Start with professional concepts
Use the language of matters, events, projects, commitments, documents, meetings, and next steps before exposing internal AI or orchestration terms.
AI Should Adapt to Professional Work—Not the Other Way Around.
Assist Mi is designed to absorb unnecessary AI complexity so professionals can work through familiar concepts, systems, and review moments without learning how to think like a model or workflow engine.
Reducing cognitive burden does not mean hiding AI involvement. Transparency, source context, uncertainty, review, audit, user agency, and advanced controls remain essential when they help people work safely and confidently.
Adapt the AI to the Work
Let professional intent lead the experience.
People should not have to master prompting mechanics, model abstractions, internal routing, or orchestration concepts before the platform can help with ordinary professional work.
Fit the workflows people already know
Work with familiar communications, calendars, documents, tasks, and role-specific systems instead of forcing every activity into an AI-native interface.
Treat usability as part of the outcome
A capable model is not enough. Defaults, naming, onboarding, review, and the final steps of follow-through must make the capability usable in real work.
Where Ava Fits
Ava is a conversational interface to Assist Mi—not a human and not the entire product.
Ava can provide a natural way to ask, review, prepare, and follow up. Behind that conversation, the platform supplies connected context, role-specific workflows, permissions, evidence, policy, action boundaries, and edition-specific capabilities.
The experience should feel natural without pretending that a person is on the other side of the interface or obscuring when AI helped produce the result.
Simplicity Is Not Secrecy
Hide unnecessary mechanics. Keep meaningful AI context visible.
Human-centered design removes friction that does not help the work. It does not remove information a professional, reviewer, administrator, or operator needs to understand and control the system.
Keep these signals visible when they matter
- When AI contributes to an interpretation, draft, recommendation, or action.
- The source context and evidence available to support important work.
- Uncertainty or confidence when it helps a professional calibrate trust.
- Status, review queues, approvals, audit history, and consequential-action boundaries.
- Advanced and administrative controls for people responsible for policy and operations.
Make these mechanics optional in everyday work
Prompt construction, model selection, object taxonomies, routing details, internal queue stages, and orchestration vocabulary can remain behind familiar product language unless exposing them helps an expert diagnose, configure, or govern the system.
The test: does this detail help the person accomplish, understand, review, or control the work? If not, the platform should absorb it.
Professional Agency
Make the next step easier without making the person passive.
Assist Mi can prepare and coordinate work, but professionals need a clear ability to inspect, change, approve, reject, redirect, or stop what happens next.
People decide where judgment belongs
Professional conclusions, relationships, exceptions, risk, approval, and accountability remain human responsibilities at the boundaries defined by the workflow.
Review is part of the experience
A review step should preserve useful context and make the proposed next action legible rather than forcing a person to reconstruct why it appeared.
Simplicity does not remove control
Everyday workflows can stay approachable while expert, admin, permission, policy, and diagnostic controls remain available to the people who need them.
One Principle, Role-Specific Work
Human-centered design follows the professional context.
The principle applies across every active edition. Each product family defines the familiar vocabulary, workflows, systems, review moments, and claim boundaries for its profession.
What This Principle Does Not Mean
Natural interaction must still be honest interaction.
- Do not hide that AI is involved.
- Do not present Ava or another assistant as a human.
- Do not remove source context, uncertainty, review, audit, or user agency for the sake of simplicity.
- Do not imply that every workflow, edition, connector, or consequential action operates autonomously.
Use AI naturally. Keep professional control explicit.
See how the principle fits the shared operating loop, then choose the edition designed around your professional environment.