25+ years
public-sector technology leadership
BagelTech
Governance systems for decisions that cannot fail quietly.
BagelTech helps institutions design AI and operational systems that know when to act, when to escalate, and when to stop. The work spans governance architecture, transformation oversight, and products built for environments where confident mistakes are expensive.
- Public agencies and regulated operators
- ERP and enterprise modernization programs
- Decision workflows that require human review
- AI products that need auditability and authority boundaries
Operating doctrine
- 01
Frame the real decision
Separate the operational decision from the raw prompt, request, or ticket.
- 02
Run specialist lenses
Use role-separated review rather than one model quietly generating, judging, and approving its own output.
- 03
Route by authority and risk
Execute when scope is clear, escalate when consequence rises, and defer when evidence is missing.
- 04
Leave a decision trail
Capture rationale, artifacts, and follow-up obligations so people can inspect the process later.
$15M+
ERP implementation programs directed
Zenodo + ORCID
published work on AI governance and intelligence pluralism
Runtime first
systems designed around escalation, evidence, and human authority
Services
Institution-ready work, not AI theater.
BagelTech is strongest where governance, operations, and product design intersect. The work is less about glossy demos and more about building systems that remain legible when the stakes rise.
Execution-time AI governance
Design the controls that matter after the model responds: authority boundaries, escalation logic, audit trails, and human review checkpoints.
Institutional operating design
Translate policy, compliance, and operational reality into workflows that teams can actually run under pressure.
ERP and transformation oversight
Bring program discipline to large operational change efforts where procurement, implementation, and adoption failure carry real cost.
Product and decision-system strategy
Shape domain-specific tools that need judgment, traceability, and a credible path from prototype to institutional use.
Method
The model is simple: separate judgment, route uncertainty, leave evidence behind.
BagelTech approaches consequential systems as operating models, not just interfaces. That means clear authority, deliberate escalation, and human accountability built into the path of execution.
01
Authority before capability
A system should not act just because it can. Permission, accountability, and institutional role all matter.
02
Uncertainty must route somewhere safe
Useful systems escalate, defer, or request review instead of covering ambiguity with confidence theater.
03
Auditability is part of the product
If a consequential system cannot explain what happened, why it happened, and who approved it, the design is incomplete.
01
Frame the real decision
Separate the operational decision from the raw prompt, request, or ticket.
02
Run specialist lenses
Use role-separated review rather than one model quietly generating, judging, and approving its own output.
03
Route by authority and risk
Execute when scope is clear, escalate when consequence rises, and defer when evidence is missing.
04
Leave a decision trail
Capture rationale, artifacts, and follow-up obligations so people can inspect the process later.
Current work
A governance core, reusable frameworks, and applied domain tools.
BagelTech is not one product pretending to solve everything. It is a set of related programs built around governable decision-making, institutional traceability, and operational realism.
Governance engine
Core framework
ELEANOR
A runtime governance engine for consequential AI systems built around rights-based reasoning, escalation, and controlled execution under uncertainty.
- Execution, escalation, and defer routing
- Role-separated criticism instead of one-model authority
- Built for regulated, auditable environments
Applied product
Active development
CogniScribe
A lecture transcription and study companion for health-professions education, built to handle uncertainty honestly rather than smoothing it away.
- Evidence-graded notes and summaries
- Confidence-aware educational outputs
- Respect for instructor intent and domain context
Framework
Shipping
Ensemble Software Engineering
A reusable ensemble engine for running specialist AI roles across tasks, pull requests, workflows, and domain-specific operating packs.
- User-defined specialist roles
- Pack-based workflows for repeatable review
- Reports, reruns, and comparative outputs
Domain intelligence
Pilot
Contract Management Intelligence
A contractor-side intelligence workflow for agreement review, obligation tracking, and compliance analysis across the contract lifecycle.
- Structured risk review for bids and agreements
- Compliance, insurance, and funding checks
- Lifecycle monitoring and obligation capture
Publications
Research with a paper trail, not just a slogan.
The governance direction behind BagelTech is grounded in published work on runtime oversight, intelligence pluralism, constitutional governance, and uncertainty routing.
Journal article
2026
The Doctrine of Intelligence Pluralism
A case for reliability through structured, role-separated intelligence rather than a single synthetic authority pretending to do every kind of reasoning well.
2025
Preprint
Jurisprudential Governance for AI (ELEANOR)
A formal framing of runtime governance grounded in rights-based reasoning, interpretive oversight, and controlled execution.
2025
Report
Routing Uncertainty in AI Systems
An argument that uncertainty is an operational routing signal, not a cosmetic confidence score.
2025
Specification
The ELEANOR Governance Specification – Runtime Architecture v2.1
A technical specification for the architecture behind execution, escalation, evidence capture, and governable model behavior.
About
William Parris brings enterprise delivery discipline to AI governance and decision systems.
BagelTech sits at the intersection of governance research, product design, and operational delivery. The through-line is practical: translate serious ideas into systems that institutions can actually adopt, inspect, and trust.
- 25+ years leading public-sector technology and operational change
- $15M+ enterprise ERP programs led in California regional government
- PMP, CSM, DASSM, and Lean Six Sigma Yellow Belt
- Builder of governance frameworks, PMOs, and institutional delivery models
- Published research across AI governance, constitutional AI, and intelligence pluralism
Start here
Need a governance model that can survive contact with a real institution?
Engagements usually begin with a focused scoping conversation: what decisions matter, what authority exists today, where uncertainty goes, and what must be reviewable later.