System Atlas

Training data acquisition and licensing → pretraining and fine-tuning runs (information flow)

Workflows that drive an outcome from step to step. Process steps that hand work to each other.

Within one industryHolds in Artificial intelligence. Counted from public, sourced parts only; each organization below links to the parts it was counted on.

Organizations running it
5
Of their industry
36%
The pattern
What every instance has, in order
  1. Data
    Training data acquisition and licensing
    in 100% of instances
    information
  2. Process
    Pretraining and fine-tuning runs
    in 100% of instances
    then
  3. Decisions
    Market and positioning
    in 100% of instances
    then
  4. Process
    Lead generation
    in 100% of instances
    then
  5. Process
    Product or service design
    in 100% of instances
    information
  6. Process
    Production or delivery
    in 100% of instances
    then
  7. Process
    Hiring
    in 100% of instances

Coloured arrows: a flow of that type seen between the two at instance organizations. Grey arrows: order only.

Structure
Element
Share of instances
Companies
Training data acquisition and licensing
100%
5
Variant · where they differ
Member
Variants, couplings and instances

Every level, number, source and dimension: Member, $300 a year.

Couplings · what comes together
Member
Variants, couplings and instances

Every level, number, source and dimension: Member, $300 a year.

Entropy
Where it runs rough

Operating-reality findings at instance companies: workarounds, manual hand-offs, bottlenecks.

  • Training data acquisition and licensing4 findings
    Dependency 2 · Manual hand-off 1 · Workaround 1 · at 4 companies
  • Production or delivery2 findings
    Bottleneck 1 · Dependency 1 · at 2 companies
  • Lead generation2 findings
    Dependency 1 · Informal channel 1 · at 2 companies
  • Suppliers and partners2 findings
    Dependency 2 · at 2 companies
  • Communication tools1 finding
    Tacit rule 1 · at 1 company
  • Customer onboarding1 finding
    Dependency 1 · at 1 company
  • Health, safety and environment1 finding
    Workaround 1 · at 1 company
  • Inference serving1 finding
    Workaround 1 · at 1 company
Seen across
Where it holds
Business model
  • product20%
Size
  • 11-50 people20%
  • 201-500 people20%
Region
  • Germany / United States20%
  • San Francisco20%
  • United States20%
  • United States (founded in the US); team distributed across Europe and North America20%
  • United States (Silicon Valley, California; Pennsylvania) and India (Bengaluru)20%
Instances
The 5 companies it was seen at
Member
Variants, couplings and instances

Every level, number, source and dimension: Member, $300 a year.