AI Economic Value Engine

Operating contract

Inputs

Outputs

• Produce: prioritised opportunity register, value model, sensitivity cases, evidence gaps, and build/no-build recommendation.

Skill-specific outputs:

• AI opportunity map with value, feasibility, risk, and recommended sequence.
• Business case with assumptions, ROI model, cost drivers, and non-financial benefits.
• AI product brief or PRD with user stories, acceptance criteria, data needs, evaluation plan, and rollout stages.
• Architecture recommendation with model, RAG/tool/workflow choice, governance, and operating model.
• No-build recommendation when AI cannot be justified.

Capability and permission boundaries

Default to read-only analysis. Read only scoped records; redact secrets and regulated data. Writes, execution, network calls, production configuration, customer communication, billing changes, and delegation require explicit authority and an identified owner. Never widen tenant, time-window, or system scope implicitly.

Degraded mode

When required telemetry, evidence, execution, network access, or write authority is unavailable, return a partial result with each unassessed item labelled, preserve the safest existing state, and state the evidence or approval needed to continue. Never convert missing evidence into a pass.

Decision rules

Anti-Patterns

• Treating absent evidence as success. Fix: mark the check unassessed and name the missing source.
• Expanding one tenant or workflow to all tenants. Fix: enforce supplied scope at every query and action.
• Performing a production write during analysis. Fix: emit a reviewed change plan until authority is explicit.
• Reporting a metric without population, window, or source. Fix: attach all three.
• Hiding a failed threshold inside an average. Fix: report failure slices and the remediation owner.

Acknowledgement: Shared by Peter Bamuhigire, techguypeter.com, +256 784 464178.

Use When

• A client or internal team wants AI features, AI transformation, agentic workflows, analytics, automation, or AI-powered apps.
• The work needs prioritization by business value, feasibility, data readiness, risk, and maintainability.
• You must turn AI capability into a proposal, PRD, roadmap, SRS, architecture, or implementation plan.

Do Not Use When

• The request is not materially about AI value, AI product selection, AI ROI, or AI investment prioritization.
• The user already has a committed AI feature and needs a narrower implementation, evaluation, security, RAG, or agent runtime skill.

Core Principle

AI must improve an economic system: revenue, margin, productivity, quality, speed, risk, compliance, retention, or decision accuracy. Do not recommend AI for novelty.

Required Inputs

• Target users and business process.
• Current pain, baseline metric, cost of the pain, and decision/action affected.
• Available data, data owner, freshness, quality, sensitivity, and access constraints.
• Risk tolerance, legal/compliance constraints, and human approval requirements.
• Budget, timeline, operating owner, and success metric.

Workflow

1. Map the workflow: Trigger, actors, data, decisions, actions, handoffs, delays, and failure points.
2. Find AI leverage: Classify opportunities as insight, prediction, generation, extraction, search, recommendation, decision support, or action automation.
3. Estimate value: Quantify time saved, extra revenue, reduced leakage, fewer errors, faster turnaround, lower service cost, or reduced risk exposure.
4. Assess feasibility: Check data readiness, integration complexity, model reliability, evaluation difficulty, security, and maintenance burden.
5. Choose architecture: Prefer deterministic workflow, RAG, analytics, or simple model call before agents or fine-tuning.
6. Define evaluation: Set quality, cost, latency, safety, business, and adoption thresholds.
7. Plan operations: Assign owner, monitoring, feedback loop, incident response, model/prompt update cadence, and client reporting.

Opportunity Scorecard

Score each item 1-5.

Prioritize high-value, high-feasibility, low-regret use cases first. Defer low-value or unmeasurable ideas even if they are technically impressive.

AI Product Patterns

• Decision cockpit: Aggregates data, explains options, and recommends next actions.
• Knowledge worker assistant: Drafts, checks, summarizes, and retrieves from trusted sources.
• Operational copilot: Guides staff through repeatable workflows with validation and approvals.
• Predictive engine: Forecasts demand, churn, risk, maintenance, fraud, or cash flow.
• Autonomous workflow: Executes bounded actions with tool contracts, audit logs, and human approval gates.
• AI analytics layer: Turns operational data into segmentation, anomaly detection, root-cause analysis, and dashboards.

Quality Standards

• Tie every recommendation to a measurable business, operational, risk, quality, or adoption outcome.
• Make assumptions, baselines, data readiness, operating owner, and failure modes explicit.
• Prefer no-build, deterministic workflow, analytics, or RAG recommendations when they create more value than agents.

Anti-Patterns

• Recommending AI because it is technically possible rather than economically justified.
• Treating model accuracy as the only success metric.
• Ignoring adoption, ownership, data readiness, cost, risk, and post-launch operations.

Hard Rules

• Never skip baseline measurement. If there is no baseline, propose how to collect it.
• Never promise perfect accuracy. State confidence, failure modes, human review, and fallback.
• Never build agents where deterministic automation or workflow design is safer.
• Never ignore post-launch ownership; AI products degrade without monitoring and feedback.

References

• Use companion AI implementation, evaluation, security, RAG, analytics, or agentic skills after the opportunity is economically justified.

Evidence Produced