Skip to content
Hazen + AI | Troy Redding

Proposed approach

Make the rules clear before scale

AI policy should make good work easier and risky work harder. The rules need to be clear enough that an engineer knows what is allowed without reading a long policy every time.

Approved

Examples

  • Public source research
  • Low consequence drafting
  • Approved internal knowledge search
  • Brainstorming with non sensitive information

Requirements

  • Approved tool
  • Normal company security controls
  • Employee reviews the result

Controlled

Examples

  • Client information
  • Proposal content
  • Project financial data
  • Technical analysis support
  • Design review

Requirements

  • Approved enterprise environment
  • Access controls
  • Defined human review
  • Named business owner
  • Evaluation before broad rollout
  • Logging where appropriate

Restricted

Not allowed

  • Confidential client information in unauthorized public AI tools
  • AI approving engineering work
  • Unverified AI output used as final regulatory authority
  • Automated decisions that bypass required professional review

From request to production

  1. 1Request
  2. 2Value and risk review
  3. 3Build, buy or partner decision
  4. 4Controlled pilot
  5. 5Evaluation
  6. 6Production approval
  7. 7Monitoring

Who should own what

Proposed responsibility split based on public role descriptions. This is not Hazen's actual reporting structure.

Director of AI Strategy & Adoption
Direction, portfolio choices, investment recommendations, governance direction, executive reporting, value measurement and team leadership.
Data and Technology / DataTech
Data and technical expertise, engineering solutions, client facing innovation and solution development.
IT and AI Adoption & Governance
Enterprise platforms, access, licensing, support, training, usage, policy execution and the tool approval process.
Legal, Risk and Cybersecurity
Contracts, confidentiality, liability, vendor review and security review.
Practice leaders
Identify useful workflows, sponsor pilots and validate business value.
AI champions
Peer support, feedback and practical examples.
Licensed engineers
Final professional judgment and accountability for engineering work.

How I would build the core AI team

This is a proposed cross-functional delivery group, not an asserted reporting structure. The AI Adoption & Governance Manager is an IT partner; the public listing says that role reports to the Director of Information Technology. Each role below is a capability to assess against existing staff before considering new hiring.

Proposed capabilities and delivery group

  1. Director of AI Strategy & Adoption

    Sets direction, chooses priorities, manages the portfolio and coordinates the proposed cross-functional delivery group.

  2. AI Adoption & Governance Manager

    Partners from IT on platform adoption, training, support, tool review and policy execution. The public listing places this role under the Director of Information Technology, not the Director of AI Strategy & Adoption.

  3. AI and machine learning engineering

    Builds and evaluates technical solutions where Hazen needs custom capability.

  4. Data engineering

    Connects approved data sources, permissions and pipelines needed for reliable AI systems.

  5. Program and product delivery

    Turns priorities into scoped work, coordinates pilots, measures results and helps move useful tools into supported production.

Practice champions and subject matter experts should remain embedded in the practices. The central team supports them rather than replacing local expertise.

Adoption should follow usefulness

This loop expresses confidence in a practical approach, not assumptions about Hazen's current organization.

  1. 01

    Define real tasks

    A respected practice sponsor and qualified reviewers choose representative work and name the support owner.

  2. 02

    Record the baseline

    Measure completion time, quality, review and rework before the pilot begins.

  3. 03

    Run a bounded pilot

    Test representative tasks and gather reviewer findings, corrections and overrides.

  4. 04

    Adjust the workflow

    Improve the tool, sources, workflow or training. Use champion feedback and office hours to resolve practical issues.

  5. 05

    Expand with evidence

    Expand only after reviewers confirm usefulness and controls, with clear ongoing support ownership.

At Metron, I led adoption across engineering, operations, sales and customer excellence, evaluated and supported Copilot, Claude and ChatGPT, and provided practical training in responsible use. I would combine that support with reviewer evidence, champion feedback, office hours and a named support owner. Repeated use matters only when the work and controls hold up.