← All case studies

Nighthawk Platform — 5-Year Roadmap

The strategy that takes fragmented operational tooling to a unified AI-native platform.

Role Platform author and lead engineer
Period 2025 – 2030 (Phase 1 complete; Phase 2 in flight)
Headline 4-phase roadmap · 8+ legacy tools consolidated
Stack MySQL 8.x · containerized web backend · Oracle · SQL Server · SAP HANA → Snowflake · ClickHouse · Docker · Kubernetes (Tilt) · Azure AD OAuth2 · Microsoft Graph · Three-tier RBAC

Problem

Most operations-heavy organizations — utilities, industrial manufacturers, regulated infrastructure — run their day-to-day work on a portfolio of disconnected point-tools. Each tool has its own login, its own database, its own update cycle, its own data model. The cost isn’t just inefficiency; it’s that no one feature can compound on another. Adding AI to a stack of disconnected tools means integrating AI ten times, once per tool. The platform problem has to be solved before the intelligence problem can be.

OG&E Grid Operations was in this state at the end of 2025: legacy SCADA viewers, ad-hoc dashboards, spreadsheet KPIs, separate on-call systems, ETL scripts no single person owned. The opportunity was bigger than any single tool replacement — it was a platform decision.

Approach

Design and author a multi-year roadmap that delivers in phases:

  1. Foundation — consolidate the tools first, under one auth model and one shared data layer, before adding any new capability on top.
  2. Intelligence — once the foundation is in place, layer AI as a single integration that every downstream feature reuses.
  3. Agentic — once AI is integrated, evolve from answering questions to taking action, under a human-in-the-loop governance framework.
  4. Enterprise — once agentic patterns are proven, extend the platform across the organization, not just one department.

The phasing order is non-negotiable. Adding AI to a fragmented platform produces fragmented AI. The platform comes first.

What Phase 1 Delivered (Complete)

  • Unified application shell — 8+ legacy tools consolidated into one platform with shared header, navigation, search, and authentication.
  • Multi-database architecture — MySQL, Oracle, SQL Server, SAP HANA (migrating to Snowflake), and ClickHouse accessible through a shared data layer with Azure AD SSO across all of them.
  • Three-tier RBAC — department · user · superadmin, with audit logging on every authorization decision.
  • Container-orchestrated deployment — Docker images orchestrated with Tilt onto Kubernetes.
  • Change Control Process v4 — formal governance for production changes, RACI matrix, capital-classification process.
  • 15+ applications now running on the platform — Resource Availability Suite, Strategic Undergrounding Analytics, Grid Operator Dashboard, Circuit Operations App, ADMS QA, Wall Map, Reliability hub, and others.

What Phase 2 Adds (In Flight, 2026–2027)

  • AI Model Integration on AWS Bedrock with Anthropic Claude models — the platform layer for four production-bound LLM capabilities (covered in detail on the AI Integration case study).
  • Predictive analytics — Accurate ETR Status, ADMS model improvements.
  • Regulatory compliance reporting — AR-PSC automated submissions.
  • Itron AMI data connection — bringing meter-level data into the platform.
  • Legacy migration — ~700 iDashboards users migrated onto the Nighthawk BI layer.

What Phase 3 Plans (2028–2030)

  • Agentic AI under a four-tier human-in-the-loop governance framework I authored:
    • Tier 1: read-only operations (no approval needed)
    • Tier 2: low-risk actions (auto-approved with log entry)
    • Tier 3: operational actions (supervisor approval)
    • Tier 4: critical actions (manager approval + audit review)
  • Five named agents: Schedule, Storm Preparation, Compliance, Dispatch Recommendation, Auto-Reporting.
  • Each agent operates inside its tier with explicit guardrails and full per-action audit trails.

Why the Pattern Transfers

This roadmap structure isn’t OG&E-specific. The Foundation → Intelligence → Agentic → Enterprise progression is the right phasing for any operations-heavy organization running on a stack of legacy systems. Utilities, healthcare, defense, finance, manufacturing — the constraint is the same: you can’t compound features on top of a fragmented platform, so the platform consolidation has to come first.

The differentiator is owning the strategy, not just the implementation. A platform engineer with a five-year roadmap negotiates very different conversations with leadership than one with a backlog of tickets.

Results to date

  • Phase 1 marked complete on schedule.
  • Phase 2 deliverables on track for 2026–2027 horizons.
  • Phase 3 governance framework authored and reviewed.
  • Stakeholder alignment from Director of AI through Grid Operations leadership.

Impact thesis — four pillars: reliability (the SAIDI lever inside the utility), cost (O&M reduction), safety, and regulatory compliance — the same scorecard the AI integration optimizes for, with platform consolidation as the precondition that makes those gains possible.

Some specifics are abstracted for confidentiality. Happy to go deeper on the technical approach, the failure modes we worked through, or the operational outcomes — jose@macias-tech.com.