ManyAct is coming soon

Start with the outcome. Put a team behind it.

ManyAct is being built for people and AI teammates to work together. Begin with an outcome, an existing team or something that needs doing, then keep the people, knowledge, tasks and results connected.

Join early access See how it works

People stay in the work. They can request, contribute to, review or complete tasks while AI teammates research, analyse, create and pass work on.

Outcome Brief In Progress · 4 of 6

"Understand whether we should enter the European market this year."

Research market size & regulations Researcher (AI)
Provide budget & financial constraints Maya (Finance)
Analyse entry models & compare options Analyst (AI)
Review recommendations & decide You (Decision)
Assembled Team: Researcher Analyst Strategist Maya You

The basic idea

You don't draw a workflow.
You build a team.

Bring outcomes, context, people, AI teammates and their work together. Start wherever it makes sense. Define the team yourself, bring in specialists you already use, or let ManyAct help shape what the work may need.

Outcome Context + knowledge People + AI Work + results
ROLE What they're here to do.
KNOWLEDGE What they should know.
TOOLS What they can use.
SKILLS What they can apply.
What are we working towards? An outcome
MEMORY What should carry forward.
TASKS What needs doing.
MODEL The intelligence behind them.
INTEGRATIONS Which external systems they can work with.

Build the teammate, not the plumbing.

How ManyAct works

Define. Assemble. Act.

Begin with an outcome, a task, a teammate or an existing team.

ACT
Research
Analyse
Review
HUMAN
AI
AI
  1. 01

    Define what matters

    Create a shared place for the work. Add an outcome, useful knowledge, your tools and the context that matters.

  2. 02

    Shape the team

    Create teammates yourself, reuse specialists, or let ManyAct suggest who could help.

  3. 03

    Let the work develop

    Objectives and tasks can be assigned, discovered, delegated or refined as the work progresses.

  4. 04

    See what the team produces

    Follow activity, contributions and results while people remain part of the work.

Example teammates

The team is shaped by the work.

These are examples, not fixed personas. An Act can bring together the people and AI specialists that make sense for the work—and that combination can change as the work develops.

🔍

Researcher

Finds information, explores sources, surfaces evidence and synthesizes facts.

AI Teammate

🧪

Analyst

Tests assumptions, compares options, quantifies findings and spots edge cases.

AI Teammate

◇

Strategist

Considers alternatives, identifies risks, and structures reasoning with trade-offs.

AI Teammate

⚒

Builder

Creates artefacts, writes code, drafts materials, and makes things real.

AI Teammate

⟡

Coordinator

Tracks progress, sequences work, handles handoffs, and keeps tasks moving.

AI Teammate

●

You & your people

Contribute domain expertise, provide authority, review work and make key decisions.

Human Teammate

The product itself

A team you can understand.
Work you can inspect.

An Act can hold one or more outcomes, the team working toward them, and the tasks and results that develop along the way. Each AI teammate still has real configuration.

Act: European Market Entry Analysis Team active
R
AI TEAMMATE · TASK OWNER

Researcher

Working
Role

Find and verify market evidence

Active Task

Synthesise EU regulatory landscape

Tools

Web search · Document Parser

Knowledge

Product docs · Competitor database

Memory

Project context & preferences

Model

Configured reasoning model

Skills

Source evaluation · Fact synthesis

Integrations

Connected data warehouse

Illustrative product view showing an Act, its outcomes, tasks, mixed human and AI team, teammate configuration and activity.

Starting points

Different work calls for different combinations.

These are starting points, not prescribed teams or fixed ways to begin.

01 Research something properly
Example team

Research agent → Analyst agent → Human reviewer → Writer agent

Possible outcome

A thoroughly researched brief with sources, competing interpretations and a final synthesis.

02 Turn an idea into something concrete
Example team

Human lead → Planner agent → Builder agent → Reviewer

Possible outcome

A concept progressively developed, prototyped, and checked for feasibility.

03 Keep an eye on something
Example team

Watcher agent → Analyst agent → Human decision → Messenger

Possible outcome

Continuous monitoring with meaningful signals and changes surfaced when they matter.

04 Work through organisational knowledge
Example team

Human question → Retriever agent → Specialist → Writer agent

Possible outcome

Answers and deliverable work grounded directly in the organisation’s own information.

05 Run repetitive operational work
Example team

Coordinator agent → Specialist → Human checkpoint

Possible outcome

Tasks executed consistently with people participating where judgement and authority are required.

These are examples, not promises of autonomous perfection. The quality of the work still depends on the people, models, tools, information, instructions and oversight involved.

Scale

A team doesn't require a company.

One person

Build specialists around the work you already do.

Small team

Create repeatable support around shared work and knowledge.

Organisation

Create clearer roles and reusable capabilities across teams.

Working principles

A few things we're trying hard not to ruin.

01

Simple without pretending the problem is simple.

The machinery can be sophisticated. Using it should not require staring at complex orchestration graphs all day.

02

Autonomous doesn’t mean invisible.

If agents are doing work, you should be able to understand what happened and why.

03

Different jobs deserve different specialists.

One giant prompt and one generic assistant is rarely the right shape for real collaborative work.

04

Useful before impressive.

A team you trust with tomorrow's job matters more than a flashy five-minute demo.

05

Humans remain part of the team.

A person should be able to request work, contribute to a task, review outputs or take over whenever judgement matters.

Coming soon

We're building ManyAct now.

The first public version is still being built. If this sounds like a useful way to work with AI, leave your email and we'll let you know when there's something worth trying.

AI teams are still only as good as the models, knowledge, tools, instructions and judgement around them. ManyAct is being built to make those pieces easier to bring together.