Dutify Blog

Context, agents, and modern work - from the team building Dutify.

A scoped credential separating private resources from public destinations

· 6 min read

Preview an AI assistant’s resource scope before you authorize it

Before approving an AI assistant, inspect exactly which workspace resources its credential can read, change, create, and disclose across private and public spaces.

ai-coworkeraccess-controlresource-scopingsecurity-review

 AI credential key locked to one selected workspace resource

· 7 min read

Resource scoped API keys keep AI Coworker writes on target

See how resource scoped API keys limit AI Coworker to the intended project, wiki space, or request board when ambiguous names could send a write elsewhere.

api-securityai-coworkeraccess-control

 AI task card awaiting human approval before being applied

· 7 min read

When AI Coworker acts by default and when it asks first

Learn how Dutify decides when AI Coworker should act by default and when a human must confirm, based on reversibility, exposure, and the cost of a wrong write.

ai-coworkerhuman-in-the-loopai-actions

 Transparent audit panels tracing an AI action from context to approval

· 7 min read

Why AI actions need a reproducible decision trail

Learn why reproducible AI actions require records of the original instruction, retrieved passages, model version, and human approval before any task or wiki change.

ai-coworkeraudit-trailsreproducibility

 AI agent reading structured tasks across a project board

· 9 min read

How task structure makes project boards usable for AI agents

Learn how clear statuses, custom fields, hierarchy, and views turn a project board into a reliable operating interface for AI agents in Dutify Cloud.

ai-agentstask-structureproject-management

 Four modern glass cards with clear symbols for work, documents, feedback, and meetings beneath one shared AI context layer

· 9 min read

What is an AI context layer?

An AI context layer gives people and agents a governed, current view of your team's work. Learn what it contains, how it works, and how to build one without creating another data dump.

agentssuite

 One physical key opening a single archival drawer while eleven other locked compartments remain closed

· 10 min read

How to connect AI agents to your team's work - without handing over everything

A practical least-privilege pattern for agent access: scoped identities, permission-aware retrieval, bounded tools, approval gates, and an audit trail.

agentsproject-management

 Many loose meeting fragments compressed into one decision card with three clipped action slips

· 9 min read

Meeting notes are not the point - decisions are

A transcript preserves what was said. Meeting intelligence becomes useful when it preserves decisions, owners, evidence, and follow-through. Here is how to build that workflow.

hubmeetings