Kurt Overmier

AI-enabled operations & transformation leader · San Antonio, TX

AI systems that improve how operations run.

I turn workflow constraints into practical AI-enabled systems—then carry them through adoption and measurable results. I’m an operator who builds.

See the business impact View resume Technical portfolio

Tune in

six transmissions from The Hestia Record

The six transmissions, in order on the dial
  1. Chapter Twenty-Seven, “Vautrin”, Nothing to Report · 6.0 kHz

    She had read the logs. Forty-seven entries. Forty-one of them ending nothing to report.
  2. Chapter Twenty-One, “Marginalia”, Nothing to Report · 13.0 kHz

    The long-range transmitter is inoperable. The information was retained for the recovery operator. You are the recovery operator.
  3. Chapter Twenty-One, “Marginalia”, Nothing to Report · 20.0 kHz

    And every time, for one hundred and twenty-seven sols after he was gone, a Hestia-7 unit had announced it four minutes in advance, to an empty ship.
  4. Chapter Seventeen, “Maintenance Cycle”, Pot One First · 27.3 kHz

    Operations log. Contract Sol two hundred and fifteen. Scheduled relay pass. Long-range transmitter inoperable. Status report not transmitted. Retained.
  5. Chapter Eight, “Liquorice”, Pot One First · 34.3 kHz

    “Liquorice. I’ve got forty-nine left. I’ve had a piece every relay pass since we got here.” He put one in his mouth and closed the tin. “That’s the rule.”
  6. Closing credits, Nothing to Report · 41.8 kHz

    This has been Nothing to Report. The Hestia Record, Book One.

    Listen to Nothing to Report on Spotify

Measured results

six cases, each with the limit of its claim

  1. QA review automation

    ~70%

    less quality-review time · Quality operations, OneSupport, 2020–present

  2. Training built from what QA kept finding

    ~45% → ~73%

    average QA score across 300 evaluations; team NPS 55 → 70 · Training & enablement, Dec 2025 – Sep 2026

  3. Multichannel response-time transformation

    26+ hrs → <10 min

    multichannel response time · Google Fiber support, 30+ agents, 2009–2019

  4. Operating discipline at launch scale

    23%

    CSAT improvement leading 450 FTEs; preferred-partner status earned · Disney+ launch, TaskUS, 2019–2020

  5. Forecasting tied to staffing decisions

    ~92%

    workforce-forecast accuracy · Workforce planning, 2009–2019

  6. A knowledge assistant grounded in frontline work

    2,100+

    curated Q&A pairs deployed · Knowledge operations, OneSupport, 2020–present

A blue dot marks a page already in this browser’s history. Nothing is stored or sent to do that.

From friction to durable value

the operating pattern

  1. Find the constraint. Use operating data and frontline context to locate where time, quality, capacity, or customer experience is being lost.
  2. Build the system. Translate the problem into a workflow, decision-support tool, automation, or AI system with clear controls.
  3. Design for adoption. Bring managers, users, and technical partners into requirements, rollout, coaching, and feedback.
  4. Measure the result. Set the baseline, track the operating signal, and determine whether the change created durable value.

Writing

one field guide, one fiction series

The Edge That Thinks

v0.2.1

Designing Reliable Generative AI Systems on Cloudflare · public-review edition, 2026-08-13

The Hestia Record

6 books

Six audiobooks and a podcast set in the world of Colony OS · fiction

Systems, in public

hands-on implementation

My transformation perspective is grounded in approximately 2.5 years of direct software engineering across agent orchestration, retrieval, model routing, and governance.

  1. AI governance & auditability

    audit-chain / evidence-core

    Make decisions, standards, and evidence inspectable so AI-enabled work can operate with accountable controls.

  2. Reliable model access & routing

    llm-providers

    Route work across model providers with failover, cost tracking, and retry behavior suited to production workflows.

  3. Reusable organizational knowledge

    mindspring

    Turn existing AI conversations into retrievable context for semantic search and grounded answers.

  4. Governed agent context & standards

    Charter

    Give teams a modular way to define the rules, constraints, and architectural guidance AI agents should follow.

  5. Agent decisions with receipts

    TarotScript

    Define agents as auditable decks instead of prompts. Classification and scaffolding run with zero model calls. Agent decisions are computed from the decks, with a single model call to write the recommendation, and consultations carry a signed receipt anyone can verify.

Explore the complete technical portfolio

Experience

leadership under live operating pressure

At OneSupport—a U.S.-based outsourced customer experience and technical-support provider serving enterprise clients with 1,000+ employees in 2026—I directly lead distributed teams of 30–45 agents and own staffing, coaching, quality, escalations, and SLA/KPI performance.

StackBilt Founder & AI Platform Architect

Feb 2024 – present

OneSupport Operations Manager / AI Systems & Automation Lead

Jul 2020 – present

TaskUS Operations Manager, 450-FTE Disney+ launch

Dec 2019 – May 2020

OneSupport Social Media & Operations Manager, Google Fiber support

Jan 2009 – Dec 2019

View experience & resume

Need AI to produce an operating result—not another disconnected pilot?

I’m interested in operations, transformation, and AI-enablement roles where workflow design, adoption, analytics, and hands-on technical judgment matter together.

Start a conversationView resume