2024
Trillbit Inc
huku.ai
Background
The future of work is AI-first.
We believe the most productive people in the world will spend their days collaborating with AI, making it central to how they think, create, and execute.
Yet today's tools treat AI as an afterthought.
They force people to work around them instead of working for the user.
We built Huku to change that.
An AI workspace that works for you.
A living, intelligent environment where human and AI agents collaborate naturally to automate routine tasks, accelerate strategic work, and unlock new levels of potential.
Core problem
Despite LLMs being powerful, they are still limited by four major barriers:
Knowledge Silos: Most company knowledge lives in PDFs, Slack threads, emails, and folders that AI cannot easily access or reason over.
Poor User Experience: Chatbots can draft an email but cannot run a marketing campaign or manage a product launch. Complex, ongoing work falls apart.
Single-Model Limits: One model cannot handle the full spectrum of tasks. Specialized tasks demand specialized intelligence.
Workflow Isolation: AI remains disconnected from the apps where real work happens. It cannot move data between tools, track progress, or close loops on its own.
Today, AI feels like texting an intern who forgets everything between conversations.
It offers flashes of brilliance, but no real memory, ownership, or sustained collaboration.
Until these barriers are broken, LLMs will remain powerful but peripheral, not the true work partners they are capable of becoming.
Solution
Key features
Multi-LLM System: Use the best model for each task — OpenAI, Anthropic, Gemini, or custom.
AI Knowledge Base ("Second Brain"): Seamlessly capture, tag, and retrieve structured and unstructured company knowledge, making it instantly LLM-usable.
AI-Native Work Interfaces: Purpose-built Docs, Sheets, Canvas for task orchestration — summoning agents inside the flow of work.
Expert Agents: Pre-trained vertical agents (Sales, Customer Support, Ops, Marketing) that can execute multi-step tasks, not just answer questions.
Fast Setup, No-Code Deployment: Teams can deploy and customize agents without engineering bottlenecks.
Deep App Integrations: Agents operate across email, files, Slack, CRM, spreadsheets — where work actually happens.
Memory + Tool Access: Agents remember context across sessions and wield the tools you already use.
My Role
Approach & Process
Research and Discovery
I began by deeply researching the emerging workflows of AI-first workers.
As the field was still rapidly evolving, I studied hundreds of AI products across categories. This included chatbots, agent platforms, knowledge bases, and automation tools. I mapped patterns, identified gaps, and closely analyzed where current solutions were falling short.
Alongside competitive research, I conducted user interviews, gathered early adopter feedback, and ran internal dogfooding experiments to validate real-world pain points.
The key insights were clear:
Knowledge remained trapped across fragmented, unstructured sources.
AI stayed isolated from the core tools people used daily.
Users lacked trust in AI agents beyond one-off tasks.
These findings directly informed Huku’s product principles.
We focused on building an AI-native workspace grounded in memory, action, and real collaboration.
Design and Development
I led the design and development of Huku’s core platform, aiming to create an AI workspace that was both powerful and human-centered.
We focused on four major pillars:
Multi-LLM Orchestration
I designed a system where tasks could be routed to the most suitable model in real time. This prevented single-model limitations and ensured that users always had access to the best AI capabilities available.Second Brain Knowledge Layer
I architected an advanced retrieval-augmented knowledge base that captured documents, emails, and files. This made all user information instantly usable for AI agents without manual tagging or curation.AI-Native Work Interfaces
I developed intelligent work surfaces across Docs, Sheets, and Canvas, allowing users to summon and collaborate with AI agents directly inside their work context. This eliminated the need for switching between chatbots and traditional tools.Agentic Workflows
We built multi-step agent workflows that could remember user preferences, coordinate actions across multiple tools, and manage complex tasks autonomously, with humans always remaining in control.
Throughout the build, I emphasized creating a workspace where agents were not just smart assistants but reliable, evolving collaborators inside the work itself.
Key Features
Huku launched with a focused set of features designed for real-world AI-first workflows:
AI-Native Work Interfaces
Users could collaborate with agents directly inside Docs, Sheets, and Canvas, without breaking their flow.Persistent Agent Memory
Agents retained context across sessions, building continuity and deeper collaboration.Cross-Tool Integration
Agents operated seamlessly across Slack, Gmail, Drive, and CRMs, executing actions inside real workflows.No-Code Customization
Users could create and adapt agent workflows without engineering support.Multi-Model Intelligence
Tasks were dynamically routed to the best AI models like OpenAI, Anthropic, and Gemini, improving output quality.
Each feature reinforced the core goal:
to create an AI workspace that worked for you and evolved with your needs.
