Coming Soon

Sage

Reintegration Companion

Sustained connection for people reintegrating after custody — in the weeks when it matters most.

Sage will support people reintegrating after release from custody. Regular check-in calls that provide connection, gently detect risk, and help people navigate the practical maze of the first weeks out — housing, Medicare and scripts, Centrelink, ID. The aim is simple: that no one walks out the gate alone.

Sage serves the person, not the system.

It is not a supervision, parole, monitoring or surveillance tool, and it does not report to corrections. It builds trust by being on the person's side — because the evidence is clear that anything perceived as surveillance destroys engagement with people leaving custody.9

We are seeking partners in reintegration, throughcare and reentry support. If your organisation works alongside people leaving custody, we would welcome the conversation.

The fortnight after release

The most dangerous weeks of someone's life are often the ones right after they get out.

In the first two weeks after leaving prison, the risk of death is around 12.7 times that of the general population — led by drug overdose, because tolerance lost in custody makes a familiar dose suddenly fatal.2 In Australia, drug-related death risk in the first fortnight runs roughly four times higher than in the weeks that follow.3

And the cycle is relentless: 44.5% of people released from Australian prisons return to prison within two years, and 53.6% return to corrective services in some form.1 Around one in two people leaving prison expect to be homeless on release.4 Continuity of care breaks at the gate — barely half of planned exits leave with a health discharge summary.5

12.7×

higher death risk in the first 2 weeks after release, overdose-led2

~4×

AU drug-death risk in weeks 1–2 vs the weeks after3

44.5%

return to prison within 2 years of release1

~1 in 2

expect to be homeless on release4

What Sage does

Connection, risk detection and navigation — concentrated where the danger is.

Connection

A warm, regular voice that calls and listens — someone on your side in weeks that are often lonely and overwhelming. Relationships and a credible new identity are what sustain desistance in the research; Sage's connection function is modelled on that literature.8

Risk detection

Sage gently listens for the signals that precede a crisis — substance-use safety, mood and loneliness, a sudden loss of housing. The overdose window is real and short, so contact is concentrated in the first four weeks.3 Sage detects and flags; it does not diagnose or treat.

Navigation

The first weeks out are a paperwork emergency: restoring a Medicare card and scripts, lodging the Centrelink Crisis Payment within its 7-day deadline, replacing lost ID, finding somewhere to sleep tonight. Sage prompts these and warmly connects people to the real services that handle them.6

Engineered human escalation

When Sage hears genuine distress, the person is warmly guided to real human help — services they have chosen, never corrections by default. This path is built, tested and rehearsed, not left to chance. People make every clinical and safety call; Sage does not.

Modelled on what works

Designed from the throughcare evidence — honest about what's proven.

Why sustained contact

Continuity of contact across release is the common active ingredient in throughcare that works — the ACT Extended Throughcare evaluation found better drug-dependence, accommodation and treatment outcomes and lower recidivism in the first 120 days; NSW and Victorian programs run 12 weeks to 9 months.7 Sage's cadence mirrors those proven windows.

Why the first weeks

Death and crisis cluster in the fortnight after release, overdose-led, because tolerance is lost in custody.2 Continuing opioid agonist treatment roughly halves post-release overdose mortality.10 Sage front-loads its calls into that window and carries the tolerance-loss safety message conversationally.

Why trust comes first

Distrust of "the system" is the dominant barrier to help-seeking for justice-involved people; anything that feels like surveillance kills engagement.9 That is why Sage serves the person and never reports to corrections. There are no published trials of AI check-in calls for people leaving custody11 — so we claim modelling on human throughcare, not proven outcomes.

Powered by Kate

Every call Sage will make is orchestrated by Kate, the intelligence engine behind all CAREPLANS AI companions. Kate manages scheduling, emotional analysis, risk detection and engineered human escalation across every persona and every vertical.

Safety, privacy & what Sage is not

On the person's side — by design, in a domain where trust is everything.

Not supervision, parole or surveillance

Sage is a support-and-connection tool that serves the person. It is not a corrections monitoring, supervision, parole, e-reporting or compliance tool, and it does not report to corrections by default. Supportive contact is a different thing from electronic monitoring, and we keep them rigidly apart.

Not clinical care or diagnosis

Sage does not diagnose, treat or provide clinical care, and it is not a medical device. It listens, detects risk and flags for people to act on. It does not replace caseworkers, parole officers, OAT prescribers, housing services or crisis lines.

Genuine distress goes to human help

Sage is not a crisis service. Its escalation path is engineered and rehearsed: a person in distress is warmly guided to the services they choose — Lifeline 13 11 14, 13YARN 13 92 76, or 000 in an emergency — never to corrections.

Data handling

Data is stored in AWS Sydney (Australia). AI processing currently runs in the United States (Anthropic and Hume), with zero-data-retention in progress. We never train on customer data, and we do not share data with corrections by default.

If you or someone you know needs support now: Lifeline 13 11 14 · 13YARN 13 92 76 · NUAA PeerLine 1800 644 413 · Link2home 1800 152 152 · In an emergency call 000.

Partner with us

We are seeking reintegration, throughcare and reentry partners to bring Sage to people leaving custody. If you work alongside this community, we would welcome the conversation.

andrew@careplans.io

Sources

  1. Productivity Commission, Report on Government Services 2026 (Justice — Corrective Services), 2022–23 release cohort: 44.5% returned to prison within 2 years; 53.6% returned to corrective services. Returned-to-prison and returned-to-corrective-services are distinct indicators; figures move each ROGS edition.
  2. Binswanger I.A. et al. (2007). Release from prison — a high risk of death for former inmates. New England Journal of Medicine, 356:157–165. Risk of death in the first 2 weeks post-release ~12.7× the general population (95% CI 9.2–17.4), led by overdose (Washington State cohort, N=30,237).
  3. Merrall E.L.C. et al. (2010). Meta-analysis of drug-related deaths soon after release from prison. Addiction, 105:1545–1554. Australia-specific relative risk ~4.0 (95% CI 3.4–4.8) for drug-related death in weeks 1–2 vs weeks 3–12; mechanism is lost opioid tolerance.
  4. Australian Institute of Health and Welfare, The Health of People in Australia's Prisons 2022 (NPHDC, published 2023): ~48% of people leaving prison expect to be homeless on release (45% short-term/emergency accommodation; 2.8% rough sleeping).
  5. AIHW, The Health of People in Australia's Prisons 2022: only ~58% of planned prison exits leave with a health discharge summary; people in custody are ineligible for Medicare/PBS (s19(2)), so continuity of care breaks at the gate.
  6. Services Australia, Crisis Payment for release from prison: must be claimed within 7 days of release; pre-release Centrelink interview possible up to 21 days before release. Restoring Medicare, scripts and ID are day 1–7 tasks (AIHW; AMA).
  7. Australian Capital Territory Extended Throughcare evaluation (UNSW Social Policy Research Centre, 2021): better drug-dependence, treatment-participation and accommodation outcomes and lower recidivism in the first 120 days. NSW Initial Transitional Support runs 12 weeks; Victorian ReConnect up to 9 months. Pilot-era evidence; graded MODERATE.
  8. Desistance literature — Sampson & Laub (social bonds / turning points) and Maruna (identity / redemption scripts; "Redeeming desistance", Criminology, 2025): relationships and a credible new identity sustain desistance.
  9. Qualitative help-seeking studies of justice-involved people (Norwegian interview studies, 2023–24; Howerton et al., BMJ, 2007) consistently find distrust of "the system" and fear of negative consequences are the dominant barriers to help-seeking. Anything perceived as surveillance suppresses engagement.
  10. Bukten A. et al. (2024), Norwegian nPRIS cohort (N=112,877 releases), Lancet Public Health: opioid agonist treatment associated with HR 0.51 for overdose mortality (0.58 all-cause) in the 6 months post-release — roughly halving post-release overdose death.
  11. No published randomised trial of proactive AI phone check-in calls for people leaving custody exists (2023–2026 literature sweep). Adjacent SMS/phone navigation evidence (e.g. CARE+ Corrections) does not transfer automatically. Sage claims modelling on human-delivered throughcare evidence, not proven efficacy.

These statistics describe population research and the human throughcare programs Sage is modelled on — not Sage's own outcomes. Sage is in development and its effectiveness is under evaluation; there are no published trials of this modality with this population.