Organization-scoped secrets, webhooks, CRM connections, telephony, billing, permissions and operational visibility.
Prism.AI Multi-Tenant Platform Proposal
Prepared for Jennifer Alvarez, Founder, Prism.ai
Technical implementation plan for multi-tenancy, integrations, automated billing, access controls, reliability and measured scale.
Executive summary
Prism.AI progresses through three technical stages: a controlled multi-tenant foundation, commercial operation, and a validated scale-stage platform. Each stage is completed against its agreed scope and validation criteria rather than against a fixed calendar date.
The technical roadmap is the primary scope. Dates and salon counts are planning references, not automatic phase triggers. AXIS will assess completion against the agreed scope, validation criteria, and measured system behavior.
The business outcome is operating leverage. Standardized tenant setup, reusable CRM adapters, automated billing, and progressively more self-service are intended to reduce the incremental effort required to activate and support each additional salon while keeping system cost and capacity visible.
Phorest + Zenoti in Phase 1, with two additional integrations added in Phase 2.
Capacity is validated by measured system behavior; reevaluation can occur before or after the target is reached.
How the engagement is run
The implementation is structured to prove each operating layer before adding more complexity or scale.
The three implementation phases
The differences are shown first so the technical detail that follows has a clear frame.
Phase 1: Foundation
Prove isolated multi-tenant operation for the first production group.
Tenant separation, CRM round-trip, telephony, tracking, CI/CD and production controls working.
Guided; manual third-party and access work expected.
Phase 2: Commercial
Operate commercially with automated Stripe billing and a broader CRM footprint.
Subscriptions, plan limits, dashboards, two added CRM adapters, quotas, margin visibility and more onboarding automation.
Assisted, with more routine setup handled inside Prism.AI.
Begins after Phase 1 scope is validated and the client confirms progression.
Phase 3: Scale
Validate a repeatable self-service operating model and scale-stage controls.
Routine zero-touch path, CRM switching, cost governance, metered billing, queue scaling, rollback and failover.
Driven by measured capacity,not salon count alone.
Phase comparison
A compact reference for what changes from one phase to the next.
Multi-tenant system architecture
The tenant layer is the central control plane. CRM, communications, billing, analytics and access remain organization-scoped.
Encrypted org secrets · org-scoped webhooks · workers across tenants · feature entitlements · usage and cost attribution
Fourteen technical workstreams
The summary stays compact; open any workstream to see the Build → Extend → Scale detail.
01 · Lead intake forms
Native builder, embeddable script, org-scoped submissions.
Conditional logic, scoring, spam/bot filtering.
Templates, A/B variants, attribution.
02 · CRM / booking support
Phorest + Zenoti end to end.
Shared abstraction + two additional adapters; four total.
Switching among implemented adapters.
03 · Customization features
Branding, sender name, business hours.
Scripts/prompts, service menus, server-side overrides.
White-label/template controls as released.
04 · Twilio A2P & toll-free
Submission flow; manual assistance.
Automated verification and status.
Bulk provisioning; subaccount isolation.
05 · Numbers, voice & AI cost
Number/forwarding APIs, Jambonz, ElevenLabs, per-call cost.
Per-org cost/margin and usage caps.
Cost governance and voice-tier switching.
06 · Business limits by plan
Plan schema stored; limits recorded.
Limits enforced; overages and upgrades.
Usage-based tiers and annual plans.
07 · SMS & call limits
Org-scoped send/receive; volume logged.
Rate limits, quotas, throttle/notify.
Burst handling and carrier failover.
08 · Stripe integration and billing
Customer billing deferred.
Checkout, subscriptions, invoices, failed-payment recovery.
Metered usage, proration and tax.
09 · Development and billing dashboards
Internal org, health and logs view.
Dev dashboard + salon billing dashboard.
Self-service account management/export.
10 · Meta and Google approvals
Submit week one; Meta OAuth migration.
Token/permission handling hardened.
Re-verification and policy monitoring.
11 · GA4 and Meta Pixel setup
Guided/manual where needed.
In-app wizard and OAuth attach.
Zero-touch routine path.
12 · Feature flags and access control
Role gating; hardcoded flags.
Per-org flag service and entitlements.
Gradual rollout, targeting, kill switches.
13 · Hosting
RLS, org secrets/webhooks, core hosting.
Autoscaling, backups, monitoring.
Read replicas, queue scaling, load testing.
14 · Staging and production
Staging/prod, CI/CD, seeded test org.
PR previews and migration discipline.
Blue-green deployment and rollback.
| Workstream | Phase 1 | Phase 2 | Phase 3 |
|---|---|---|---|
| Lead intake | Native builder; embeddable script; org-scoped submissions | Logic, scoring, spam/bot filtering | Templates, A/B variants, attribution |
| CRM / booking | Phorest + Zenoti | Shared abstraction + two added adapters | Switching among implemented adapters |
| Customization | Branding, sender, business hours | Scripts/prompts, service menus, server-side overrides | White-label/template controls |
| A2P / toll-free | Submission flow; assisted registration | Automated verification/status | Bulk provisioning; subaccount isolation |
| Numbers / AI cost | Forwarding APIs, Jambonz, ElevenLabs, per-call cost | Per-org cost/margin, usage caps | Cost governance; voice-tier switching |
| Plan limits | Schema stored | Limits enforced; overages/upgrades | Usage tiers; annual plans |
| SMS / calls | Org-scoped send/receive | Rate limits and quotas | Burst handling; carrier failover |
| Stripe | Customer billing deferred | Checkout, subscriptions, invoices, recovery | Metered usage, proration, tax |
| Dashboards | Internal ops view | Dev + salon billing dashboard | Self-service management/export |
| Meta / Google | Initial submission + Meta OAuth | Permission/token hardening | Re-verification and policy monitoring |
| GA4 / Pixel | Guided/manual setup | Wizard + OAuth attach | Zero-touch routine path |
| Feature flags / access | Role gating | Per-org flags + entitlements | Gradual rollout + kill switches |
| Hosting | RLS, org secrets/webhooks | Autoscaling, backups, monitoring | Read replicas, queues, load testing |
| Staging / production | Staging/prod, CI/CD | PR previews | Blue-green deploys + rollback |
Phase 2 CRM shortlist
Two additional integrations are selected for Phase 2. The shortlist is evaluated by integration fit, with Meevo retained as a secondary candidate pending deeper endpoint validation.
Vagaro
API-capable salon platform suitable for booking/client integration.
Mindbody
Mature developer ecosystem with booking and business-data integration paths.
Square Appointments
Bookings connects with Square customer, location, team and catalog data.
Boulevard
Appointment-oriented integration model with event-driven integration capability.
Meevo
Salon-management platform retained as a secondary Phase 2 candidate pending validation of the specific booking, client, service and event capabilities Prism.AI requires.
Capacity review and phase progression
Salon count provides a business planning target. AXIS will use observed system behavior, completed scope, and validation results to assess when additional capacity or phase progression is appropriate.
AXIS monitors the indicators that reflect actual system load and operating demand.
Where monitoring indicates that the current operating range may need to expand, AXIS will complete a system audit, review the findings with the client, and use a minimum two-week validation period to support its recommendation, unless reliability or security requires earlier action.
Move forward when measured demand justifies it. Stay in the current phase when validated thresholds remain healthy,even if the salon target has been reached.
Phase validation criteria
Each phase is validated against the agreed technical outcomes and scope. A later phase does not begin automatically because a planning date or salon target has been reached.
Phase 1 validation
Phase 2 validation
Phase 3 validation
Phase transition process
The planning timeline does not create an automatic commitment to the next phase.
AXIS phase pricing and payment schedule
Each phase is commercially independent. The initial 50% is due before work begins on that phase. The remaining 50% is paid in four equal installments due 30, 45, 60, and 75 calendar days after the actual phase commencement date.
Support and availability schedule
The support period is tied to the actual commencement of each phase, consistent with the Service Agreement.
Implementation and operational support for the applicable phase are included during this period.
If agreed scope or validation criteria remain incomplete because of work within AXIS's responsibility, AXIS continues the required work without an added support fee solely because four months have elapsed.
Where AXIS cannot proceed because required client access, credentials, approvals, or decisions are outstanding, the support period is paused for the documented blocked period.
Monday through Friday, excluding mutually agreed closures. Guaranteed 24/7 support is not included unless separately agreed.
Estimated operating cost by lead volume
Operating cost varies primarily with lead volume and the modules a salon uses. The planning case below uses 25% utilisation of variable usage lines; fixed platform and account costs remain unchanged.
Commercial structure and change control
AXIS service fees remain separate from operating costs, and custom engineering is handled through a simple scope-control process.
AXIS commercial structure
The AXIS phase fees and payment structure are shown in the pricing section above. Each later phase is activated only after the current phase has been validated and the client confirms that it wishes to proceed. Any work beyond the stated phase scope is reviewed separately. A separate service agreement will document the final commercial and operating terms. Third-party platform costs, Stripe processing and selected CRM/API charges remain separate variable costs.
Standard setup boundary
Existing Prism.AI configuration is standard setup. Requests that require new platform functionality, unsupported integrations, custom workflows, specialized metrics or changes to released Prism.AI behavior will be reviewed as additional engineering scope.
Detailed cost model
Internal planning detail. Vendor list prices checked August 2026. Consumption per lead remains modelled and should be validated during beta.
Unit rates and consumption assumptions
| Input | Rate / assumption | Basis |
|---|---|---|
| Claude Opus 5 voice | $0.100 / call minute | ~4 turns/minute, ~880 output tokens/turn, cached context, output billed at $25/MTok. |
| Claude Opus 5 text | $0.775 / inbound message | One agentic run: ~400k input tokens across loop + ~3k output. |
| Cache reads | $0.150 / run component | 300k tokens at $0.50/MTok, 75% of input. |
| Fresh input | $0.300 / run component | 60k tokens at $5.00/MTok, 15% of input. |
| Cache writes | $0.250 / run component | 40k tokens at $6.25/MTok, 10% of input. |
| Output | $0.075 / run component | 3k tokens at $25/MTok. |
| ElevenLabs agent minute | $0.080 | Overage rate; $0.16 if concurrent calls exceed plan limit. |
| Telephony minute, blended | $0.0094 | 60% outbound at $0.013, 40% inbound at $0.004 over SIP trunking. |
| jambonz licence | $35.00 / month | Per salon, per tenant. Fixed regardless of call volume. |
| Outbound SMS segment | $0.012 | Base + carrier surcharge, 160 characters per segment. |
| Inbound SMS segment | $0.0079 | Twilio inbound. |
| US local number | $1.15 / month | One per salon. |
| A2P campaign | $10.00 / month | Standard use case; low-volume mixed is $1.50. |
| Instagram DM delivery | $0.00 | No per-message carrier cost through Meta API; inference only. |
| Phone minutes per lead | 20.0 min | Full lead lifecycle assumption. |
| Inbound SMS per lead | 2.5 messages | Calibrated to 200–300 messages/month at 100 leads. |
| Outbound SMS per lead | 5.0 segments | Replies plus follow-up sequence. |
| Inbound Instagram DMs per lead | 4.0 messages | DM threads modelled as chattier than SMS. |
What one lead costs, by module
| Module / component | Volume per lead | Rate | Cost per lead | Per 100 MAX | Per 100 AVG |
|---|---|---|---|---|---|
| Phone System AI add-on | |||||
| ElevenLabs agent minutes | 20.0 min | $0.080 | $1.600 | $160.00 | $40.00 |
| Opus 5 voice inference | 20.0 min | $0.100 | $2.000 | $200.00 | $50.00 |
| Telephony minutes, blended | 20.0 min | $0.0094 | $0.188 | $18.80 | $4.70 |
| Phone variable subtotal | $3.788 | $378.80 | $94.70 | ||
| SMS AI add-on | |||||
| Opus 5 agentic text runs | 2.5 runs | $0.775 | $1.938 | $193.75 | $48.44 |
| Outbound SMS segments | 5.0 seg | $0.012 | $0.060 | $6.00 | $1.50 |
| Inbound SMS segments | 2.5 seg | $0.0079 | $0.020 | $1.98 | $0.49 |
| SMS variable subtotal | $2.017 | $201.72 | $50.43 | ||
| Instagram AI add-on | |||||
| Opus 5 agentic text runs | 4.0 runs | $0.775 | $3.100 | $310.00 | $77.50 |
| Instagram variable subtotal | $3.100 | $310.00 | $77.50 | ||
| All three add-ons, variable only | $8.905 | $890.52 | $222.63 | ||
MAX assumes full modelled usage by every lead. AVG applies 25% utilisation to all variable lines. Fixed costs do not change.
Monthly cost per salon, by lead block
| Line | 100 leads | 200 leads | 300 leads | 400 leads | 500 leads |
|---|---|---|---|---|---|
| Fixed: platform + jambonz + number + A2P | $60.65 | $60.65 | $60.65 | $60.65 | $60.65 |
| Phone variable, AVG | $94.70 | $189.40 | $284.10 | $378.80 | $473.50 |
| SMS variable, AVG | $50.43 | $100.86 | $151.29 | $201.72 | $252.16 |
| Instagram variable, AVG | $77.50 | $155.00 | $232.50 | $310.00 | $387.50 |
| Cost to serve, AVG | $283.28 | $505.91 | $728.54 | $951.17 | $1,173.81 |
| Cost per lead, AVG | $2.83 | $2.53 | $2.43 | $2.38 | $2.35 |
| Cost to serve, MAX | $951.17 | $1,841.70 | $2,732.22 | $3,622.75 | $4,513.27 |
| Cost per lead, MAX | $9.51 | $9.21 | $9.11 | $9.06 | $9.03 |
Add-on combinations, MAX
| Configuration | 100 | 200 | 300 | 400 | 500 |
|---|---|---|---|---|---|
| Base only | $14.50 | $14.50 | $14.50 | $14.50 | $14.50 |
| Base + SMS | $227.38 | $429.10 | $630.83 | $832.55 | $1,034.28 |
| Base + Instagram | $324.50 | $634.50 | $944.50 | $1,254.50 | $1,564.50 |
| Base + SMS + Instagram | $537.38 | $1,049.10 | $1,560.83 | $2,072.55 | $2,584.28 |
| Base + Phone, including $11.15 number | $439.45 | $818.25 | $1,197.05 | $1,575.85 | $1,954.65 |
| All three add-ons | $951.17 | $1,841.70 | $2,732.22 | $3,622.75 | $4,513.27 |
Base allocation shown at the Phase 1 rate of $14.50. Phase 2 uses $9.20 and Phase 3 uses $8.30 per salon.
Lever 1: average phone handle time per lead
| Minutes / lead | 100 leads | 200 | 300 | 400 | 500 |
|---|---|---|---|---|---|
| 10 min | $189.40 | $378.80 | $568.20 | $757.60 | $947.00 |
| 15 min | $284.10 | $568.20 | $852.30 | $1,136.40 | $1,420.50 |
| 20 min, current assumption | $378.80 | $757.60 | $1,136.40 | $1,515.20 | $1,894.00 |
| 25 min | $473.50 | $947.00 | $1,420.50 | $1,894.00 | $2,367.50 |
Lever 2: token budget and cache hit rate
| Scenario | Per run | Text cost / lead | 100 leads | 300 | 500 |
|---|---|---|---|---|---|
| 400k tokens, 75% cached, current | $0.775 | $5.038 | $503.75 | $1,511.25 | $2,518.75 |
| 400k tokens, 90% cached | $0.480 | $3.120 | $312.00 | $936.00 | $1,560.00 |
| 200k tokens, 75% cached | $0.425 | $2.763 | $276.25 | $828.75 | $1,381.25 |
| 100k tokens, 75% cached | $0.250 | $1.625 | $162.50 | $487.50 | $812.50 |
| 400k tokens, no caching | $2.075 | $13.488 | $1,348.75 | $4,046.25 | $6,743.75 |
Price floors at target gross margin
| Planning case | 100 leads | 200 | 300 | 400 | 500 |
|---|---|---|---|---|---|
| AVG cost to serve | $283 | $506 | $729 | $951 | $1,174 |
| 50% margin floor | $567 | $1,012 | $1,457 | $1,902 | $2,348 |
| 60% margin floor | $708 | $1,265 | $1,821 | $2,378 | $2,935 |
| 70% margin floor | $944 | $1,686 | $2,428 | $3,171 | $3,913 |
| Maximum utilisation exposure | |||||
| MAX cost to serve | $951 | $1,842 | $2,732 | $3,623 | $4,513 |
| 50% margin floor | $1,902 | $3,683 | $5,464 | $7,246 | $9,027 |
| 60% margin floor | $2,378 | $4,604 | $6,831 | $9,057 | $11,283 |
| 70% margin floor | $3,171 | $6,139 | $9,107 | $12,076 | $15,044 |
These are model-derived floors, not client pricing recommendations. They exclude support, sales, and overhead.
Items to validate during beta
This is estimated, not measured. Instrument average handle time per lead in Phase 1 beta and refit before plan pricing.
At 500 leads, 10,000 agent minutes per month are modelled. Exceeding plan concurrency changes the rate from $0.08 to $0.16 per minute.
Voice is modelled at $0.10 per minute using light turns. If voice uses the same agentic tool loop as text, the phone add-on may be materially under-modelled.
Instrument token counts per run in beta. Token budget and cache discipline are the largest cost levers in the model.
At 50 salons, the licence line alone is $1,750 per month. Volume or multi-tenant terms should be explored before Phase 2.