# Deep Economic Assessment Rubric

Applied ONLY to high-priority opportunities (the ones that survived the four-metric cut). Thirteen criteria across four categories, each scored 1, 3, or 5. Interpolating (2, 4) is allowed when the executive argues for it.

Explain each criterion in one plain sentence before scoring it the first time. The executive should leave able to run this rubric at their company without you.

## Category 1: Deep Functional Fit

The same four metrics from the funnel, now examined deeply per opportunity rather than comparatively across the list.

| Criterion | Score 1 (Unfavorable) | Score 3 (Neutral/Moderate) | Score 5 (Highly Favorable) |
| --- | --- | --- | --- |
| **Risk (Data, Security, Regulatory)** | High compliance risk, severe data privacy concerns, or heavy reputational risk. | Manageable risks with standard mitigation; standard security protocols apply. | Zero PII involved; negligible regulatory friction; highly secure. |
| **Business Impact (Top/Bottom Line)** | Marginal or highly speculative impact on primary business goals. | Moderate, steady impact on revenue generation or cost savings. | Transformational impact; creates a massive competitive moat or revenue lift. |
| **Feasibility (Technical & Data)** | Unproven technology; requires a massive, complex data engineering overhaul. | Mature tech, but requires some data cleaning and API integration work. | Turnkey or near-turnkey; data is already clean, structured, and highly accessible. |
| **Human vs. AI (Mission & Perception)** | Violates core values; customers would perceive it as cheap, impersonal, or trust-breaking. | Neutral alignment; customers accept it as a standard, invisible backend efficiency. | Deeply aligns with the mission; customers actively praise this and see value-add. |

## Category 2: Ethics

| Criterion | Score 1 | Score 3 | Score 5 |
| --- | --- | --- | --- |
| **Ethics & Responsible AI** | Black-box decision-making; high potential for systemic bias, unfairness, or societal harm. | Standard guardrails in place; moderate transparency with human-in-the-loop oversight. | Fully explainable, transparent, and proactively mitigates bias; promotes equity. |

## Category 3: Economics

| Criterion | Score 1 | Score 3 | Score 5 |
| --- | --- | --- | --- |
| **Capital Expenditures (CAPEX)** | Prohibitive upfront investment required (custom infrastructure, high initial licensing). | Moderate upfront costs; standard enterprise software/hardware investments. | Minimal upfront capital needed (e.g., leveraging existing SaaS/API infrastructure). |
| **Operating Expenses (OPEX)** | Unpredictable, runaway recurring costs (e.g., massive compute/inference fees). | Predictable recurring costs that scale linearly with standard business growth. | Highly favorable unit economics; OPEX shrinks as volume scales. |
| **Cash Flow Analysis / ROI** | Heavy initial cash drain; payback period exceeds 24-36+ months. | Near-term cash neutral; steady payback within 12-18 months. | Immediate ROI; rapid positive cash flow generation within 0-6 months. |
| **Financial Returns (NPV & IRR)** | Negative or marginal Net Present Value; IRR below the company hurdle rate. | Positive NPV; IRR comfortably meets the company's baseline capital requirements. | Exceptionally high positive NPV; IRR vastly outperforms standard investments. |
| **Levelized Cost** | AI cost per unit/transaction is higher than or equal to the current legacy/human baseline. | Cost per unit is slightly lower, justifying the transition over time. | Drastically lowers the cost per unit/transaction to near zero at scale. |

Plain-language explainers to use when teaching the economics rows:
- CAPEX: what you pay before the first unit of value ships.
- OPEX: what you pay every month it runs; the danger is inference costs that scale faster than value.
- Cash flow / ROI: how long until the money comes back.
- NPV and IRR: whether this beats the other things the company could do with the same capital. If the executive doesn't have a hurdle rate handy, 10 to 15% is a workable default for the exercise.
- Levelized cost: the total lifetime cost per task divided by tasks performed, compared head-to-head with the human baseline. This is the number that reveals which opportunities are newly viable.

## Category 4: Organizational Impact

| Criterion | Score 1 | Score 3 | Score 5 |
| --- | --- | --- | --- |
| **Process Design & Modeling** | Requires tearing down and rebuilding core operational processes from scratch. | Integrates into existing workflows with moderate process re-mapping. | Seamlessly slots into or instantly simplifies existing, complex workflows. |
| **Clock Speed & Velocity** | Adds layers of review; creates bottlenecks due to human-in-the-loop latency. | Maintains current operational speeds but increases accuracy or volume. | Exponentially accelerates decision-making, time-to-market, and workflow execution. |
| **Revenue per Employee** | Negligible impact on individual output; requires new dedicated AI headcount. | Noticeable boost to individual productivity; stabilizes headcount needs. | Acts as a massive multiplier; allows the business to scale 10x without adding headcount. |

## Scoring and verdict

- Total = sum of all 13 criteria. Range 13 to 65.
- **Passes deep assessment: total >= 39 (an average of 3) with no category averaging below 2.** A single 1 on Risk or Ethics is a flag to resolve, not an automatic kill; name the mitigation that would raise it.
- Rank passing opportunities by total score. That ranked list IS the AI implementation roadmap.
- The rule from the process: implement the highest-priority opportunities that pass deep economic assessment, and use those as the roadmap moving forward. Every high-priority opportunity must make it beyond the sandbox.