Eric Driscoll Consulting

Approach

From evidence to durable operating value.

Start with the decision. Add only the analysis, tooling, and governance the value and risk can justify.

I call this decision engineering: design the evidence, economics, interface, decision rights, and feedback around a specific choice, then build only what the decision justifies.

North star

Evidence reveals the opportunity. Economics tests the value. Action captures it. Systems sustain it.

Formal method

Identify. Validate. Operationalize. Strengthen.

The depth of work scales with materiality, uncertainty, risk, and the operating environment.

  1. 01Evidence

    Identify the opportunity

    Define the business question, source, baseline, limits, and confidence needed to support a decision.

  2. 02Economics

    Validate the economics

    Test lifecycle cost, benefit, risk, effort, opportunity cost, uncertainty, and materiality.

  3. 03Action

    Operationalize the decision

    Name the owner, authority, timing, output, acceptance condition, and next decision.

  4. 04Systems

    Strengthen the system

    Build the learning into ownership, process, feedback, proportionate controls, review, and an explicit exit.

See the method applied

Operating principles

Six conditions for useful AI.

A useful AI workflow should create lasting value while preserving control, capability, and purpose. I assess the whole workflow against these six conditions.

ConditionWhat it means in practice
Durable net valueBenefits remain after technology costs, human effort, checking, rework, and recovery are counted.
Meaningful human controlQualified people have the information, time, and authority to understand, challenge, and intervene.
Accountability and recoveryDecisions have clear owners; failures have usable escalation, fallback, and recovery paths.
Deepening learningPeople continue developing domain expertise, AI literacy, and independent judgment.
Career pathwaysWork preserves opportunities to gain experience and take on greater responsibility.
Job purposePeople retain agency and useful responsibilities, with manageable demands.

Control and recovery need to work before consequential use. Value and human outcomes need evidence over time.

Adapted by Eric Driscoll Consulting from CIVIC-AI Collaboration (Wu, Ziems et al.), When Does AI Augment Work? A Workflow-Level Framework for Human-Agent Collaboration (2026), arXiv:2609.12482v1, licensed under CC BY 4.0. Practical wording adapted by EDC. No author or institutional endorsement is implied.