Data owners

Understand what your organisation is already producing.

Support threads, code reviews, teaching interactions, lab records, quality judgements: ordinary operations produce material that may contain learning signals AI organisations ask for. bnd exists to find out, confidentially, whether yours does.

Discuss a Data ReservoirMap a Potential Reservoir
01 / The engagement

What a discovery engagement looks like.

No raw data moves at the start. Most of this is conversation and description, and you can stop at any point.

01

A confidential conversation

What your organisation does, which processes repeat, and what records they leave behind. Nothing transferred.

02

Reservoir inventory

Together we list the candidate reservoirs: where sequences of failure, correction, intervention, and outcome may live.

03

Sample definition

You decide what may be reviewed. A small, controlled sample is enough to test whether a signal exists.

04

AI-assisted description

Inspection produces a structured description of what the source contains: recurring patterns, rare cases, gaps. Provisional by design.

05

Demand comparison

The description is compared with what AI organisations currently request. Sometimes the honest answer is that demand does not value it yet.

06

A qualified package

If demand and material meet, bnd proposes a package: scope, rights, evidence, delivery, and payment. You approve or decline.

02 / Owner controls

Set the rules once. Value will move within them.

Owners are not asked to review every sentence or negotiate every excerpt. The rail is being designed around standing permissions.

Define permitted purposesStanding rule
Exclude sensitive categories by defaultStanding rule
Require anonymity or attributionStanding rule
Choose manual review or automatic licensing within your rulesOwner mode
Set minimum commercial termsStanding rule
See every completed transactionAudit
Revoke standing permission for future matchingControl

The rail is being designed with two permission stages: standing permission for private matching inside the rail, and release permission before any package leaves. Both are owner decisions.

03 / Rights

The practical unit is a clear rights package.

Buyers cannot predict every downstream experiment, and owners cannot review every future use. What works is truthful permissions, clean provenance, and enforceable boundaries.

PurposeCommercial AI model training and evaluation
IdentityAnonymous
Derivative modelsPermitted
Raw redistributionProhibited
RetentionDefined term
PaymentTo the owner, with a percentage toll to bnd

Example rights package: illustrative, not a standard contract

04 / Readiness

What makes a reservoir strong.

  • Provenance: you can say where the material came from and under what terms.
  • Signal structure: the records capture failure, correction, intervention, or outcome, not only final results.
  • Outcome evidence: confirmation of what finally worked is increasingly a buyer requirement.
  • Rights clarity: your organisation controls the material, or knows exactly which parts it controls.
  • Volume or rarity: scale matters for some requirements; rare, hard-to-recreate material matters for others.

Not every archive contains a valuable signal, and no data is guaranteed to have commercial value. The diagnostic and the discovery conversation exist to find out honestly.