Core Team

Agentic AI has an interface problem.

Before advanced work can begin, people are often asked to choose models, assemble context, connect tools and manage permissions.

What if the interface began with the way people already understand help?
A creative professional overwhelmed by several devices and a cluttered work desk
The same creative professional calmly working with paper sketches, with one phone on the desk
Core Team3 agents working on 5 tasks

Meeting scheduled at 5:00 PM

The consequences

What breaks in today’s general-purpose AI applications?

ChatGPT, Claude and similar blank-chat interfaces look simple, but dependable agentic work still rewards the people who know how to configure the system behind them.

01Advanced capability stays hiddenMany people remain in basic question-and-answer chat.
02Every new task begins againRole, context, tools and boundaries must be rebuilt.
03Results depend on configuration skillPeople who understand the system receive more capable outcomes.

If people go to different experts for different tasks, why should every AI task begin with the same blank chat box?

One person choosing between financial, repair, language and vacation-planning experts

Hide the system.
Show the work.

Early concepts exposed a marketplace, chat and a dashboard as separate places to learn. Reducing the model to Find, Brief and Track lets people begin with the work, while the system stays underneath.

1

Find the right help

Browse specialists by role and capability.

Core Team Discover screen showing specialists
2

Explain the situation

Send a message, photo or voice note.

A conversation with Marco, the cooking specialist
3

Track what happens next

See progress, approvals and completed work.

Core Team dashboard showing progress and decisions
Future possibilities

Where this model could go next.

Core Team could help democratise advanced AI: useful specialist capability becomes accessible without requiring technical fluency.
01An accessible specialist marketplaceAnyone could find trusted help by need, evidence and boundaries—not by technical setup or prompt skill.Needs clear quality standards, pricing and accountability.
02Specialists that become more efficientWith consent, recurring work could take less explanation as a specialist learns selected preferences and routines.Needs visible memory controls, correction and selective forgetting.
03Community knowledge that compoundsCommunities could build shared local databases that improve answers and reduce repeated work, while personal context remains separate.Needs consent, provenance, privacy and moderation.

Hi, I’m Chinmay Phatak.

I’m a UX designer exploring how agentic AI could become useful without asking people to assemble and operate complex systems.

I have experienced that setup burden myself and seen the same barrier keep other people inside basic chat. Core Team is my attempt to design a more familiar doorway.

Core Team is a self-initiated design project; the demo is a working prototype built to test the idea.

See the Demo Contact me