the situation
Between 2023 and 2025, the organization handled ~400 conceptual projects per year, while critical review capacity remained concentrated in a small number of specialists. That created bottlenecks and increased the risk of inconsistent early-stage submissions. These early checks are decisive for improving maturity ahead of the next stage-gate, reducing rework and avoiding preventable cost and schedule impacts.
the IDEA
Convert a dense internal standard (and its supporting guidance) into a repeatable, auditable review flow. The goal was not to replace technical judgment, but to shift left quality by standardizing the “did we include what we need?” check, so specialists could spend time where expertise is truly required.
the solution
A Copilot Studio AI agent that reviews a conceptual project submission against the organization’s minimum information requirements for the relevant project category, and returns a table-based adherence check for human review.
High-level flow:
- User uploads the conceptual project file
- Agent runs a guided, topic-based check against minimum requirements (by project type)
- Agent outputs results in a structured table for review and action
Boundary (by design): the agent does not do a deep technical evaluation. It checks whether the submission contains the required information and whether each section meets the intent (fully/partially), with clear escalation to specialists when needed.
what i did
Led my team in:
- Codifying an internal engineering standard into a reusable, topic-based review model (consistent, complete, auditable)
- Defining operating guardrails (minimum-requirements verification vs. technical review) to protect quality, credibility, and trust
- Building a Copilot Studio agent, including logic by project category and a structured output aligned to existing stage-gate routines
- Aligning cross-functional stakeholders on scope, “good enough” thresholds, and escalation points to specialists (human-in-the-loop)
- Establishing KM governance (ownership, review cadence, version control) so the asset stays accurate as standards evolve
why it mattered
This solution turned expert-only review knowledge into a scalable decision-support capability, expanding the potential for consistent coverage across the full volume of conceptual projects.
- Efficiency signal: Average savings of ~12–12.5 minutes per conceptual project, freeing specialist time for higher-value technical evaluation.
- Quality/compliance signal: Increased adherence earlier in the authoring process by giving project owners a consistent, guided checklist before formal review.
- Scalability signal: A standardized, replicable approach that can be extended with new checks and automations across other early-stage artifacts.
- Value signal (modeled): Potential soft savings of up to ~US$1.2K per project tied to improved early-phase maturity and reduced rework (converted from a local estimate; internal performance metrics beyond public reporting were not accessible due to platform changes).
Cinthia Blanchard 
Leave a comment