What's New in IRIS

IRIS — What's New (V3.4 Auto-Approve · V3.3 On the Floor · V3.2 Self-Sufficient · IRIS 3.0)

The latest platform release is V3.4 (July 2026) — IRIS Auto-Approve — on top of V3.3 "On the Floor" (June 2026) and V3.2 "Self-Sufficient" / IRIS 3.0.

New in V3.4 — IRIS Auto-Approve (optional add-on)

Distinct from Generate / Fill Gaps — IRIS Auto-Approve approves existing open shift claims and accepted swaps only. It does not create assignments from scratch.

Strict-clean gating — rest, hours, craft, scope, overlap, and leave must all pass; no admin override path
Real-time mode — instant approval when staff claim or accept a swap
Backlog mode — Living Roster → IRIS Auto-Approve → preview, whitelist run, undo, finalise before publish
Two-tier entitlement — platform availability, then org toggles for open shifts and swaps separately
Windows and exclusions — min notice, max horizon, date bounds, stream/craft/staff exclusions
Full audit trail — Logs & Audit → IRIS Auto-Approvals, badges, Assignment Intelligence column
Leave never auto-approved

Setup guide: IRIS Auto-Approve. Staff guide: What staff should expect.

New in V3.3 On the Floor

Duty Board:

Daily staffing view in the sidebar — check-ins, coverage summary, no-show alerts, notices, and proxy roles
Huddle Mode (LCD) for wall displays
Configurable tabs (Staffing plus CMS static/updates pages)
Two-tier enablement: platform availability, then org switch

AUDITOR role:

Dedicated read-only role for compliance officers
Reports (all exports) and Audit Log on the Reports page
Living Roster read-only; no IRIS, no admin write access

Apply Master Roster:

Override compliance rules toggle at apply time (default on) — bulk create despite overridable breaches; reason stamped per shift
Full skip reporting: leave, conflict, inactive period, no vacancy, rule violation, vacancy full, plus created via compliance override count
Faster apply and corrected created counter on large multi-stream runs

New in V3.2 Self-Sufficient (IRIS 3.0)

IRIS speaks for itself:

Live narration — a thinking rotator describes what IRIS is doing through every fortnight batch, so you never wait in silence
End-of-session summary — fill rate, distinct staff rostered, stretch shifts used, rule notes, and unfilled slots, surfaced in chat in plain language as soon as Generate finishes
Free-text parity — typing "roster the AM shift on Monday" gets the same lively narration as the full Generate workflow
Fill the Gap progress — per-stream progress messages so it's clear what IRIS is considering, not just a spinner

Honest diagnostics:

Rejection-reason aggregation — "couldn't fill" now reads as "couldn't fill because 78% of rejections were craft-group mismatches"; a clear signal of where qualifications gaps live
Pre-existing over-max called out separately — staff already pushed over their fortnight max by the master roster are reported distinctly from anything IRIS introduced
Per-fortnight leave-adjusted measurement — summary metrics use the same leave-adjusted limits the engine applied at runtime
Under-minimum sampling with blocked-by reason — staff who got zero IRIS shifts are surfaced alongside why (hard-filter vs scoring)

Fairness as a first-class axis:

Tier-first sort — under-minimum staff who haven't yet received a shift in the current batch are scored ahead of staff already at or over their minimum; the V3.1 contract becomes structural, not just a soft preference
Per-batch IRIS shift tracking — the spread is intentional and inspectable in the summary

Fast enough to stay with:

~15× faster full-organisation Generate — large departments complete in roughly 2 minutes where they previously took ~30
Single-flight concurrency — two simultaneous generations for the same organisation are no longer possible; a clear message replaces silent state pollution
Fortnight carry-in correctness — starting mid-fortnight now counts week-1 shifts already assigned (or published) against the fortnight's hour bucket

Correctness floor raised:

AI rules now load and evaluate regardless of how ruleType or ruleLevel is cased in the database
Crash on the final fortnight is fixed — generation completes cleanly even on long full-year runs
Roster Preview, Apply Master Roster, and Fill Gaps modals no longer overlap the IRIS chat panel — the end-of-session summary is finally readable while reviewing the preview

From V3.1 Fair Minimums (still current)

For Roster Builders:

Near-hard minimum hours — Generate and Fill Gaps strongly prioritise under-minimum staff through a single shared hours policy
Tolerated over-max notes — one bounded exception when shift granularity makes the max unreachable without dropping below minimum; visible in Roster Health, not counted as violations
Multi-stream generation — Generate with IRIS, Apply Master Roster, and Fill Gaps now process all selected streams with fortnight batching
Admin hours separation — administrative hours excluded from clinical min/max calculations so admin-heavy staff get the correct clinical load
Honest AI provenance — run logs record the model that actually ran; soft rules that cannot be enforced show warnings instead of silently doing nothing

For Staff:

Help knowledge base — IRIS answers "how do I...?" questions using built-in guides (leave amendments, open shifts, swaps, and more)
Personal data queries — ask about your shifts, leave, availability, hours, and fairness stats

From V3.0 Cadence (still current)

Leave-aware hours — Fill the Gaps and staff hour calculations account for approved leave and partial leave
Vacancy eligibility — IRIS can check whether a specific staff member is eligible for an open vacancy
Roster Health upgrades — craft-group filter, auto-expand conflicts on open, unassign shifts directly from the health modal
Day-of-week fairness — improved fairness distribution analysis across weekdays
Smart shift removal — IRIS suggests which shifts to remove when reducing hours, with fortnight-aware checks
Activity insights — expanded plain-English summaries of organization activity logs

From V2.0 Phoenix (still current)

Fill-the-Gaps Intelligence — smarter gap detection with distribution context
Roster Health integration — analyze violations and suggest remediations with confidence scores
Expanded tool library — 20+ tools for scheduling, availability, vacancy, and compliance
Conversation memory — IRIS remembers the last 10 messages for natural back-and-forth conversation
Streamed responses — see IRIS thinking and answering in real time

How IRIS works behind the scenes

IRIS uses a tool-use architecture — instead of dumping all data into one prompt, IRIS calls purpose-built tools on demand (querying assignments, checking availability, validating compliance). This produces faster, more accurate, and more contextual responses.

Credit system

Each interaction uses credits:

Operation	Credits
Chat question	5
Staff chat	2
Generate roster	50
Report	10
Activity insights	3

Your organization's admin manages credit allocation.