Living landscape

Enterprise AI Adoption

Explore documented AI business activity, including adoption, partnerships and deployments—and see what each source actually establishes.

Published snapshotBuilt from current wiki evidence

Compare

Compare evidence

Choose 2–4 entities for a side-by-side matrix. Unknowns stay visible instead of being inferred.

4 of 4 selected

Decision view

What the wiki can establish

Evidence gaps remain visible as “Not established” instead of being filled by inference.

Use this matrix to compare published evidence—not to rank organisations or products. Each detailed claim links to its supporting wiki entry. Documented means that a source reported it, not that the wiki independently verified a causal outcome.

Business question1PasswordCyberAgentAtlassianPwC
What was rolled outNo source-linked detail in this snapshotNo source-linked detail in this snapshotNo source-linked detail in this snapshot
  • Microsoft's customer story says PwC rolled out Microsoft 365 with Copilot across the firm, with more than 200,000 Copilot licenses enabled and 40.8 million Copilot actions recorded over six months.Vendor-reported · Read entry 1
  • It reports more than 285,000 users in over 100 countries.Adopter-published · Read entry 1
Business workflow
  • 1Password deployed Codex across planning, implementation, review, testing, and production investigation.Vendor-reported · Read entry 1
No source-linked detail in this snapshot
  • Atlassian uses Claude for complex and long-running workflows in Rovo Chat, Rovo CLI, and early-access Rovo Max, while Claude Code is used within engineering and was piloted in non-technical teams.Vendor-reported · Read entry 1
  • Its April 2026 engineering account says the system used Anthropic Claude 3.5 Sonnet to generate review suggestions, alongside separate quality checks.Source-reported · Read entry 1
No source-linked detail in this snapshot
Reported result
  • Its OpenAI customer story reports a 20.9% productivity improvement for a core engineering cohort and a 10.9% reduction in median pull-request cycle time.Vendor-reported · Read entry 1
  • For its Finance AI Hackathon, 1Password says 90 team members built 21 agents and that those agents save the team 5–10 hours per person per week.Source-reported · Read entry 1
No source-linked detail in this snapshot
  • Atlassian reported a 30.8% reduction in median pull-request cycle time with its Claude-backed Rovo Dev Code Reviewer and 35.6% fewer human-written review comments.Source-reported · Read entry 1
No source-linked detail in this snapshot
Measurement basis
  • The company estimates a 553% ROI and about $784,000 of annual engineering capacity for a modelled group of 50 consistent users.Vendor-reported · Read entry 1
  • The ROI is modelled, not a realised cash saving or an independently verified causal effect.Vendor-reported · Read entry 1
No source-linked detail in this snapshotMeasurement basis included in reported result aboveNo source-linked detail in this snapshot
Limits and unknowns
  • Expansion to finance and marketing is described as a rollout, but the published productivity figures concern engineering and should not be generalized to those teams.Vendor-reported · Read entry 1
  • The ROI is modelled, not a realised cash saving or an independently verified causal effect.Vendor-reported · Read entry 1
No source-linked detail in this snapshot
  • The same story reports millions of Rovo Chat users but does not provide a controlled measure of Claude-specific productivity or customer outcomes.Vendor-reported · Read entry 1
  • Anthropic's customer story says more than five million AI agents ran in customer business workflows in one month, but does not attribute all those executions to Claude; Rovo uses multiple models.Vendor-reported · Read entry 1
  • It does not independently validate the adoption scale in Anthropic's customer story, nor establish a Claude-specific productivity outcome.Adopter-published · Read entry 1
  • These are vendor-published customer figures; the story does not provide an independent causal evaluation or the calculation method for the time-savings valuation.Vendor-reported · Read entry 1
  • The separate $25 million saving came from platform consolidation and should not be attributed to AI.Vendor-reported · Read entry 1
  • The page does not provide the calculation behind “capacity created”; usage volume is not the same as business value.Adopter-published · Read entry 1

Not directly comparable: reported results may use different measures, populations, baselines, methods or dates. Read each measurement basis and source before drawing a conclusion.

Snapshot classificationsContext, not cited per-cell claims

These classifications summarise each entity dossier; they are not individually cited claims. Open a linked claim above for source-specific detail.

Classification1PasswordCyberAgentAtlassianPwC
Adoption stageProductionProductionProductionScaling
Business functionsFinance, Software DeliveryNot establishedNot establishedNot established
Outcome categoriesCycle Time, ProductivityNot establishedCycle TimeProductivity
Source date2026-09-082026-06-182026-08-282026-02-25
Add or change entities53 available
  • 1PasswordOrganisation · 2 sources
    Remove 1Password
  • AARPOrganisation · 1 source
    Limit reached
  • AWSOrganisation · 1 source
    Limit reached
  • AccentureOrganisation · 2 sources
    Limit reached
  • AdobeOrganisation · 1 source
    Limit reached
  • Akash SystemsOrganisation · 1 source
    Limit reached
  • AmazonOrganisation · 1 source
    Limit reached
  • AnthropicOrganisation · 1 source
    Limit reached
  • Applied MaterialsOrganisation · 1 source
    Limit reached
  • AtlassianOrganisation · 3 sources
    Remove Atlassian
  • BBVAOrganisation · 2 sources
    Limit reached
  • Bristol Myers SquibbOrganisation · 1 source
    Limit reached
  • CertiniaOrganisation · 1 source
    Limit reached
  • ChocoOrganisation · 1 source
    Limit reached
  • CiscoOrganisation · 2 sources
    Limit reached
  • CoinbaseOrganisation · 1 source
    Limit reached
  • CreatioOrganisation · 1 source
    Limit reached
  • CrowdStrikeOrganisation · 1 source
    Limit reached
  • CursorOrganisation · 1 source
    Limit reached
  • CyberAgentOrganisation · 2 sources
    Remove CyberAgent
  • CyberArkOrganisation · 1 source
    Limit reached
  • DoorDashOrganisation · 1 source
    Limit reached
  • FacebookOrganisation · 1 source
    Limit reached
  • FluxOrganisation · 1 source
    Limit reached
  • FortinetOrganisation · 1 source
    Limit reached
  • Gilbert + TobinOrganisation · 1 source
    Limit reached
  • GitLabOrganisation · 2 sources
    Limit reached
  • GoogleOrganisation · 2 sources
    Limit reached
  • HPEOrganisation · 2 sources
    Limit reached
  • IBMOrganisation · 1 source
    Limit reached
  • ImpelOrganisation · 1 source
    Limit reached
  • IntuitOrganisation · 1 source
    Limit reached
  • MetaOrganisation · 1 source
    Limit reached
  • NetAppOrganisation · 1 source
    Limit reached
  • OpenAIOrganisation · 8 sources
    Limit reached
  • Oriole NetworksOrganisation · 1 source
    Limit reached
  • PermisoOrganisation · 1 source
    Limit reached
  • PictetOrganisation · 1 source
    Limit reached
  • PwCOrganisation · 2 sources
    Remove PwC
  • PythianOrganisation · 1 source
    Limit reached
  • QontoOrganisation · 2 sources
    Limit reached
  • QuantiumOrganisation · 1 source
    Limit reached
  • Red HatOrganisation · 1 source
    Limit reached
  • RedditOrganisation · 1 source
    Limit reached
  • Regal RexnordOrganisation · 1 source
    Limit reached
  • S&P GlobalOrganisation · 1 source
    Limit reached
  • Samsung ElectronicsOrganisation · 2 sources
    Limit reached
  • Sun LifeOrganisation · 1 source
    Limit reached
  • TecbanOrganisation · 1 source
    Limit reached
  • VercelOrganisation · 1 source
    Limit reached
  • WorkdayOrganisation · 1 source
    Limit reached
  • World Bank GroupOrganisation · 1 source
    Limit reached
  • ZapierOrganisation · 2 sources
    Limit reached