A Shared Back Office for Golden Heart Senior Care
01Purpose and scope
This plan describes how Propel Team will take over after-hours scheduling for the fifteen Golden Heart Senior Care operators. A pilot with three to five operators begins on October 15, 2026, and the full network is scheduled to launch on January 1, 2027. The document is intended for operators, who need to know what will change in their agencies and what they are expected to do, and for Propel Team staff, who need a common reference for the protocol, the delivery model and the economics.
The pilot covers one workflow, which is filling shifts that open outside business hours. When a caregiver calls, texts or emails to say she cannot work a scheduled visit, the message will reach Propel Team instead of the owner, and Propel Team will find a replacement, update the schedule and notify the people affected. Candidate screening, hiring, new hire onboarding, new client onboarding, billing, payroll and marketing are excluded from the pilot because each has its own data requirements and decision rights, and adding them before the first workflow is stable would slow the pilot and make its results harder to interpret. Section 11 describes how the pilot is designed so that those workflows can be added later without rebuilding the foundation.
Figures marked VERIFY are planning assumptions that the pilot or vendor diligence must confirm before they are used in contracts or operator pricing.
02The case for a shared desk
After-hours coverage carries a high fixed cost because someone must be available for every hour outside business hours, regardless of how many calls arrive. A week contains 123 hours outside a weekday schedule of 8 a.m. to 5 p.m., and covering those hours with an on-call coordinator at a loaded rate of about $20 an hour costs roughly $10,650 a month. Most agencies with $1 million to $3 million in revenue cannot justify that expense, so the owner or a scheduler usually carries the phone, and some shifts that could have been filled go unfilled.
When fifteen agencies share one desk, they pay for the coverage hours once and divide the cost among themselves. Their calls are also spread across the night, so the combined workload is large enough to keep coordinators practiced without requiring fifteen times the staff. AI agents reduce the human workload further because most call-offs follow a predictable sequence of recording the absence, finding eligible caregivers, offering the shift and confirming the replacement. Coordinators can then concentrate on the calls that require judgment.
These savings depend on harmonization. A coordinator can handle a call from any operator efficiently only if the operators follow a common protocol, and each operator-specific convention adds training time and increases the likelihood of errors.
03The pilot workflow
What the desk does
Between the end of each operator's business day and the start of the next, the desk receives every scheduling-related call, text and email sent to the operator's after-hours line. It records call-offs and late arrivals, identifies the affected shifts and runs the operator's fill ladder to find replacements. Once a caregiver accepts, the desk confirms the assignment in writing, updates the schedule directly or through write-back depending on the integration, and notifies the client or family contact if the operator's policy requires it. Every interaction is logged. Each morning the desk sends the operator a handoff report listing filled shifts, open shifts, escalated decisions and items that need follow-up.
What the desk does not do
The desk does not make clinical judgments, accept new clients, discipline caregivers or change pay rates beyond the limits each operator sets. In an emergency it directs the caller to 911 where appropriate and then contacts the operator's on-call decision-maker immediately. Any report suggesting abuse, neglect or exploitation is passed to the operator at once, because the operator is the licensed provider and holds the reporting obligations under state law. The operator remains responsible for care delivery in every case, and the desk acts on the operator's behalf within written limits.
Service levels
The targets below are starting points. The pilot will measure each one against the operators' current performance, and the launch targets will be set from the pilot results.
| Measure | Pilot target |
|---|---|
| Inbound calls answered by the AI agent | Within 2 rings, at all hours |
| Texts and emails acknowledged | Within 2 minutes |
| Escalations picked up by a human coordinator | Within 5 minutes |
| Open shifts with at least 2 hours' notice that are filled | 85% VERIFY against pilot baseline |
| Operator alerted to a shift that cannot be filled | At least 60 minutes before shift start |
| Morning handoff report delivered | By 7:30 a.m. operator local time |
| Interactions logged with outcome | 100% |
04Mapping current workflows
Operators tend to describe their after-hours process as simple and consistent, although in practice it often varies with the person on call. Mapping produces a written record of what actually happens, and that record is the basis for the protocol in Section 5. The work runs from October 19 to November 6 for the pilot operators and continues into early December for the remaining operators.
Method
Mapping begins with the operator questionnaire in the Questionnaire tab, which records the agency's roles, systems, after-hours procedure, policies and other workflows, and from which a draft SOP and workflow map are generated for each agency. A structured interview of the owner and the person who currently handles after-hours calls follows, and it tests the questionnaire answers against the interviewees' account of how calls are answered, how call-offs are recorded, how replacements are chosen, which incentives are offered, when the owner is contacted and how clients are informed. A Propel Team analyst then reviews a week of actual after-hours contacts with the operator's permission, since practice under time pressure frequently differs from the description given in an interview. The analyst also checks the system answers in the questionnaire against sample exports from the agency's scheduling system.
Each operator receives a current-state map, drawn as a swimlane diagram across the caregiver, the person on call, the client and the scheduling system, along with a variance log. The variance log records each point where the operator's practice differs from that of the other operators, and during harmonization every entry is resolved either as a configurable setting or as a practice the operator agrees to change. The network comparison in the Questionnaire tab produces the first entries in the log by placing operators' answers side by side.
Baseline measures
During mapping, each operator also provides any available data on after-hours volume, such as weekly call-offs, the share of open shifts filled and the typical time to fill. Many operators will have limited records, in which case the week of reviewed contacts serves as the baseline. A baseline is needed so that the desk's performance in January can be compared with the arrangement it replaced.
05Operational harmonization
Harmonization separates the steps of the protocol from its settings. The steps are identical at every operator so that a coordinator in any region can handle a call from any operator correctly. The settings belong to each operator and reflect legitimate differences in how agencies are run, such as the amount of overtime an owner will approve or whether families are contacted at night about a caregiver change. Each entry in the variance log is assigned to one of these two categories.
The after-hours protocol
Every call-off follows the same sequence. The desk receives the contact, verifies the caregiver's identity against the roster, identifies the affected shift and checks how much notice remains. It then runs the fill ladder, offering the shift to eligible caregivers in the operator's chosen order and recording each offer and response. When a caregiver accepts, the desk confirms in writing, updates the schedule and notifies the client contact if the operator's policy requires it. If no caregiver accepts, the desk escalates to the operator's on-call decision-maker no later than the operator's alert threshold. Every step is logged, and the shift appears in the next morning's handoff report.
Operator settings
| Setting | What the operator decides |
|---|---|
| Coverage window | When business hours end and begin each weekday, and which holidays receive full-day coverage |
| Fill ladder order | The priority given to the client's preferred caregivers, caregivers who have served the client before, proximity, and caregivers with available hours below the overtime threshold |
| Overtime authority | Whether the desk may assign a shift that puts a caregiver into overtime, and up to how many hours |
| Shift incentive cap | The largest bonus the desk may offer without contacting the operator |
| Wake threshold | Which situations justify contacting the owner overnight, and how long before shift start an unfilled shift triggers a call |
| Client notification | Whether and when clients or families are told about a caregiver change, and through which channel |
| Escalation contacts | The ordered list of people the desk may call, with backup contacts |
| Outreach quiet hours | The hours during which caregivers should not be contacted about open shifts |
| Restricted-routing clients | Clients whose records may be handled only by the US desk, such as Medicaid or VA clients where contracts require it |
Decision rights
Disagreements over who was authorized to make a decision are a frequent source of friction in outsourced operations, so the decision rights are documented before the desk begins taking calls.
| Decided by | Decisions |
|---|---|
| AI agent | Record call-offs, verify identity, offer shifts to eligible caregivers in ladder order, confirm replacements within standard pay and incentive limits, send client notices that policy pre-approves, and produce the handoff report |
| Human coordinator | Calls outside the agent's confidence or policy, distressed callers, caregiver disputes, incentives up to the operator's cap, and choices among eligible caregivers for clients with particular needs |
| Operator | Overtime or incentives above the cap, client cancellations or reductions, any safety, clinical or abuse concern, disciplinary matters, and exceptions to the operator's own policies |
The data contract
The desk receives only the information it needs to perform the workflow, consistent with the HIPAA minimum-necessary standard and to limit the number of fields that must be kept synchronized. The contract covers caregivers (identifier, name, phone, consent to receive texts, languages, skills and certifications, availability, home ZIP code and hours already scheduled this week), clients (identifier, service ZIP code, care requirements relevant to matching, caregiver preferences and exclusions, payer routing flag and notification contact), shifts (identifier, client, caregiver, start, end and status) and each operator's settings. The desk writes back only shift reassignments, status changes and notes.
Integration tiers
Propel Team cannot require operators to change scheduling systems, so integration is tiered according to what each system supports. The tier determines how current the desk's information is, which affects how quickly and reliably shifts can be filled.
| Tier | Systems | Method |
|---|---|---|
| 1 · Native | ThinkHealth | Live read and write of rosters, availability and shifts VERIFY API scope |
| 2 · Connected | WellSky and other platforms with API access | Reads through the vendor API at short intervals, with write-back where permitted VERIFY per vendor and plan, including API fees |
| 3 · Bridged | Aanii and systems without accessible APIs | Scheduled exports (nightly at minimum, intraday where possible), with changes entered by coordinators and checked by the operator each morning VERIFY export options |
Operators on Tier 3 will receive slower service, because a change a caregiver makes to her availability during the afternoon may not reach the desk until the next export. Section 7 therefore recommends a price credit for operators on ThinkHealth to encourage migration without requiring it.
06Global delivery model
The human team is divided across three regions so that each team works during its own daytime. Overnight shifts tend to reduce accuracy and increase staff turnover, and assigning each block of US hours to a region where those hours fall within the working day avoids that problem while creating scheduled handoffs between regions.
| Site | Block (Eastern) | Pilot coverage | Candidate locations |
|---|---|---|---|
| Americas | 4 p.m. – 12 a.m. | Weekday and weekend evenings, plus the US-based desk | United States for restricted work, and Costa Rica, Colombia, the Dominican Republic or Mexico for general work |
| Asia | 12 a.m. – 8 a.m. | Every night, including the early-morning increase in call-offs | Philippines (recommended first site), with India, Vietnam or Indonesia as later options |
| Europe | 8 a.m. – 4 p.m. | Weekend and holiday daytime during the pilot, with weekday business-hours work as later workflows are added | Portugal, Poland or Romania, with South Africa as an alternative in the same hours |
The Sites tab compares each candidate location on cost, English proficiency, healthcare BPO experience, talent pool, alignment with the assigned hours and infrastructure, and it ranks the locations within each region according to weights that can be adjusted. The Philippines is recommended for the Asia site because it has an established workforce experienced in US healthcare support and strong English proficiency, both of which matter when coordinators speak with caregivers and families early in the morning. The Europe site carries the smallest workload during the pilot and could be deferred. Establishing it now would, however, give the business-hours workflows planned for 2027 a location without requiring a second recruiting effort.
Work that remains in the United States
HIPAA permits offshore access to protected health information when business associate agreements and appropriate safeguards are in place. Some state Medicaid programs and VA contracts impose stricter limits on offshore access, and those requirements vary by state and by contract VERIFY per operator state and payer. Because most operators serve private-pay clients, the majority of the work can be performed at any site. Clients flagged for restricted routing are handled only by the US desk, and because the flag is stored on the client record, later workflows will apply it automatically.
Handoffs and quality review
At each change of region, the outgoing team posts a written handoff listing open shifts, pending offers and unresolved escalations, and the incoming lead confirms receipt before the outgoing team ends its shift. A quality lead reviews a daily sample of AI and human interactions against a scorecard, and the findings are used for coordinator coaching and for revisions to the agent's instructions.
07Cost model and pricing
The pricing objective is a 10% margin over cost, at a price below what an operator would pay to provide comparable coverage on its own. The model below covers the after-hours scope across all fifteen operators after the January 1 launch. Labor rates are fully loaded estimates and are the largest source of uncertainty in the model.
Assumptions
| Assumption | Value |
|---|---|
| After-hours coverage per week | 123 hours (weekday 5 p.m. – 8 a.m. plus weekends) |
| Seats on shift | Americas 2, Asia 2, Europe 1 (230 seat-hours per week) |
| Loaded hourly rates | Americas nearshore $11, Asia $8, Europe $17, US desk $32 VERIFY |
| Shrinkage (training, breaks, absence cover) | 15% |
| Contacts fully resolved by AI without a human | 60% VERIFY in pilot |
| AI, telephony and messaging per operator | $300 per month VERIFY |
Monthly cost at launch (15 operators)
| Line | Monthly |
|---|---|
| Coordinators, Americas: 102 seat-hours per week at $11 | $4,858 |
| Coordinators, Asia: 112 seat-hours per week at $8 | $3,880 |
| Coordinators, Europe: 16 seat-hours per week at $17 | $1,178 |
| Shrinkage at 15% | $1,487 |
| Quality lead, Asia (1.0 FTE), and team lead, Americas (0.5 FTE) | $2,942 |
| US-based desk for restricted work (0.5 FTE) | $2,771 |
| AI agents, telephony and messaging (15 × $300) | $4,500 |
| Contact-center software, QA tooling and integration maintenance | $2,400 |
| Program management and overhead allocation | $2,500 |
| Total | $26,516 |
| Cost per operator | $1,768 |
| Price per operator at a 10% margin (cost ÷ 0.9) | $1,964 |
Recommended pricing
A single flat price would overcharge small operators and undercharge large ones. The recommendation is to tier prices by monthly billed hours, which correlates with call-off volume and is a figure every operator already tracks.
| Tier | Monthly billed hours | Price per month | Assumed operators |
|---|---|---|---|
| Small | Under 2,500 | $1,450 | 4 |
| Standard | 2,500 – 6,000 | $1,975 | 8 |
| Large | Over 6,000 | $2,750 | 3 |
| Network revenue and margin | $29,850 | 11.2% | |
The distribution of operators across tiers is an assumption VERIFY against operator billed hours, and the tier prices should be adjusted once actual billed hours are known so that the blended margin remains near 10%. For an operator with $2 million in annual revenue, the Standard price equals about 1.2% of revenue. The recommendation also includes a monthly credit of $150 for operators on ThinkHealth, funded by the reduced coordinator time that a live integration requires, to give operators a financial reason to migrate.
Alternatives available to operators
Operators will compare this offer with the arrangements available to them. Hiring an in-house overnight coordinator costs about $10,650 a month for one person and provides no backup coverage, quality review or reporting. An answering service costs roughly $300 to $1,500 a month VERIFY and records messages for the operator to act on the next morning. Continuing the current arrangement involves no direct expense, although it depends on the owner or a scheduler being reachable overnight and leaves some shifts unfilled.
Sensitivity and pilot economics
The assumption with the greatest effect on cost is the share of contacts the AI resolves without human involvement. If that share is 40% instead of 60%, the desk needs one additional seat in both the Americas and Asia, which adds about $5,000 a month and raises the average price required for a 10% margin to about $2,325. If the share reaches 80%, one Asia seat can be released, saving about $2,200 a month. The pilot should therefore measure the AI resolution rate before launch prices are finalized.
The pilot will operate at a loss because a three-site desk costs nearly as much to staff for four operators as for fifteen. The recommendation is that pilot operators pay nothing during shadow mode, pay half of their tier price during supervised live operation in December, and receive their launch price fixed for twelve months. Propel Team should budget the pilot deficit, estimated at $40,000 to $50,000 excluding engineering VERIFY, as a startup cost of the business.
08Implementation timeline
The program runs eleven weeks from kickoff to launch. That schedule is short for establishing a new desk across three regions, so the plan relies on defined go/no-go gates to control the risk of launching before the desk or an operator is ready. The Timeline tab shows the full schedule, and the Board tab lists the individual tasks.
| Phase | Dates | Exit criteria |
|---|---|---|
| Mobilize | Oct 15 – Oct 30 | Charter approved, pilot cohort confirmed and pilot BAAs signed |
| Map | Oct 19 – Nov 6 | Current-state maps, variance logs and system inventories signed off by pilot operators |
| Harmonize | Nov 2 – Nov 20 | Protocol v1 approved, pilot configuration sheets complete and data contract fixed |
| Build and staff | Nov 2 – Dec 4 | Integrations tested for each tier, AI agent passing scripted tests, and pilot desk hired and trained |
| Pilot | Nov 23 – Dec 18 | Shadow mode (Nov 23 – Dec 4), then supervised live operation (Dec 7 – Dec 18), with service levels met for 10 consecutive days |
| Network onboarding | Nov 2 – Dec 23 | Every remaining operator has completed the readiness checklist in Section 9 |
| Launch | Dec 21 – Jan 1 | Dress rehearsal passed, change freeze from Dec 24, all ready operators live Jan 1, and hypercare through Jan 15 |
Two gates govern the launch. The pilot go/no-go decision on December 4 determines whether the pilot moves from shadow mode to live operation. The network go/no-go decision on December 18 determines, operator by operator, which operators launch on January 1. An operator that has not completed its readiness checklist will join a second wave on January 15, since problems during the first nights of service would undermine operator and caregiver confidence in the desk. Shadow mode includes Thanksgiving and the launch falls on New Year's Day, and call-offs typically rise on both holidays. Those periods will test the desk under heavier load, so the desk must be fully staffed for both, and operators should keep their existing on-call arrangements in place as a backup throughout shadow mode.
09Operator implementation guide
This section is addressed to each operator's owner and implementation lead. Pilot operators follow the earlier dates shown on the Board and Timeline, and the remaining operators follow the network onboarding dates listed below. Completing every item is a condition of launching on January 1.
- Nov 6Name an implementation lead who can commit two to three hours a week, along with an on-call decision-maker and a backup for escalations.
- Nov 13Complete the operator questionnaire in the Questionnaire tab, which covers your roles, systems, after-hours procedure, policies and other workflows, and send the exported file to the implementation team. Pilot operators complete it by October 23.
- Nov 20Complete the mapping interview and, if you agree, allow Propel Team to review a week of after-hours contacts. Send any data you have on call-off volume and fill rates.
- Dec 4Complete your configuration sheet, which records your fill ladder, overtime authority, incentive cap, wake threshold, client notification policy, escalation contacts and restricted-routing clients.
- Dec 4Sign the business associate agreement and the services addendum.
- Dec 11Update your caregiver roster so that every caregiver has a current mobile number, recorded consent to receive texts, availability, skills and any client preferences or exclusions. Automated texting without recorded consent creates exposure under the TCPA, so the desk will not text caregivers whose consent is missing.
- Dec 11Grant integration access to your scheduling system, or set up the scheduled export if your system is on Tier 3.
- Dec 18Configure after-hours call forwarding and the after-hours text number, and test both with the desk.
- Dec 23Send the caregiver and client announcements using the templates Propel Team provides, so that caregivers and clients know whom they will reach after hours.
- Dec 21–23Take part in training and in the dress rehearsal, during which the desk handles simulated call-offs for your agency from start to finish.
- Jan 1 →Read the morning handoff report each day during hypercare and respond to any flagged item by 10 a.m. Weekly review calls continue through January.
Shared services frequently fail when owners continue answering after-hours calls out of habit. Caregivers then learn to contact the owner directly, and the desk does not develop familiarity with the agency. Operators who join are asked to route all after-hours calls to the desk from the first night and to raise any concerns through the morning report.
10Governance and risk
A steering group made up of Propel Team leadership, Golden Heart franchise leadership and one pilot operator will meet weekly through launch and monthly afterward. The group owns the protocol, approves changes to service levels and pricing, and makes each go/no-go decision. Protocol changes are versioned and announced to all operators at least one week before they take effect.
| Risk | Mitigation |
|---|---|
| The eleven-week schedule slips | Two go/no-go gates, with a January 15 wave for operators who are not ready |
| Hiring for three sites within six weeks | Begin with an established BPO partner at the Asia and Americas sites, with an option to bring the work in-house in 2027 |
| AI resolves fewer contacts than assumed | Measure the rate during shadow mode, staff for the measured rate and adjust launch pricing before contracts are signed |
| Tier 3 data is out of date | Intraday exports where possible, morning checks by the operator and the ThinkHealth migration credit |
| Offshore restrictions for certain payers | Restricted-routing flag on client records, with flagged work handled by the US desk |
| Missing consent to text caregivers | Roster update in the operator checklist, and no automated texting to caregivers without recorded consent |
| Owners continue answering after-hours calls | Announcements to caregivers and clients, call forwarding, and the morning report as the channel for concerns |
| Liability for a missed visit | The services agreement states that the operator remains the provider and that the desk acts within written limits on a reasonable-efforts basis VERIFY with counsel |
11Designing for later workflows
After-hours fill-in is the first of several workflows the desk will take on, and decisions made during the pilot will determine how much later workflows cost to add. The principles in this section are intended to keep that cost low, so that adding candidate screening in 2027 requires considerably less effort than the pilot did.
Each workflow is defined as a protocol together with a set of operator settings. When an operator needs different handling, the change is made as a new setting available to every operator, which keeps the desk's work consistent across agencies. The data contract is extended as workflows are added, so screening adds candidate fields to the existing caregiver record and billing adds visit and payer fields to the existing shift record. Agent skills are built as separate, versioned components, and each new version is tested against recorded interactions before release so that a change to one workflow does not disrupt another.
Decision rights are documented for each workflow before it goes live, using the same structure as in Section 5. Compliance routing is attached to the work item instead of to a particular team, so that billing and payroll, which involve more sensitive data, apply the restricted-routing flag and the minimum-necessary standard without additional engineering. Each workflow is measured during mapping before any part of it is automated, because those baselines are how operators will judge the value of the service. Pricing is set per workflow so that operators can adopt services in sequence.
The recommended order after launch begins with candidate screening and hiring, because that workflow involves little protected health information, has high volume and has a clear measure of success in hires per month. New client and new hire onboarding would follow, then billing and payroll once the integration tiers are mature, and then inbound and outbound marketing. Operators that move to ThinkHealth before the later workflows are introduced will be able to adopt them sooner and at lower cost.
12Appendix: board synchronization
The Board tab saves to the viewer's browser by default, so each person sees a separate copy. Setting SYNC_URL at the top of this file connects the board to a shared backend so that operators and Propel Team staff work from the same board. The backend must support two operations. A GET request returns the board as JSON containing a version number and the list of cards. A PUT request saves the board and includes, in an If-Match header, the version it was based on. If another user has saved a newer version, the server returns status 409 with that version, and the page loads it. Each operator receives its own bearer token in SYNC_TOKEN. The board is a project tracker, so protected health information should not be entered on it.