Designing Projects Persistent Context and Team Collaboration for an Platform

Projects is a persistent workspace for one ongoing effort, so context, knowledge and teams stay together.

Overview

Projects is a persistent workspace for one ongoing effort, built into an AI for work platform that brings together agents, workflows, search, memory and knowledge. It keeps chats, files, instructions, memory and agents in one place, so context carries over from chat to chat and teams can share it.

One-line summary Chat is great for single questions but breaks down for ongoing work. I designed Projects so people can keep context, knowledge and AI behavior together instead of re-explaining everything in every new chat.

Quick Facts

My Role: Product Designer
Time Line: 26 Days
Platform: Web & Native
Domain: Enterprise AI
Team: 1 PM + 2 Designers + 5 Engineers + 3 QA
My Contribution: Research, Interaction & UI Design, Dev Handoff and Usability Review

User Pain Points

In a plain chat product every chat starts from zero, so people doing ongoing work repeat themselves, lose track of earlier work and stop trusting what the AI knows.

Pain point	What users say	Impact	How Projects solves it
Repeating context	I paste the same background into every chat.	Wasted time, inconsistent answers	Project knowledge and instructions apply to every chat
Scattered work	I can’t find the chat where we worked this out.	Hard to find and resume work	Chats grouped in one project, with an overview to resume from
Inconsistent output	I have to tell it the tone and format every time.	Rework and heavy editing	Project instructions with starter templates
Working alone	My team can’t use the same context I set up.	Duplicate effort across the team	Shared projects with roles and shared knowledge

Business Need

Projects let people keep working on the same thing, so they stay longer, bring in teammates and become worth paying for.

Help the business in four ways:

Keep users: Context that carries over gives people a reason to return.
Grow teams: Shared projects bring colleagues in, so usage spreads.
Earn more: Teams relying on shared context need more seats and controls.
Stand out: Agents, workflows and live knowledge go beyond ChatGPT and Claude.

Problem

People at work have ongoing efforts, not one-off questions, and plain chat makes them start over each time

Context loss

every new chat starts from zero, so people re-paste the same background and re-upload the same files before they get a useful answer.

Inconsistent output

tone, format and rules have to be re-stated each time, so the same task gives different results from one chat to the next.

No shared ground

team mates cannot work from the same knowledge and instructions, so each person builds a private version that drifts from the rest.

How might we let people organize AI work around ongoing efforts, so context persists, knowledge is reusable and output stays consistent, without adding setup effort that feels like admin?

Research and Competitive Analysis

I studied how ChatGPT Projects and Claude Projects frame the same idea, then compared them to my product’s needs.

Project as a container for chats, files and instructions.
Project-level instructions that shape every conversation inside.
Knowledge files that the AI can draw on across chats.
A sidebar entry point with project chats listed under the project.

Users and Insights

Projects serve three primary personas and one secondary persona. I ranked them by how often they use projects and how much they influence adoption.

Operator

Produces the same weekly outputs from the same documents and format.

Goals

get the same format, tone and structure every week without rewriting prompts. Have her rules for sources, wording and checks apply automatically in every chat. Start a task in seconds, with background and files already in place, and spend her time reviewing output instead of rebuilding context.

Team Lead

Leads a team where new people join and others step away mid-effort.

Goals

have everyone on the team work from the same instructions and knowledge. Onboard a new member in minutes, with the full context already there, and keep output consistent in tone and format across the team. Hand over work without losing context when someone steps away.

Workspace Admin

Responsible for what is shared, with whom, and outside the workspace.

Goals

set safe sharing defaults so teams can start without waiting for approval. Limit sharing outside the workspace and restrict public links, and see who has access to each project and what changed through an audit log. Avoid orphaned projects by transferring ownership when an owner leaves.

Use case Set up and share a weekly report project

This use case shows the full path from creating a project to a team using it, and ties together the personas, flows and sharing rules in this case study.

Primary actor → Team Lead
Goal → Give the team one place where the weekly report is produced from the same sources, rules and format every time
Trigger → The team keeps re-pasting the same background and producing reports in different formats
Preconditions → Priya has a workspace account and can create projects; sharing is enabled by the admin; the report's source files or connected sources exist
Postconditions → A shared project exists with instructions, knowledge and members; Maya can produce the report in minutes and every chat follows the same rules

Key Insights

I keep pasting the same background into every chat → persistence is the core value.
I don't know what the AI actually remembers → transparency builds trust.
Setup should not feel like configuring software → progressive disclosure.
My team needs the same answers I get → shared context.

Design Principles

They kept the design focused on what people need from ongoing AI work, and I used them as a test whenever I compared options.

1

Start simple scale with depth.

Let users create a project with just a name, then add context and structure as their needs evolve.

2