| name | saas-ai-cost-of-tenant-calculator |
| description | Use when producing or reviewing the saas ai cost of tenant calculator component of a business plan; applies its specialist evidence, decisions, and acceptance tests instead of neighbouring pipeline skills. |
| metadata | {"portable":true,"compatible_with":["claude-code","codex"]} |
SaaS AI Cost-of-Tenant Calculator Skill
Overview
A tier design that ignores per-tenant AI cost will produce a margin disaster. This skill installs the calculator: a tenant-level computation of monthly AI cost from first principles, with sensitivity, scenarios, and a worked African vertical-SaaS example. It is the operating tool behind saas-ai-unit-economics-and-cogs and the input to saas-ai-pricing-strategy.
The calculator answers four questions an AI-aware investor will ask within the first 30 minutes:
- What does the median tenant cost you in AI per month?
- What does the top-decile tenant cost you?
- Where does the model break — what tier × usage profile turns negative-margin?
- What levers protect margin without degrading the customer's experience?
Use When
- A SaaS plan has customer-facing AI features (chat, summarisation, classification, generation, RAG, agentic flows)
- Tier design or packaging decisions are being made
- A tier's margin is suspicious and needs verification
- Pricing changes are being modelled
- An investor asks "what is your AI cost per tenant?" — this skill produces the answer in a defensible format
- Plan must pass
meta-ai-bankability-and-investor-readiness
Do Not Use When
- AI is internal-efficiency only (no per-tenant cost) — use
14-ai-integration cost-benefit framework instead
- Product is in pre-design stage with no architecture decisions — return after architecture is set
- Tenant cost is dominated by non-AI infra (hosting, storage, payment fees) — use
saas-unit-economics-and-cohort-model first; this is a supplement
Required Inputs
- Tier list with ARPU and AI-feature inclusion per tier
- Architecture: foundation model(s) used, embedding model, vector store, eval pipeline, model-router policy
- Provider pricing snapshot (per-1k input tokens, per-1k output tokens, per embedding, per fine-tune token, per GPU-hour, per vector-store row)
- Per-tenant usage assumptions (queries/month, prompt tokens/query, completion tokens/query, embeddings/month, documents indexed)
- Cache-hit ratio (semantic + exact-match)
- Fine-tune amortisation policy (e.g. $5,000 fine-tune / 1,000 active tenants / 12 months)
- Eval pipeline cost / month and active-tenant count for allocation
- Reserve policies (hallucination, retraining)
- Currency exposure (USD AI cost vs local-currency ARPU)
Workflow
- Capture the architecture stack — one paragraph + component list (model, embedding, vector DB, router, eval, observability, cache).
- Build the per-query cost formula per
references/saas-ai-cost-of-tenant-calculator.md:
per-query cost = (prompt tokens × input rate + completion tokens × output rate) × (1 − cache-hit) × model-mix-weighted-rate
- Build the per-tenant monthly cost formula:
per-tenant monthly AI cost =
queries × per-query cost (after cache + model mix)
+ embeddings × embedding rate
+ fine-tune amortisation share
+ eval pipeline allocation share
+ GPU reservation share (if dedicated)
+ retraining contribution share
+ hallucination reserve allocation
+ AI overhead allocation
- Build the per-tier roll-up — apply the per-tenant formula for each tier's usage profile. Output: tier × cost-per-tenant × cost-as-%-of-ARPU.
- Build the sensitivity matrix — three axes minimum:
- usage (×0.5, ×1, ×2, ×4)
- model-mix downshift (cheaper-model share 25% / 50% / 75%)
- cache-hit ratio (20% / 40% / 60% / 80%)
- and a fourth: FX (UGX or NGN ±20%)
- Identify the break-even tenant — at what usage × pricing does the tier go negative-contribution? This is the design-of-the-tier constraint.
- Identify the top-decile tenant cost — the long-tail tenant is the margin killer; usage caps or fair-use policy are tier-design outputs.
- Apply mitigation levers in scenarios — list 6-8 levers with expected impact (see Mitigation Lever Library below).
- Produce the worked example — at least one realistic African vertical SaaS worked through end-to-end.
- Wire to pricing decision — outputs feed
saas-ai-pricing-strategy.
- Wire to living plan — cost-per-tenant becomes a monthly KPI; provider pricing watch monthly; cache-hit ratio weekly.
Mitigation Lever Library
| Lever | Typical impact | Cost / risk |
|---|
| Semantic + exact-match cache | -30 to -60% | Cache infra + freshness risk |
| Prompt-token compression (system-prompt minimisation, schema prompting) | -10 to -25% | Engineering time |
| Completion-token truncation / structured output | -15 to -30% | Quality risk if poorly designed |
| Model-mix routing (cheap model for routine; premium for complex) | -30 to -50% | Router + eval to detect drift |
| Per-tier usage caps with overage | -20 to -40% on cost variance | Customer-experience friction |
| Local model fallback (Llama 3, Mistral, Gemma) | -50 to -90% on routed share | Hosting infra + quality risk |
| Distillation (fine-tune a small model on a big model's outputs) | -60 to -85% on routed share | Eval rigour mandatory |
| Batch processing for non-real-time | -30 to -50% on batched share | Latency change |
| RAG over fine-tuning when data freshness matters | -50 to -80% vs fine-tuning | RAG infra complexity |
| AI-as-add-on tier (separate tier, separate price) | Margin-protective | Sales-cycle complexity |
Quality Bar
- Per-tenant cost has explicit formula, not a vibe
- Prompt tokens vs completion tokens separately treated
- Cache-hit ratio explicit and defensible
- Model-mix policy stated
- Embeddings line-itemed when RAG is used
- Eval cost allocated
- Reserves modelled when product is high-stakes
- Sensitivity covers the four core axes
- Break-even tenant identified
- Top-decile tenant identified
- Mitigation levers prioritised with expected impact
- Worked example shown end-to-end
- FX modelled if cost-currency ≠ revenue-currency
- Living-plan cadence assigned
Anti-Patterns
- "We estimate AI cost at 5%" with no calculation
- One token rate (no input/output split)
- 20% cache-hit assumed without instrumentation
- Single-model assumption when product needs routing
- Embeddings ignored when RAG is described in product section
- No long-tail / power-user accounting
- "We'll switch to cheaper models" without quality evaluation
- Mitigation levers stated as line items without quantified impact
- Reserves omitted in regulated verticals (health, finance, legal)
- USD-only modelling when ARPU is local currency
Outputs
- Architecture stack description
- Per-query and per-tenant cost formulas with assumptions
- Tier roll-up table (tier × ARPU × AI cost × % of ARPU × contribution)
- Sensitivity matrix on usage / model-mix / cache / FX
- Break-even tenant and top-decile tenant analysis
- Mitigation lever playbook with prioritisation
- Worked African vertical-SaaS example
- Hand-off to
saas-ai-pricing-strategy and saas-ai-unit-economics-and-cogs
Living-Plan Cadence Defaults
| Element | Cadence | Owner | Variance threshold |
|---|
| Per-tenant AI cost (median) | monthly | CFO + Head of AI | +20% MoM |
| Per-tenant AI cost (top decile) | monthly | Head of AI | >2.5× median |
| Cache-hit ratio | weekly | Head of AI | -10pp from baseline |
| Token usage per query | weekly | Head of AI | +25% week-over-week |
| Model-mix share | monthly | Head of AI | shift >15pp |
| Provider pricing | as-published + monthly | Head of AI / CTO | any provider change |
| FX corridor | monthly | CFO | move >5% from plan |
References
references/saas-ai-cost-of-tenant-calculator.md — full formula spec, worksheet structure, sensitivity matrix, worked examplereferences/saas-ai-pricing-architecture.md — how cost links to pricingskills/10-financial-projections/saas-ai-unit-economics-and-cogs/SKILL.md — sister skillskills/14-ai-integration/references/saas-ai-feature-roadmap-in-business-plan.md — roadmap disciplinebook-extractions/tod-building-multi-tenant-saas-architectures-extraction.md — multi-tenant cost realities
Africa / Uganda Application Notes
- FX overlay is mandatory; tier ARPU is local currency, AI cost is USD-denominated.
- Cache-hit ratios are typically higher in vertical-SaaS African contexts (40-60% vs US 20-30%) because query patterns repeat within tight vertical contexts.
- Local-language tokens are typically 1.5-2.5× more tokens per equivalent English content because African languages are under-represented in tokenisers; line-item this if local-language inference is core.
- If hosting in af-south-1, africa-south1, or local providers (Liquid, Cassava, Raxio, MainOne), GPU pricing is typically 1.5-3× US/EU.
- Public-sector / NGO tenants are typically high-touch / high-doc-volume → top-decile AI cost; separately analyse.
- Mobile-first customers using WhatsApp interfaces produce different token profiles (shorter prompts, more turns) than browser-using customers; model both.
July 2026 Portable Contract
Required Inputs
| Input artefact | Source/provider | Required | Behaviour when absent |
|---|
| Approved commercial assumptions, contracts, usage/cost evidence, accounting framework, opening position, and projection horizon for saas ai cost of tenant calculator | Client records, approved operating model, finance owner, and accounting doctrine | Yes | If absent, contract terms, usage evidence, framework, or cost drivers are unavailable, isolate the affected schedule, label it unassessed, and do not force the model to balance with a plug. |
| Finalised business brief, target reader, country, and stage | Client intake and engagement owner | Yes | Stop section decisions and route the missing context to client intake. |
| Reconciled upstream assumptions that this section consumes | Named pipeline owners | Conditional | Record the dependency, affected claim, owner, and recovery step; do not substitute an invented value. |
Outputs
| Artefact | Consumer | Observable acceptance condition |
|---|
| Unit-economics model with task or tenant cost bridge | Plan author and target decision-maker | The artefact answers the section decision and traces each material conclusion to the supplied evidence. |
| saas ai cost of tenant calculator exception and handoff note | Downstream section owners | Every blocked or conditional item names its consequence, owner, evidence request, and restart condition. |
| saas ai cost of tenant calculator release record | Reviewer or plan assembler | Records the checks completed, failures, unassessed items, professional review required, and release state. |
Evidence Produced
| Evidence | Format | Acceptance condition |
|---|
| Formula trace, source/assumption register, three-statement or schedule reconciliation, and finance-gate record | Source-linked table, calculation, or annotated prose | The evidence is reproducible from named inputs and distinguishes verified fact, management assumption, and inference. |
| saas ai cost of tenant calculator decision record | Decision note | States the selected action, rejected credible alternative, countercase, rationale, and risk accepted or avoided. |
| saas ai cost of tenant calculator review trace | Gate entry | Identifies the date, input versions, reviewer role, failed checks, recovery owner, and any check that remains not assessed. |
Capability and Permission Boundaries
For saas ai cost of tenant calculator, the controlling focus is tenant-level token, model, vector, tool, storage, and support cost allocation. This skill may inspect records and calculate planning scenarios in read-only mode; it may not post entries, change ledgers, set accounting policy, certify IFRS treatment, or release statutory values without authorised professional review. Its normal mode is read-only analysis and drafting. Any mutation, external communication, spending, certification, or professional conclusion outside that boundary requires explicit authority and must remain traceable to the approving role.
Degraded Mode
For saas ai cost of tenant calculator, loss of evidence about tenant-level token, model, vector, tool, storage, and support cost allocation activates degraded mode. If the controlling saas ai cost of tenant calculator evidence is unavailable, the same boundary applies. When contract terms, usage evidence, framework, or cost drivers are unavailable, isolate the affected schedule, label it unassessed, and do not force the model to balance with a plug. Return the verified subset, label the affected decision qualified or not assessed, explain the downstream consequence, and state the smallest evidence request or authorised action that permits recovery. Do not convert the missing check into a pass.
Decision Rules
| Choice or condition | Action | Failure or risk avoided |
|---|
| For saas ai cost of tenant calculator, commercial billing, cash receipt, service delivery, and accounting recognition occur in different periods | model each event separately, reconcile the bridge, and route judgemental treatment to the finance reviewer | Cash, revenue, liability, and margin can be conflated into a misleading forecast |
| For saas ai cost of tenant calculator, A current legal, regulatory, tax, accounting, market, or platform claim controls the saas ai cost of tenant calculator decision | Verify the controlling source, effective date, jurisdiction, and reviewer status before release | Stale external facts become permanent plan assumptions |
| For saas ai cost of tenant calculator, The evidence reconciles with neighbouring sections and the countercase does not overturn the choice | Complete unit-economics model with task or tenant cost bridge, attach the evidence and release record, and hand off named dependencies | Premature release and repeated downstream rework |
Workflow
- Define the exact saas ai cost of tenant calculator decision, intended reader, jurisdiction, business stage, and permission boundary.
- Collect approved commercial assumptions, contracts, usage/cost evidence, accounting framework, opening position, and projection horizon and map each material conclusion to its source; stop the affected conclusion when an input could change it.
- Apply the specialist methods and directly linked references already contained in this skill, retaining its domain thresholds, calculations, and Uganda or East Africa context where applicable.
- Compare the credible alternatives, test the countercase and failure path, and apply the decision table rather than selecting a template default.
- Produce unit-economics model with task or tenant cost bridge with the evidence, exception, and handoff records; reconcile every shared assumption with its owning section.
- Run the section quality checks, applicable finance or professional review, and anti-slop gate. If a gate fails, correct the evidence or decision and return to the responsible step.
Quality Standards
- Unit-economics model with task or tenant cost bridge must answer a real decision for the named bank, investor, DFI, grant, board, or strategic-partner reader.
- Formula trace, source/assumption register, three-statement or schedule reconciliation, and finance-gate record must be source-linked, dated where facts can change, and sufficient for another reviewer to reproduce the conclusion.
- The section exposes its countercase, stop condition, recovery action, and effect on neighbouring sections.
- No unavailable source, calculation, tool, or professional review is reported as passed; finance and statutory judgements follow the governing doctrine.
- Language remains specific to saas ai cost of tenant calculator, uses British English naturally, and passes the repository anti-slop gate without promotional filler.
Anti-Patterns
- In saas ai cost of tenant calculator, treating an unavailable approved commercial assumptions, contracts, usage/cost evidence, accounting framework, opening position, and projection horizon as confirmed. Correction: qualify the affected conclusion and issue the named evidence request.
- Producing unit-economics model with task or tenant cost bridge that restates the brief but makes no choice. Correction: record the choice, rejected alternative, rationale, countercase, and implication.
- Ignoring a conflicting upstream assumption. Correction: return it to its owning section and resume only from a reconciled version.
- Reporting an unavailable check as passed. Correction: mark it not assessed and narrow the release state.
- Claiming compliance, assurance, bankability, or investor readiness from narrative quality. Correction: run the applicable gate and retain its evidence.
- Copying the worked example into a client plan. Correction: use the method only and replace every fact with verified engagement evidence.
Worked Example
One tenant uses long documents, premium models, and high vector retrieval while another uses short prompts. Allocate observed drivers per tenant instead of spreading cloud invoices evenly.
References
- Use the verified project evidence register and the owning upstream pipeline section for saas ai cost of tenant calculator; no local deep-dive reference is declared.
- For saas ai cost of tenant calculator claims involving money, tax, grants, reserves, revenue, cost, valuation, or financial statements, apply the Chwezi finance doctrine and record the required professional-review state; illustrative figures never become client facts.
