Software platform needs discovery, end to end, in two rounds.
Proven Discovery is built around one belief: the best platform decisions come from real conversations with the people closest to the workflow. The platform automates the parts that do not need a human (scheduling, transcription, synthesis) and keeps you in charge of the parts that do (scope, decisions, sign-off).
The roles
Two roles. Different tools, different links, different responsibilities.
You. The person running discovery.
- Creates the project and writes the platform needs statement
- Adds stakeholders individually or via bulk import
- Activates the project, triggers synthesis, shares for review
- Sees every transcript, summary, status, and diagnostic
The people you are interviewing.
- Receives a private, tokenized link by email
- Talks to Claude in voice or text, on their own time
- Reviews the generated platform brief and leaves section level feedback
- Never logs in. Never sees the admin dashboard.
Frame the platform need
Vague platform needs statements produce vague interviews. The intake chat exists to sharpen the framing before a single stakeholder is invited.
Brainstorm
Open the platform intake chat and talk through the workflow, current systems, and software need. Claude pushes back, asks clarifying questions, and helps you write a platform needs statement worth interviewing against.
Add context
Drop in process docs, requirements notes, customer transcripts, internal write-ups, or vendor research. The model uses them to ground its questions and to suggest whom you should be interviewing.
Promote to project
When the brief feels ready, click Create Assessment from Brief. Company, platform needs statement, and a starter list of stakeholders pre-fill the project form. Edit and save.
Round 1: current-state platform discovery
The first conversation is for understanding the current workflow, current tools, data handoffs, and platform gaps. Stakeholders are encouraged to describe how things actually work today.
Activation
When the project moves from Draft to Active, every stakeholder gets a personal interview link. Links are tokenized, single-session, and unique to that person.
The conversation
Claude follows the conventions of a senior discovery interview. It probes, summarizes, asks for examples, and changes course when something interesting comes up. Stakeholders can use voice or type.
Length and style
The default target is six core questions. Most interviews run 10 to 15 minutes. Stakeholders see a soft progress indicator, not a question count, so they answer naturally.
Status tracking
The admin dashboard shows who is invited, in progress, or done in real time. When all stakeholders complete Round 1, the project moves to Round 1 Complete.
Does not pitch features. Does not ask leading questions. Does not require a stakeholder to know the answer before they can move on.
Round 1 synthesis
When every Round 1 interview is complete, you trigger synthesis. Claude reads every transcript and produces a structured summary of the current platform landscape.
What gets produced
Recurring themes, current tools, workflow gaps, integration gaps, data pain, and notable signals. The summary highlights where stakeholders agreed and where they did not. Stored on the project and editable.
Why it runs as a separate step
Round 2 questions are generated from the Round 1 summary. Treating synthesis as its own checkpoint lets you read it, sanity check it, and re-run it if a stakeholder needs to be re-interviewed.
Resilience
Synthesis retries up to three times on transient failures. If it still fails, the error is surfaced in the project diagnostics with enough context to act on.
Anonymization
If anonymization was enabled at project creation, names are stripped from synthesis output and exports. The setting is locked once the project leaves Draft status.
Round 2: platform requirements
The second conversation is grounded in Round 1. Claude asks questions informed by what stakeholders actually said, then defines the capabilities, constraints, and success criteria the platform needs to support.
Informed prompts
Each Round 2 conversation is seeded with the Round 1 themes, so the model can ask, in context: what would the platform need to support, and what would have to be true for it to actually work.
Tradeoffs and edge cases
Round 2 deliberately surfaces conflict. It asks where stakeholders would push back, what they would sacrifice, and what they refuse to compromise on. This is the data that makes the platform brief honest.
Same flow, different focus
Same tokenized link experience, same voice or text, same length. Stakeholders who have not finished Round 1 cannot start Round 2.
Resending and reminders
The Resend Invites action sends the right round to each stakeholder automatically. Stakeholders who have already completed their current round are skipped.
Platform brief generation
Once Round 2 wraps, you trigger synthesis a second time. Claude reads every Round 2 transcript plus the Round 1 summary and produces the full platform requirements brief.
- •Platform need and current state
- •Stakeholders, users, and roles
- •Functional and non-functional requirements
- •Constraints, dependencies, and assumptions
- •Open questions and contested points
- •Success metrics and acceptance criteria
Citable, not just summarized
The brief references where requirements came from. You can see which stakeholders raised what, where there was consensus, and where positions conflicted.
Editable
The output is yours. Edit it inline, regenerate sections, or export and modify in your tool of choice. The platform tracks your edits separately from the generated text.
Stakeholder review
Sharing the platform brief for review gives every stakeholder a second tokenized link. They read the document, leave section level feedback, and you decide what to act on.
They agree with this section as written.
An observation or note. Non-blocking.
They disagree or want something addressed before approval.
A specific proposed change to the text.
A heat map of the platform brief: which sections are clean, which sections have comments, which sections have unresolved concerns. Approve when it is ready. The system notifies stakeholders the platform brief has been finalized.
Status and diagnostics
Every project has a clear status and a diagnostics view that explains, in plain English, what is blocking the next step.
If a project gets stuck because of a transient failure or a withdrawn stakeholder, you can override the status manually. We recommend reading the diagnostics first. Most stuck states have a real cause.
Settings and configuration
Sensible defaults out of the box. A small set of knobs when you need them.
question_targetdefault 6, range 3 to 20How many core questions Claude aims for per interview round. Higher means deeper conversations and longer interviews.
default_anonymizedefault falseWhether new projects start with anonymization on. Useful in regulated or politically sensitive orgs.
email_templateoptionalCustom invitation email copy. Falls back to the default template if not set.
signature_urloptionalImage URL appended as a signature on outbound emails.
Privacy and data
Discovery conversations are sensitive. The platform is designed to keep them that way.
Tokenized stakeholder links
Stakeholders authenticate by single-use interview tokens and persistent review tokens. Neither requires a Clerk session. Tokens are unguessable.
Optional anonymization
When enabled at project creation, names are stripped from synthesis and exports. This setting cannot be changed once the project leaves Draft status.
Row level security
The Supabase database enforces row level security on every project, stakeholder, and feedback record. Admin operations use a single shared API key.
Full export, no lock-in
Download the full project as JSON at any time. Includes transcripts, both synthesis outputs, stakeholder summaries, and review feedback.
Ready to run your first project?
Early access is opening in waves. Join the waitlist and we will reach out when your spot is ready, usually within a week or two.