AI Leadership · A personal perspective
Move fast without flying blind.
America has powerful AI tools and capable institutions. The hard part is turning them into better public service inside real missions. I would start with a few bounded workflows, measure time, cost, quality, and safety against a baseline, and expand only when the evidence shows public value.
I serve as an elected commissioner of the South Walton Mosquito Control District in Florida. Public service has taught me to begin with a defined mission, a responsible decision-maker, and an outcome residents can examine. Separately, I founded the Good Samaritan Institute and Good Combinator, where I work on applied AI, workforce development, and practical tools for the public good.
In 2025, T-Mobile gave its Tipping Point Designationto the South Walton Mosquito Control District for its drone-powered mosquito control program. The district earned that recognition. My lesson from the work is practical innovation in a public service mission: solve a specific operational problem, measure what changes, and keep the people responsible for the mission in charge.
These records support different claims. Their limits matter as much as the links.
South Walton Mosquito Control District; 2025 T-Mobile Tipping Point Designation
The recognition belongs to the district. T-Mobile reports a 75% reduction in treatment time for its drone program; that figure is T-Mobile’s account, not an independently audited result here.
T-Mobile announcementDelivered a 2026 Florida Association of Special Districts session on AI
A documented speaking engagement; it does not establish federal experience or deployment outcomes.
Session evidenceGood Combinator’s published Forward Deployed Engineer curriculum
A published training pathway, not proof of a federally tested workforce model.
Examine the curriculumCentralize capabilities, not missions.
Shared services, accountable agents, and qualified teams can support distinct agency missions, with each mission owner setting the objective and measuring the result.
NIST’s CAISIalready works on AI evaluation and security measurement. Its work with GSA and USAiaddresses evaluation in federal procurement and real workflows. Any broader implementation effort should learn from those institutions and identify a useful operating gap before proposing another layer. I have no affiliation with CAISI, GSA, or USAi.
Good Combinatorconnects learning, deployment readiness, and applied projects. Itspublished Forward Deployed Engineer curriculumis a concrete example of training people to translate a mission problem into a reviewable implementation. The methods and materials could be examined and adapted; the organization is not a ready-made federal system.
We have defined an internal pilot gate: store traces outside an agent’s writable environment; use stable keys for consequential writes; test retries, delayed commits, and duplicate delivery; report safety and useful task completion together; and bind a single-use deployment approval to the exact evaluated artifact. These are proposed controls for our pilots, not an externally validated federal standard or evidence that every control has been proven in deployment.
A proposed first 90 days, subject to the authority, partners, and resources of any actual role.
Days 1–15
Meet mission owners and existing federal AI teams. Inventory shared capabilities, procurement paths, evaluation methods, and a short list of recurring service problems. Produce a map of owners and gaps.
Days 16–30
Select two or three workflows with willing mission owners. Record the baseline, data authority, human review points, success measures, and stop conditions before choosing tools.
Days 31–60
Pair qualified implementers with agency staff. Keep agent authority narrow, preserve evidence outside the agent’s control, and test errors, retries, and recovery alongside useful work.
Days 61–90
Compare measured service value and risk with the baseline. Publish a decision record for each pilot: expand, revise, or stop, including what can be shared and what remains mission-specific.
Review the public materials, then request a 30-minute working session. I would use the time to examine one mission problem, the baseline we would need, and a small test that could produce a decision.
Request a 30-minute working sessionThis is a personal perspective. No endorsement by the South Walton Mosquito Control District or any federal entity is implied. Any public service role would require appropriate disclosures and handling of potential conflicts, including an independent review of interests related to Good Combinator and its materials.
Doug Liles · Updated September 27, 2026, 10:09 AM CDT · Public claims link to their source; the organizing proposal and first 90 days are my perspective.