We design and build AI systems that fit the way you already work.
Arbutus Labs is a specialized design and implementation team. We learn how your business runs, build automations into the tools you already use, and stay on to keep them working. You keep the judgment; the system brings you the decision.
Queen's University
- Custom AI Agents & RAG
- Prompt Engineering & Automation
- API & Workflow Integration
- Python, TypeScript, SQL
- Claude, OpenAI & Cloud APIs
Working With
Builders and developers, real estate, law, and professional services.
What We Build
From a single workflow to your whole operation.
Arbutus Labs designs, builds, and maintains custom AI agents and automations that run on your own emails, files, and tools. Whatever the work, the system draws on your real information and brings you the decision when your judgment matters.
Nothing gets sent or changed until you approve it.
How We Work
Systems built around how your business actually runs.
Every build starts with your real operation: your inbox, your files, your tools. We write a knowledge base about your business, and the system acts on what it's confident about, brings you the decision when it's not, and learns from every correction. Here is how a typical system runs.
Built on Claude and OpenAI, with custom API and webhook integrations across email, calendar, CRM, and cloud storage. Prompt engineering and confidence-based routing decide what to handle and what to bring you.
Oversight & Safety
Starts in approval mode
Every system begins with human sign-off on every action. It only acts independently once you trust it.
Confidence routing
High confidence: acts silently. Medium: acts and flags for review. Low: asks before doing anything.
Full visibility
Every action is logged. You can see exactly what was done, when, and why, and correct anything instantly.
Your data stays yours
Systems connect through scoped, revocable permissions to accounts you already control, on established enterprise infrastructure. Your data stays in your systems and is never used to train AI models.
In Practice
One example of what we've built, in full. The integrations and the purpose change with every business; the approach stays the same.
From inbox to the right budget line in a single tap.
The problem
Invoices arrived by email across multiple active projects. Each one meant reading the paperwork, pulling out the details, entering them into the right budget by hand, and hoping overruns got noticed before they became expensive.
What we built
A complete invoice and budget-tracking system. Invoices arrive by email, appear in the owner's chat within seconds with the details already extracted, and land in the right project budget with a single tap.
The result
Budgets stay current in real time, with automatic warnings the moment any budget line runs hot. The owner makes the call. The system does the paperwork around it.
Other Examples
How It Works
Find, build, scale.
01 —Find the highest-impact work
We start with your team and leadership, mapping where AI and automation will move the needle most across the business.
02 —Design and build
Systems are built into the tools you already use, with human oversight in place, and proven on real work before anyone relies on them.
03 —Measure and scale
We track the impact, refine what works, and expand across more of the operation as each system earns trust.
Get In Touch
Start with a conversation.
A short call about how your business runs and where a system could take work off your hands. If it's not a fit, we'll say so.