Institutional memory
Ask how work was done and get a source-backed answer from what actually happened. When someone leaves, their practical know-how does not have to leave with them.
Keep what your company knows.
Scroll
The overlooked layer
What's holding enterprise AI back isn't the model. It's context. How your company actually works lives in its people and leaves when they do. Arvet keeps that knowledge and puts it to work.
What Arvet does
Workflow automation
Arvet observes a task, learns the real steps, and runs it automatically. Because the automation is built from how the work actually happens, it handles the practical details rigid scripts miss.
Process intelligence
See where time goes, which handoffs create delays, and where work gets stuck. Every insight is grounded in how the organization actually operates.
Proof
Built in real offices,
not a lab.
$1T
The estimated annual cost of voluntary turnover to US businesses. Knowledge loss is one of the operational problems departures create.Gallup, 2019
2 wks
In one live trial, Arvet learned a worker's recurring routine and automated it within two weeks.
3 sites
Already running with design partners in a PR firm and two family-office environments.
How it works
From approved activity to usable memory
-
01
Define the boundaries
Your organization decides what Arvet can observe. Approved work applications are included; everything else stays out.
-
02
Build the memory layer
Arvet turns screenshots and process activity into a structured understanding of workflows, decisions, and relationships.
-
03
Answer and automate
Ask questions through Claude or ChatGPT and get answers grounded in real work. When Arvet recognizes a repeated workflow, it can run it automatically.
-
04
Improve with use
Every workflow makes the memory richer, the answers sharper, and the automations more capable.
You decide what's captured, where it lives, and who can see it.
Whitelist-first capture
Nothing is captured until it has been explicitly approved. Non-work applications stay outside the boundary.
Modular controls
Set capture and access rules by role, team, company, or region. Sensitive teams and regulated jurisdictions can be excluded without limiting the rest of the organization.
Data stays under your control
Your data remains yours. Arvet builds the memory inside your organization's boundary, where you control access, retention, and use.
Train agents on how your company actually works.
Role-specific agents
Build finance, operations, and client-service agents grounded in the real expertise of your best people.
The context layer
General models know the world. Arvet teaches them how your company works. The model can change; your organizational memory persists.
Compounding intelligence
The longer Arvet runs, the more deeply it understands your workflows, decisions, and institutional knowledge.
Who's building Arvet
Two founders, already building in the field.
We met in computer science class at QASMT and have been best friends since. We've been building with AI since the GPT-3 Playground era and now work inside a family office, where we see the knowledge-retention problem firsthand. Arvet is already running in three offices across finance and professional services.
Jet Kozarcic
Co-founder & CTO
Jet studies computer science and artificial intelligence at UQ, works in AI professionally, and runs a funded AI startup. He leads Arvet's technology, product, and engineering and has already deployed it in real workplaces.
William Liu
Co-founder · Finance & go-to-market
William studies Advanced Finance and Economics at UQ, has experience across finance and PR, and contributes to the product build alongside Jet. He leads finance and go-to-market. The idea for Arvet began after a conversation with an EY-Parthenon partner made the scale of institutional knowledge loss clear.
Questions, answered
The things buyers ask first.
Is this employee surveillance?
No. Arvet is built to preserve knowledge and automate work, not judge employee performance. Capture is off by default and limited to the work applications your organization explicitly approves.
Where does our data live?
Inside your organization's boundary. Your data remains yours, and you control who can access the memory Arvet builds.
What gets captured?
Periodic screenshots and process events from explicitly approved work apps and browser tabs. Everything outside that whitelist is excluded.
What does the employee get?
Each user gets a searchable timeline of their own work and can use Arvet, Claude, or ChatGPT to recall previous decisions, files, and workflows. Repeated tasks can then be automated.
What if we want to leave or switch?
You own the context Arvet creates. The memory is model-neutral and portable, so you can change AI providers without starting again.
How do you handle security & compliance?
Arvet is being built for enterprise deployment from the beginning, with whitelist-first capture, role-based permissions, and regional controls. SOC 2 and other formal certifications are part of the roadmap.
Which AI models does it work with?
Arvet is model-neutral. It connects through the Model Context Protocol (MCP), so you can query your work history from assistants such as Claude or ChatGPT while keeping the same organizational memory.
Who is Arvet for?
Arvet is starting with knowledge-intensive finance and professional-services organizations, including family offices, investment firms, consultancies, and accounting teams.