Changelog
Every entry below is live in the product. We publish verified changes only — no roadmap items and no reconstructed history.
Last updated 4 August 2026Current system status
Showing 10 changes
The first published release of the TaaSFlow changelog. Everything below is live in the product today; no earlier history has been reconstructed.
You can check availability yourself before raising a question. A service reads Unknown when it cannot be measured, so a green state always means something was actually checked.
When processing is delayed or an action is temporarily unavailable, the workspace says which area is affected instead of failing silently.
Each notification states what happened, what it affects, why it matters, when, and who acted, with one clear next action. Critical items stay until resolved.
Stalled stages, thin evidence, narrow pools and compressed scores are linked to a specific action and the data behind it, so numbers turn into decisions.
Each metric answers one decision question — and says when it cannot
How long until this role has someone worth a conversation?
4.2 days
median across 6 example roles
-4.9 daysvs previous example window (9.1 days)
| Item | Days to first qualified candidate | What it means |
|---|---|---|
| Role A | 3.1 | — |
| Role B | 4.2 | — |
| Role C | 5.8 | — |
| Role D | 9.4 | — |
Measured from the moment a blueprint is locked to the first evidence-backed candidate delivered for review.
How much of the scoring is backed by a quote from the source?
81%
of scored requirements carry a source quote
21 of 26 assessments have a verified quote attached.
Requirements without a quote are never counted as met — they surface for reviewer confirmation instead.
Review unverified evidence
5 assessments are waiting on a reviewer.
Which stage is losing the people you want?
Fewer than 5 completed decisions in this example window.
Drop-out is only reported once there are enough decisions to be meaningful. Until then the product says so instead of drawing a chart.
Role families, requirements, evidence types, scoring weights, compliance checks and approval controls now differ by vertical, and the differences are visible in the interface.
You can see the current stage, what has completed, and what is waiting on a decision without asking for an update.
Review · waiting on your team since 3 Aug
Review · waiting on your team since 3 Aug
Runs, queues and failures are inspectable with retry, pause, cancel and escalate controls. Inspector output is redacted, and every control writes an audit record.
Every run records actor, action, time, status and result
Blueprint agent
Compiled 8 requirements from intake + JD
09:02 · Rubric v4 locked
Discovery agent
418 profiles assessed against the blueprint
09:14 · 26 shortlisted for screening
Evidence agent
26 CVs parsed · quotes attached to 21
09:31 · 5 need reviewer confirmation
Scoring agent
12 scored under rubric v4
09:41 · Score run written to audit trail
Coordination agent
2 interview slots proposed
09:46 · Waiting on your confirmation
Each integration shows whether it is connected and responding, so you are not guessing which connections are live.
Tenant isolation is enforced in the database itself, so one organisation's data cannot be read through another account.
A blank screen now explains whether work is in progress, blocked, or genuinely empty — with no fabricated progress percentages.