The Fortress Inmate Phone System (IPS) is the modern alternative to the legacy inmate calling vendor — built for small and mid-sized jails that want affordable per-minute rates, full call recording and AI-assisted monitoring, and a billing surface that ties directly into the inmate trust account. One platform replaces the wall-phone hardware, the third-party recording archive, and the call-monitoring vendor — at a fraction of the operating cost. Built on the same telephony stack as Fortress Communicator (the staff-side softphone), IPS inherits the same SIP backbone, the same WebRTC fork (LiveKit), the same FreePBX trunking, and the same operator confidence: you are not running a side-project, you are running the same platform the vendor’s own staff uses to make outbound calls every day.
What it does
- Provisions secure outbound calling to an approved-numbers list for every inmate. Calls connect through Fortress’s SIP backbone with WebRTC clients on tablets and traditional handsets in housing pods.
- Records every non-privileged call automatically, stores recordings with chain-of-custody metadata, and surfaces them in the Call Monitoring review queue with AI-generated transcripts and risk-segment highlights.
- Bills per-minute charges against the inmate’s trust balance through the same ledger used by Commissary & Trust — no separate IPS deposit account, no reconciliation headache.
- Honors approved-call lists, time-of-day restrictions, PREA-protected-victim blocks, and attorney-privileged exclusions as first-class policy objects — not after-the-fact filters.
How it works
Calls originate from a tablet WebRTC client or a pod handset, route through the Fortress IPS SIP server, and terminate at the called number through a tier-1 carrier. Recording, monitoring policy, and billing all attach at the SIP layer, so the recording and the billing row are co-signed by the same call id. Pre-paid balances debit in real time; calls cut off cleanly when the balance drops below one minute of forecast cost. The system speaks both G.711 (for legacy handset deployments that need analog gateway interop) and Opus (for the WebRTC tablet client) — encoding is negotiated automatically at SIP INVITE time.
Families top up an inmate’s balance through Top-Off at fortressdms.app — a deposit lands in the trust ledger within 60 seconds, ready to fund the next call. Per-minute rates are 40% markup on raw wholesale, well under what the incumbent inmate-calling vendors charge. Pre-paid is mandatory: there is no all-you-can-call plan, ever. Outbound minutes always debit a real balance, which keeps the model honest and the unit economics defensible to a county auditor who is going to ask whether the calling-rates were fair.
Administrators configure approved-number lists, time-of-day windows, PREA blocks, attorney-privileged numbers, and per-inmate monthly minute caps from a single staff page. Changes take effect on the next call setup; the prior policy is archived so a post-incident review can answer the question “what policy was in force when this call connected?”
Data captured
- Per call: inmate id, called number, peer name, start & end timestamps (UTC, displayed facility-local), duration, billed amount.
- Recording: encrypted file reference, chain-of-custody hash, retention class.
- Monitoring lifecycle: order id, trigger, analyst decisions, linked SAR or ThreatAssessmentCase ids (canonical case ids 1, 2, 3).
- Billing: ledger row in the trust account with per-minute rate and the call id that generated the charge.
Compliance & retention
IPS is PREA-aware: protected victims are blocked at policy level and the block is logged for audit. Attorney-client privileged calls are flagged at call-setup and excluded from recording and monitoring; the exclusion itself is auditable. Recording retention follows the facility’s records schedule; calls referenced in a Threat Assessment inherit the longer SB-68 retention window. Billing rows are append-only, just like every other money movement in the system — corrections happen as offsetting transactions, so the ledger always tells the truth about what happened.
Operators looking to migrate from an incumbent ICS vendor get a scripted cut-over: existing approved-number lists import via CSV, the existing call-history record can be archived out-of-band, and pre-paid balances on the inmate trust can be topped up immediately through Top-Off. Most cut-overs are complete in one shift; the per-minute savings show up in the first invoice cycle and the AI-assisted monitoring savings show up the first time an analyst clears a queue in an hour instead of a day.
We switched away from one of the big legacy vendors two years ago and the cost difference alone was enough to get admin buy-in immediately — glad to see more options coming into this space. My question is around the AI monitoring piece: how does it handle flagging calls in languages other than English, particularly Spanish, which we deal with constantly? That’s always been a weak point with other platforms we’ve evaluated.
We switched away from our old legacy provider two years ago and the savings alone basically paid for a software upgrade elsewhere in our budget. The AI monitoring piece is what I’m most curious about here — does it flag specific keywords in real time or is it more of a post-call review system?
We’ve been locked into one of the big legacy vendors for years and the per-minute rates are honestly embarrassing — glad to see competition driving this down. Quick question: does the trust-account billing integrate directly with the JMS inmate ledger, or is it a separate reconciliation process at end of shift?
We switched off our old legacy vendor two years ago and the per-minute savings alone justified the whole transition — curious whether this system integrates directly with the JMS for trust account deductions or if that’s still a manual reconciliation step. That part always ate up way too much of my staff’s time at booking.
We switched off one of the big legacy vendors two years ago and the per-minute savings alone paid for a facility equipment upgrade within eight months — wish we’d done it sooner. Quick question: does the AI monitoring flag calls in real time or is there a delay before staff get the alert? That distinction matters a lot for our threat assessment workflow.
We switched off our legacy vendor two years ago and the per-minute savings alone justified the whole transition — wish we’d done it sooner. My one question is how the AI monitoring handles calls in languages other than English, because we have a significant Spanish-speaking population and that’s always been a gap with older systems. Would love to see a breakdown on that specific feature.