From the OpenAI edition of November 2, 2026
OpenAI introduces LifeVal, a vendor evaluation of agentic everyday tasks including accessing government services
OpenAI · State and local government services; K-12 enrollment (citizen-facing) · United States and international
- Publisher
- OpenAI
- Original publication
- November 2, 2026
- Source retrieved
- 2026-11-03
- Event date
- 2026-11-02
What happened
OpenAI describes LifeVal, an internal evaluation of everyday tasks built from the American Time Use Survey, the Multinational Time Use Study and a 3,995-respondent survey, spanning areas including accessing government services and education. Example task scores include 93.1% (kindergarten enrollment), 98.0% (flood volunteer coordination) and 100% (finding a home and adding a county deed to a records cart).
Why it matters
Indicates OpenAI is optimizing ChatGPT agents to act on resident-facing government and school workflows (enrollment, county records, emergency prep), which could change how residents interact with public portals.
Evidence and measured results
Vendor-built and vendor-scored benchmark; the page shows selected examples, not full methodology, grading rubric, task counts, environments, or independent replication.
Limitations and uncertainty
Vendor claim; scores are on curated tasks with unknown grading; real government portals, accessibility, and failure modes were not reported; no SLED agency involvement.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-11-03; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Residents increasingly may use AI agents to interact with government and school portals not designed for them.
- Stakeholders
- Digital services directors
- County records offices
- School district enrollment offices
- Discovery questions
- How do portals detect and handle AI agent sessions?
- What transactions require human verification?
- Value hypothesis
- Preparing portals for agent traffic may reduce errors and support residents.
- Potential engagement
- A bounded test of two resident journeys using AI agents in a sandbox.
- Unsupported claims
- Benchmark scores reflect real portal performance
- Agents are safe for unsupervised transactions
Pre-sales engineering
Role takeaway
- Fit
- Relevant to citizen-facing services as an emerging pattern; vendor eval only.
- Architecture and integration
- Agents use browser use, computer use and connectors per vendor.
- Prerequisites
- Sandbox portal environments
- Agent usage policy
- Constraints
- No methodology published
- Varied portal authentication
- Security
- Identity verification and authorization for agent-submitted transactions must be assessed.
- Proof of value
- Run agent on representative portal tasks and measure completion and error rates
Delivery
Role takeaway
- Work
- Agent-readiness testing and policy for public portals.
- Dependencies
- Portal vendors
- Legal guidance on delegated access
- Ownership
- Agency digital services team.
- Skills and adoption
- Staff testing of agentic tools.
- Governance checkpoints
- Policy on delegated agent access
- Privacy review
- Acceptance criteria
- Documented completion and error rates for tested journeys
- Risks
- Erroneous transactions
- Unclear liability
- Accessibility gaps
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Agentic browser/computer-use interactions with public portals imply bot traffic and automated form submissions on government sites.
Governance
Who approves, reviews and stays accountable for outcomes?
Agencies may need policies on AI agents acting for residents (authorization, identity, liability).
Security and privacy
What data, permissions and controls need testing?
Agents transacting on county records or enrollment systems raise identity-verification and data-handling questions.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Could aid residents navigating complex services, but accessibility outcomes were not measured.
Procurement
What should contracts, pricing and exit terms secure?
No procurement content.
Operating model
Which teams own the service once it runs?
Citizen-engagement teams may see agent-mediated traffic.
Publication history
- 2026-11-02OpenAI · Issue 054 resources
Stable resource ID: openai-lifeval-everyday-tasks-government-services-eval-20261102