AI-Wiki

A leader's guide to data strategy in the era of agentic AI

5 min read

A leader’s guide to data strategy in the era of agentic AI

In Formula 1, victories depend on turning millions of data points into winning decisions. Organizations can apply this same precision to unlock value from underutilized data assets. Drawing from F1’s model, we present a three-part framework for data excellence: customer-focused decisions, dynamic people-powered strategies, and market-responsive structures.

— AWS Events YouTube channel

A ~34-minute executive-forum keynote from the AWS Summit Sydney, delivered by an AWS Enterprise Strategist (a self-described “former CIO”; not named in the transcript), using Formula 1 as the running analogy for data excellence. It is the data-strategy companion to the wiki’s AWS team-structures sources (Allen / Brovich), and the wiki’s most focused single source on re-architecting enterprise data for machine consumers.

TL;DR

  • The data consumer has changed: machines, not humans. Four eras of data strategy: BI (reports) → big data (lakes) → generative AI (content) → agentic (machines are the primary consumer). “Agents need metadata to reason. They need memory to learn. They need context to act.” Data formatted for humans — bold, italics, indentation, PDFs — is overhead to a machine (more tokens, no signal). Stripe is cited converting human docs to markdown so agents can reason (headers→hierarchy, lists→arrays, links→navigable).
  • The sobering stats. Gartner: 80% of data-governance initiatives will fail by 2027. HBR: 99% of organisations invest in data but only 29% see meaningful value. 92% of agentic-AI users worry about data quality/governance; half of data leaders say poor data quality blocks AI implementation completely. Diagnosis: organisations “prioritised data volume over data value” — celebrating petabytes collected instead of value delivered.
  • The framework: Reimagine → Rewire → Remember.
    • Reimagine — challenge every assumption about what data is worth. Have a business strategy informed by AI, not an “AI initiative” chasing technology. Four questions every data investment must pass or be rejected: (1) How will it create value? (2) Is this the most direct path? (3) How quickly can we deliver? (4) Can an agent consume this without a human translating? “If a human needs to interpret the data, you have failed the test.”
    • Rewire — restructure teams and ownership. Data products must be purpose-built, agent-ready, metadata-enriched, with clear ownership at the most granular level (“shared responsibility = shared neglect”; the CDO is “a terrible place to put ownership”). Decentralise teams by default, centralise only what speeds you up; move data engineers into the business units where decisions happen. F1 analogy: the driver gets 8 numbers, the pit wall gets 200, the factory gets 2,000 — all purpose-built from the same source.
    • Remember (value) — strategy is “what you do every day,” not a document. Don’t gather-clean-store-then-lake (“$50M, 5 years, one fired CIO, no value”); instead identify a business problem → get the data that solves it right → lather, rinse, repeat. Audit decision delays, pick 2–3 high-impact decisions, make them real-time. McKinsey: across 200+ managers, fast + high-quality decisions outweighed almost anything else in value creation.
  • What agents need from data — three things: (1) machine-readable structure (parse without guessing), (2) semantics (relationships between entities — extract people/places/time, then their associations; a knowledge-graph framing), (3) memory across agents and interactions (so agents learn). Traditional data gives the what; agentic AI demands the why — “the more why we can give it, the more autonomous we can permit these agents to be.”
  • Minimum viable governance — guardrails, not roadblocks. Strict lockdown “creates the problems it was meant to prevent” — people go underground and build without safeguards. Instead: open by default (vast majority open within the org; a limited slice protected; only a small slice restricted), real-time policy enforcement and audit trails built-in not bolted-on, eight data tenets (customer-value-drives-investment, minimise decision distance, default-open access, clear ownership, trust-through-monitoring, automation — “if it’s not automated, it is not done” — and real-time everything). “You can’t audit your way to trust; agents need trust at speed.” Governance should accelerate value like an F1 safety car manages risk without stopping the race.

What was actually ingested

The full ~34-minute keynote (ASR transcript, ~319 lines, light proper-noun cleanup). The presenter is an AWS Enterprise Strategist / former CIO; the ASR did not capture a self-introduction, so attribution is to the AWS Events channel with the presenter named generically in body. F1 is a rhetorical scaffold throughout (AWS is the F1 title sponsor); the load-bearing content is the data-for-machines thesis, the four-question test, data products, and minimum-viable-governance.

Why this matters to the wiki

This is the data/context substrate beneath the wiki’s agent cluster. It operationalises, from the data-engineering angle, claims the wiki holds elsewhere: Argenti’s “AI transformation follows data transformation”; the harness Context layer (data products as the agent-consumable, metadata/trust-signal-enriched input — the same problem Chopra’s Headroom solves at the compression/routing layer); the knowledge-graph construction story (LLM entity-extraction → associations → context). The “can an agent consume this without a human translating?” test is a crisp, quotable addition to enterprise-ai-adoption.

Dynamic-capabilities (W&W) reading

  • digital-transforming/redesigning-internal-structures — decentralise data teams by default, move data engineers into business units, granular data-product ownership.
  • digital-transforming/improving-digital-maturity — the Reimagine→Rewire→Remember maturity path; data products with contracts/metadata/quality signals.
  • digital-seizing/strategic-agility — real-time-data over batch, “evolution not revolution,” continuous testing and learning over annual planning.
  • contextual/internal-enablers — minimum-viable-governance, the eight data tenets, and metadata/lineage as the infrastructure that enables agentic value.

Linked entities and concepts

Relationships

Relationships

Supports

Graph View

Backlinks