Research
140 reports, datasets and source records
Research reports
- Adversarial evidence on absent, negative, and overstated AI benefits
- AI cost economics — research for Knauf Digital
- Executive synthesis
- AI cost governance: measure the work, choose the intervention
- Optimisation landscape
- Worked examples and teaching boundaries
- Financial evidence for application enablement
- Broader evidence for the opening
- Prompt caching: absolute API cost evidence
- Review-transcript evidence audit — 28 October 2026
- Second review: research answers and editorial decisions
- Third review: evidence and editorial decisions
- Enterprise AI budget: chart assumptions and research
- Agent economics: evidence audit for the Knauf opening
- Independent audit of the proposed opening facts
- Enterprise evidence for AI cost control
- Unit prices and work per result: evidence audit
- AI cost optimisation: govern spend, improve execution, prove value
- Optimise the cost of a completed task, not a token in isolation
- Give each AI workload an owner, an envelope and an escalation rule
- Connect operational telemetry to invoices and business outcomes
- Choose the intervention that improves accepted output per euro
- Select the operating model from demand, constraints and lifecycle cost
- Prove value with matched outcomes and a credible counterfactual
- Source register: use current primary evidence without overclaiming
- Compare eligible deployments by fully loaded cost per accepted task
- Navigara adds a useful engineering-spend lens, but does not close the task-cost or hosting-TCO evidence gaps
- Navigara public-page inventory and coverage
Data and calculations
- 2026-10-22-ancillary-pricing
- 2026-10-22-measurement-schema
- AI optimisation decision record
- 2026-10-22-provider-pricing
- 2026-10-22-self-host-sensitivity
- 2026-10-22-worked-example
- 2026-10-27-financial-chart-data
- 2026-10-27-leadership-trends
- 2026-10-28-caching-benchmark
- 2026-10-28-cost-scenarios
- 2026-10-28-cost-trends
- 2026-10-28-third-review-chart-data
- 2026-10-29-enterprise-ai-budget-allocation
- 2026-10-22-agent-economics-claims
- 2026-10-22-claim-ledger
- 2026-10-22-adaptive-budget-comparisons
- 2026-10-22-claim-ledger
- 2026-10-22-fixed-threshold-price-chart
- 2026-10-22-o1-sampling-chart
- 2026-10-22-source-manifest
Sources and evidence
- 2026-10-26-ai-benefit-source-register
- Source index
- 2026-10-22-source-index
- 2026-10-22-source-register
- 2026-10-28-caching-benchmark
- 2026-10-28-review-evidence
- 2026-10-28-second-review-sources
- 2026-10-28-third-review-sources
- 2026-10-29-enterprise-ai-budget-allocation
- A01: Pyramid Principle pack
- A02: Dreambase Data Stories
- A03: Dreambase Data Presentation
- A04: PPT Prepare
- A05: Consultant
- A06: Presentation Master
- A07: Slidev Presentation Design
- A08: Enterprise AI Skills: storyline, critic and deck pipeline
- A09: McKinsey Problem-Solving Skills: Build a storyline
- A10: MBB Decks
- A11: McKinsey PPT Design
- 2026-10-22-adoption-audit
- 2026-10-22-sources
- C01 — OpenAI API pricing
- C02 — Claude Platform pricing
- C03 — Gemini Developer API pricing
- C04 — Prompt caching
- C05 — Batch API
- C06 — Reasoning models
- C07 — Context caching
- C08 — NVIDIA NIM LLM benchmarking overview
- C09 — Lambda on-demand GPU instances
- C10 — How to Build a Generative AI Cost and Usage Tracker
- C11 — Optimize resource consumption to minimize hosting costs
- C12 — Architecting generative AI applications for production
- C13 — OpenAI API changelog: GPT-6.1 Sol release
- C14 — MLPerf Inference Datacenter benchmark methodology
- C15 — Introducing the MLPerf End-to-End RAG Inference Benchmark
- 2026-10-22-sources
- 2026-10-27-sources
- G01 — FinOps for AI Overview
- G02 — Managing AI Value Using FinOps Practice Operations
- G03 — Unit Economics
- G04 — Allocation
- G05 — Effective Cost — FOCUS Specification 1.4
- G06 — Consumed Unit — FOCUS Specification 1.4
- G07 — Tutorial: Create and manage budgets
- G08 — Managing your costs with AWS Budgets
- G09 — Usage and Cost API
- G10 — How the agent loop works
- G11 — Semantic conventions for generative client AI spans
- G12 — Semantic Conventions for GenAI agent and framework spans
- G13 — Shadow AI discovery in Global Secure Access
- G14 — Usage Optimization
- G15 — FOCUS Specification 1.4
- Governance source register
- 2026-10-22-sources
- K01 — Knauf Digital mandate
- K02 — Product & System Data work
- K03 — myKnauf workflow context
- K04 — knauf.com relaunch
- K05 — Multilingual document access
- K06 — Building-performance ML example
- K07 — Supply-chain AI context
- K08 — Current group scale
- 2026-10-22-sources
- 2026-10-27-sources
- O01: Building effective agents
- O02: Latency optimization
- O03: RouteLLM: Learning to Route LLMs with Preference Data, v4
- O04: Automatic Prefix Caching
- O05: Prompt caching
- O06: Batch API
- O07: How we built our multi-agent research system
- O08: Evaluation best practices
- O09: Quantization
- O10: Batchers
- O11: Optimization and Tuning
- O12: Speculative Decoding, v0.21.0
- 2026-10-22-sources
- P01: The Compelling Communicator: An interview with McKinsey’s Julie Goran
- P02: The Minto Pyramid Principle Concept
- P03: The Minto Pyramid Principle Course
- P04: How the Duarte Big Idea Formula Can Make Your Presentation the Best
- P05: Advanced stickynoting
- P06: How the Design of Presentation Slides Affects Audience Comprehension
- P07: Principles for Reducing Extraneous Processing in Multimedia Learning
- P08: Slidedocs
- P09: Advancing the Knowledge Base on Effective Presentation Slide Design: Three Pilot Studies
- P10: The equity story you need for the long-term investors you want
- P11: The 10/20/30 Rule of PowerPoint
- 2026-10-22-sources
- 2026-10-09-knauf-navigara-page-inventory