The mission lifecycle
What happens between a goal and a result
Eight steps. You are in control at every one of them.
- 1
You describe the goal
Plain language, the way you would say it out loud. Optionally add a deadline, a budget, the time you have, a success metric, restrictions and resources you already own.
- 2
Parallel asks what it actually needs
Only questions whose answers would change the plan. Each one comes with a control (a choice, a slider, a date) and a line explaining why it matters. Never a long form.
- 3
It diagnoses before it plans
Parallel states what it understands, what constrains the goal, which routes exist, which one it recommends and which ones it is rejecting, with the reason for each.
- 4
One recommended plan
Not three options to choose between. One route, broken into day-anchored stages and measurable steps, with milestones, risks, assumptions, what it needs from you and what agents will handle.
- 5
Your agent team is built
Pressing the button generates a custom agent team from the plan. Each agent gets a name, a role, an objective, tools, required inputs, a schedule, success metrics and a permission level.
- 6
You set the boundary
Before activation, confirm what agents may do automatically, what must be approved, and what is prohibited outright. Credentials are never displayed again once saved.
- 7
The mission runs
Agents work through the schedule and produce deliverables. Anything consequential arrives in Approvals with the proposed action, the reason, the expected effect, the risks and a full preview.
- 8
It measures, adapts and reports
Planned against actual, every day. When results diverge, Parallel names the assumption that broke and changes the smallest thing that could fix it. Then it tells you what it changed and why.
What Parallel will and will not claim
A planning system that overstates its confidence is worse than no planning system.
It will say
- "Recommended": this is the route with the best expected return under your constraints.
- "Expected 4 to 7 replies": an estimate from measured data, not a promise.
- "Assumption: you can respond within a few hours."
- "Confidence: medium. Based on 27 messages."
- "Simulated": on anything the demo environment did not really perform.
It will not say
- “87% chance of success”: a made-up number dressed as analysis.
- "Guaranteed three clients."
- That an email was sent when it was only drafted.
- That it researched something it could not reach.
- That a plan is working when the numbers say otherwise.