Chapter 08 of 9
Quick Reference
The LegacyForward.ai framework at a glance — summary tables for Signal Capture, Grounded Delivery, Legacy Coexistence, and Core Principles.
4 min read
Overview
This chapter consolidates the LegacyForward.ai framework into summary tables — one reference point for the core questions, phases, gates, patterns, and principles across all three pillars. Use it alongside the detailed chapters, not instead of them.
The diagram below illustrates the cost-of-delay principle that underpins the Kill Early discipline: the later an initiative is terminated, the more it costs to stop it, and the less value the organization recovers.
The LegacyForward.ai Framework at a Glance
| Pillar | Core Question | Stages / Phases | Gate Types | Key Anti-Patterns |
|---|---|---|---|---|
| Signal Capture | Where does this create net new value that we cannot achieve any other way? | Hypothesis → Validation → Tracking | GO/NO-GO; GO/PIVOT/KILL; Kill Triggers | Adoption Trap; Sunk Cost Spiral; Vibe-Coded Commitment; Perpetual Pilot; Automation as Transformation |
| Grounded Delivery | Are we delivering toward the value hypothesis, or toward a definition of done that does not measure value? | Frame → Explore → Shape → Harden → Operate | GO/NO-GO; GO/PIVOT/KILL; GO/REDESIGN; GO/CONTINUE/KILL; Operate Forever | Frame Skipped; Velocity as Progress; Test Generation Illusion; Sunk Cost at Gates; Ship and Forget |
| Legacy Coexistence | How do we create AI value in the enterprise environment we actually have? | Data Exhaust → Sidecar → Gateway → Shadow Pipeline → Legacy-Aware Agent (pattern complexity increases) | Pattern selection gate; Integration validation gate; Trust graduation criteria | Greenfield Fantasy; Wrapper Illusion; Integration Afterthought; Screen Scraping Default; Strangler Fig Misconception |
Signal Capture: Value Assessment at a Glance
| Stage | Primary Question | Key Activities | Gate |
|---|---|---|---|
| Hypothesis | Can we state the value in one sentence? | Value statement; Transformation test; Value ceiling estimate | GO / NO-GO — no hypothesis, no funding |
| Validation | Is the hypothesis real, or plausible? | Data validation; Feasibility validation; Organizational validation; Economic validation | GO / PIVOT / KILL |
| Tracking | Is value being captured? | Leading indicators; Lagging indicators; Kill trigger monitoring | Kill if thresholds breached |
Grounded Delivery: Phase Summary
| Phase | Objective | Primary Deliverable | Gate |
|---|---|---|---|
| Frame | Define what "good" looks like before building | Probabilistic success criteria; Evaluation dataset design; Integration map; Failure mode analysis | GO / NO-GO |
| Explore | Validate what is technically possible against real data | Evaluation dataset; Capability map; Legacy data access validation | GO / PIVOT / KILL |
| Shape | Design production architecture | Deterministic/non-deterministic separation; Fallback paths; Monitoring design; Coexistence pattern selection | GO / REDESIGN |
| Harden | Build and evaluate against quality thresholds | Production system; Monitoring infrastructure; Legacy integration validation; Red-team results | GO / CONTINUE / KILL |
| Operate | Monitor, maintain, and feed evidence back | Quality metrics; Drift detection; Value tracking; Trust graduation decisions | No exit — permanent |
Legacy Coexistence: Pattern Selection Guide
| Pattern | Latency Requirement | Legacy System Modification | Complexity | Best For |
|---|---|---|---|---|
| Data Exhaust | Batch (hours to days) | None | Low | Historical analysis; pattern recognition over accumulated data |
| Sidecar | Near-real-time | None | Medium | Augmenting legacy processes with AI insight without replacing them |
| Gateway | Real-time | None (gateway handles translation) | Medium-High | Agent or API access to legacy systems without modern interfaces |
| Shadow Pipeline | Batch (validation) | None | High | Gradual transition from legacy process to AI process with evidence |
| Legacy-Aware Agent | Mixed (agent handles asynchrony) | None | Highest | Autonomous AI operations spanning modern and legacy systems |
Core Principles Summary
| Principle | In One Line |
|---|---|
| Kill Early | The most expensive initiative is the one that should have been killed in month two and ran for two years. |
| Non-Deterministic by Default | Every process that touches an AI system must be designed for systems whose outputs cannot be predicted with certainty. |
| Legacy Is a Feature | Thirty years of transactional data is the highest-value input your AI can have — design to access it, not replace it. |
| Value Before Technology | Start with a business problem. The question is never "what can we do with AI?" |
| Operate Forever | Deploying without operating is abandonment on a delayed schedule. |
| Coexist Deliberately | Integration patterns must be selected before development begins, not figured out after. |
The LegacyForward.ai Framework is the foundation of the LegacyForward.ai Series. For more, visit legacyforward.ai.