Onboarding

Onboarding is the first-run flow that gets TensorPM from a blank app into a first usable project.

Verified current flow:

1. Welcome screen.
2. TensorPM asks what it should call you.
3. You choose the assistant activity level.
4. You enter one project prompt.
5. TensorPM generates the first project from that prompt.
6. After project generation, you can open the project or configure AI for continued work.

Onboarding is not the same as the full project creation wizard. The wizard is a separate structured project setup flow.

What the user actually enters

The important project input in onboarding is a plain-language project prompt.

Example:

We are launching a small customer portal for our SaaS product. The goal is to let customers manage invoices, users, and support requests. We want an MVP in eight weeks with a small engineering team and a limited design budget.

TensorPM turns this prompt into an initial project structure.

Name step

TensorPM first asks what it should call you. This sets the local display name used for assistant responses.

You can provide a name or skip the step. This is personalization only; it is not project context.

Activity level

After the name step, TensorPM asks how active the assistant should be.

The current levels are:

• On standby: TensorPM steps in only when asked.
• Daily dreaming: TensorPM can run a daily project review.
• Dreaming: TensorPM can use stronger background assistance for progress, priorities, and attention points.

This controls settings such as daily project AI and background AI.

Optional AI configuration

During onboarding, there is an optional Configure AI path.

It can be used to configure:

• TensorPM account login or registration
• local AI endpoint
• direct provider API key

This is optional. The main onboarding flow can continue with TensorPM's onboarding path. Provider setup can also be done later in Settings -> AI.

Project prompt step

After activity selection, onboarding moves to the project prompt step.

One prompt is enough to start; TensorPM derives the initial context, structure, and assumptions from it.

Write enough for TensorPM to infer the basic shape:

• what the project is
• intended outcome
• important deadline or timeframe
• budget or resource constraint if known
• team or stakeholders if relevant
• known risks or unknowns

You do not need to fill structured fields manually in onboarding.

Follow-up behavior

The onboarding assistant is designed to generate quickly.

It may ask one concrete follow-up question when a missing detail would materially improve the project. It should not turn onboarding into a long interview. After at most a small number of follow-ups, it generates with reasonable assumptions.

What TensorPM generates

From the onboarding prompt, TensorPM creates an initial project profile with structured context such as:

• project description
• goal
• scope
• success criteria
• timeframe
• budget if inferable
• requirements
• methods or technologies
• milestones
• dependencies
• risks

The generated project is a starting point, not a finished plan.

After onboarding

After the project opens:

1. Review Context.
2. Correct assumptions in goal, scope, timeframe, and budget.
3. Add or refine Action Items.
4. Run Guidance only after the core context is believable.
5. Configure AI in Settings -> AI if you want BYOK, local AI, or account-based provider access for ongoing work.

What onboarding does not do

Onboarding does not:

• ask the user to complete the 11-step wizard
• import documents
• choose Empty Wizard / From Prompt / From Document modes
• require workspace creation
• replace reviewing the generated project

Those flows exist elsewhere in TensorPM.

Good onboarding prompts

Good prompts include constraints.

Plan a 200-person company summer party in Munich for July 2026. Budget is EUR 15,000. We need venue, catering, agenda, approvals, invitations, and risk handling for weather and attendance.

Restore a 1967 Mustang Fastback to roadworthy show condition by December. Budget is around EUR 45,000. Work includes engine, paint, interior, electrical, parts sourcing, inspection, and classic-car documentation.

Weak prompts are too vague:

Make a website.

Better:

Create a bilingual marketing website for TensorPM with product pages, documentation, pricing, blog, legal pages, and download links. It should be ready for a public launch and support English and German.

Next steps

• Full structured setup: Project Creation Wizard
• Understand generated fields: Project Structure
• Configure providers later: Account & AI Modes