explicitly against full AI agent replacement for Human Problem Solving. AI should enhance people.
Our product goal is quick intelligence for people. Fast context. Better judgment. Better dialogue. Yes, our demo can look like replacement. That is marketing pizazz and visual effects. The real product direction is human capability amplification with people accountable for final decisions.
The Agosec Keyboard product
Agosec Keyboard is an AI-first keyboard that lives where your conversations already happen. It combines fast web search, multi-turn AI chat, and screenshot context so you can research, reason, and reply without context switching.
This is exactly where HITL matters most. The keyboard generates options, summaries, and drafts. You choose what to send, what to revise, and what to reject. AI is the copilot in the typing flow. The human remains the pilot.
Product decisions follow the same rule: automate repetitive mechanics, keep judgment human. That means faster execution for users without giving up accountability, nuance, or trust.
What we believe
- 1. Keep humans in charge.
- 2. Use AI to automate boring repetition.
- 3. Keep hard decisions human.
- 4. Learn from AI output before shipping it.
- 5. Delay gratification. Understand first. Then apply.
Transparency is part of the product
Agosec is currently run by one engineer. Sammy Kao. We keep our TODOs public. You can inspect what we are building, what we are fixing, and what is next.
Next step is budget transparency. Marketing budget visibility is planned. We want users to see how resources are allocated.
Useful numbers
Short facts. Straight facts. All selected to support assistance over full replacement.
ILO: jobs are more likely to be augmented than destroyed
The ILO concluded generative AI is more likely to augment jobs than destroy them. This supports an assistance model over full worker replacement.
View source14% productivity lift with AI assistance
NBER found support agents became 14% more productive when humans used AI assistance. The gain came from augmentation during work, not autonomous replacement.
View source25% faster and 40% higher quality with human plus AI
BCG and Harvard researchers found consultants using GPT-4 were faster and produced higher quality work on suitable tasks. Human plus AI outperformed humans alone.
View source19% worse when people over-trusted AI
The same BCG study found performance dropped by 19% when users relied on AI beyond its capability frontier. AI alone is not the answer. Human problem solving is still required.
View source44% of worker skills will change in 5 years
World Economic Forum reports 44% of worker skills will be disrupted. The response is upskilling and bigger human problem solving, not removing humans from the loop.
View sourceSee the work, then shape the work
If this approach matches how you think AI should be built, go inspect the public roadmap or read the latest writing.