Independent proposal by Savy Employment Consultant LLC. Not an approved City and County of Denver program, procurement award, partnership, or government endorsement.
A bounded, human-in-the-loop proposal to recover staff capacity on document-heavy civic workflows — without replacing a single judgment call a human is trusted to make today.
Every ceiling above is a cap, not a budget commitment — the pilot is scoped to spend only what the 90-day evaluation proves is warranted, with no autonomous action ever taken on a city record without a named human's sign-off.
Denver's civic workforce loses measurable hours every week to repetitive, document-heavy tasks that sit upstream of any discretionary decision — the exact category of work where a restricted, auditable AI agent can recover capacity safely.
Assembling, formatting, and pre-filling recurring city forms and packets.
Locating the right approved document, policy, or precedent from internal knowledge bases.
Structuring and routing incoming requests before a human case worker touches them.
Producing first-pass replies and notices for human review, never auto-sent.
Keeping case files current — status updates, checklists, and administrative housekeeping.
Three bounded phases, each with a clear exit gate — the pilot never auto-extends or auto-expands scope without a fresh decision from the City.
Audit current workflow volume and time cost, select a narrow pilot scope, and establish the human approval chain before any agent writes a single draft.
Stand up the restricted-tool agent in draft-only mode. Every output is queued for human review — nothing reaches a resident or a record unreviewed.
Measure recovered hours, error rate, and staff/resident satisfaction against the baseline, then deliver a clear go / no-go recommendation to the City.
The pilot is deliberately designed as one node in a larger civic capacity-building model — not a standalone tool deployment.
The restricted agentic pipeline itself — recovering staff hours on document prep, retrieval, intake, and case administration.
Upskilling city staff across every level to use, supervise, and govern AI tools responsibly — not just the pilot team.
Aligning the pilot’s workforce-development data with CivicFlow-style case management and WIOA-funded training pathways already in use by city workforce boards.
Least-privilege by design. No new data center, no new standing infrastructure to maintain — five controls, applied in sequence, keep a human accountable for every outcome.
The agent is grounded exclusively in a vetted, City-approved knowledge base — internal policy documents, forms, and precedent the City explicitly uploads. No open internet scraping, no unvetted third-party data sources feed the system at any point.
Illustrative model—not a forecast or guaranteed cash saving. Figures represent potential staff-hour capacity recovered under modeled assumptions, not realized budgetary savings, headcount reductions, or guaranteed outcomes.
A hard 3-year ceiling, broken into four accountable phases.
Six dedicated tracks, because every level of city government touches AI governance differently.
Strategic oversight, risk framing, and go/no-go decision literacy.
Day-to-day use of the approved tools within their restricted scope.
Reviewing and approving agent outputs before anything is finalized.
Access control, audit-log maintenance, and rollback procedures.
Tool-whitelisting, integration maintenance, and technical escalation.
Internal train-the-trainer capacity so the City scales without ongoing dependency on Savy Agent.
All six tracks are structured to align with U.S. Department of Labor (DOL) digital-literacy and AI-upskilling frameworks under WIOA, so training hours can plug directly into existing city workforce-development reporting rather than creating a parallel system.
Denver is the proving ground — the service model is deliberately built to be repeated, not rebuilt, for the next city that wants it.
For individuals and organizations who want to help fund the pilot's research and development ahead of any formal city procurement decision.
Named acknowledgment in the pilot’s public evaluation report.
Supporter benefits, plus a briefing on pilot findings ahead of public release.
Partner benefits, plus recognition as a founding R&D supporter of the Denver AI Beacon pilot.
Founding support funds independent research and pilot development costs only. It does not purchase, influence, expedite, or guarantee any City and County of Denver procurement decision, contract award, partnership, or endorsement. No tier confers access to city officials, preferential treatment, or lobbying services of any kind.
This proposal is a starting point for a conversation with the City and County of Denver — not a finished deal. Three bounded ways to engage:
All inquiries route directly to tec@savyagent.co.