Second Shift is a library of turnkey departments: 53 suites spanning Finance, Tax, Legal, and 50 more. Each suite is a team of specialist packs, and each pack has a fixed job, a least-privilege tool set, a risk ceiling, and a human who signs off. You don't prompt them; you turn them on.
A pack is a single specialist agent with one fixed job; a suite is the department that bundles them. Each pack drafts and recommends; a credentialed human reviews and commits. 2,020 packs across 53 suites, ready to switch on.
Every agent has one bounded persona and one task: a 13-week cash forecaster, a 1099 extractor, a citation verifier. No open-ended prompt to engineer.
Each agent gets a least-privilege tool set and a risk ceiling it cannot exceed: a narrow allow-list, an explicit deny-list, bound to every action.
Agents draft and recommend; they don't commit. Work routes to a credentialed reviewer who certifies and releases it, maker-checker, built in.
The finance org as agents: controllership, FP&A, treasury, assurance, and reporting. A hard “never move money without a human” guardrail, amount-aware approval tiers, a segregation-of-duties linter, and OFAC/SDN sanctions screening.
A documents-to-draft workflow: classify and extract W-2, the 1099 family, K-1, and 1098 into a standardized workpaper, then a first-pass 1040 plus the resident-state return. Every line cites its source document. It never files.
Contracts, litigation support, privacy regimes (GDPR/CCPA/PIPL/APPI plus US state laws), AI and emerging-tech counsel, antitrust, and board governance, with citation verification that strips any authority it can't confirm.
The same discipline, pre-tuned for the buyer. Corporate-function suites cover the org chart; industry suites ship bundled with that industry's regime and jurisdiction content. Each line is a suite, with its specialist-pack count.
People & HR · 88
Sales & GTM (RevOps) · 81
IT & GRC · 76
Legal · 76
Finance · 70
Product & Engineering · 67
Operations · 53
Marketing · 52
Strategy & Corp Dev · 51
Data & Analytics · 43
Customer Experience · 41
Security Operations · 36
Procurement · 34
Tax · 31
Facilities & EHS · 20
Trust & Safety · 20
Executive Office · 20
Enterprise Risk · 20
ESG & Sustainability · 20
Process Automation · 20
Knowledge Management · 16
Insurance · 53
Banking · 51
Healthcare · 50
Retail · 47
Logistics · 47
Manufacturing · 47
Construction · 45
Real Estate · 40
Utilities · 40
Government Contracting · 40
Professional Services · 40
Education & Nonprofit · 38
Telecom & Media · 37
Hospitality · 35
Pharma & Life Sciences · 31
Private Equity & VC · 30
Capital Markets · 30
Travel & Aviation · 30
Oil & Gas · 27
Aerospace & Defense · 25
Agriculture · 25
Public Sector · 25
Automotive · 25
Chemicals · 24
Semiconductors · 24
Crypto & Digital Assets · 24
Water Utilities · 24
Renewables & Cleantech · 24
Maritime · 24
Medical Devices · 24
Mining & Metals · 24
Food, Beverage & CPG · 24
Before a pack ships, every agent clears lightwork domains-lint: a bounded persona, a least-privilege allow-list, an explicit deny-list, and a risk ceiling. The same gate, across all 2,020.
Packs attach to any platform tier. On Gold and Platinum, the governance and learning layers wrap around them (capability tokens, the policy engine, the signed audit, and provable improvement), so a department doesn't just work, it clears a review and gets better under proof.
An agent factory synthesizes a new pack — from a description of the job, or a demonstration of a person doing it — through the same governed intake: clamped to a least-privilege envelope and cleared by domains-lint before it ships. It improves with every run, and never widens an existing pack's envelope. See Daybreak Forge →
Pick the pack closest to the work in front of you. We'll scope it on your data, under governance you can prove.
