Appendix C: Budget Planning Templates
This appendix provides the budget templates referenced in Chapter 5. Each template is designed for a specific phase and scale of AI transformation investment. Use them as starting points calibrated to your organization’s specific situation, not as precise forecasts.
All figures are approximate and reflect typical costs for B2B SaaS companies in the 30 to 250 employee range as of 2025. Actual costs will vary significantly with team composition, model selection, usage volume, and organizational context.
Template 1: Sprint Phase Budget (90 Days)
This template covers the first 90-day implementation sprint described in Part IV. It assumes an internal team executing the sprint without external consultants.
Personnel Costs
The personnel cost for the sprint is the most significant and most commonly underestimated line item. The cost is not a new hire; it is the opportunity cost of existing engineering and product time redirected to AI work.
| Role | Time Allocation | Duration | Monthly Loaded Cost | Sprint Cost |
|---|---|---|---|---|
| AI lead engineer (internal) | 80% | 3 months | $14,000 | $33,600 |
| Supporting engineer (internal) | 40% | 3 months | $12,000 | $14,400 |
| Product manager (internal) | 30% | 3 months | $11,000 | $9,900 |
| Engineering manager (oversight) | 10% | 3 months | $16,000 | $4,800 |
| CS/domain expert (calibration support) | 20% | 2 months | $9,000 | $3,600 |
| Personnel subtotal | $66,300 |
Loaded cost includes salary, benefits, and employer taxes. Adjust multiplier for your organization.
Infrastructure and API Costs
| Item | Monthly Cost | Sprint Total |
|---|---|---|
| LLM API (generation) | $400 to $1,200 | $1,200 to $3,600 |
| Embedding API | $50 to $150 | $150 to $450 |
| Vector database (managed) or pgvector infrastructure | $75 to $300 | $225 to $900 |
| Additional cloud infrastructure (storage, compute) | $100 to $400 | $300 to $1,200 |
| Infrastructure subtotal | $625 to $2,050 | $1,875 to $6,150 |
Infrastructure costs scale significantly with production traffic volume. Sprint estimates assume development and limited beta usage, not full production load.
Tooling and Software
| Item | Cost |
|---|---|
| Evaluation and observability tooling (if not using custom scripts) | $0 to $500/month |
| Development tools and licenses specific to AI work | $0 to $200/month |
| Data preparation tools (if needed) | $0 to $500 one-time |
| Tooling subtotal (3 months) | $0 to $3,600 |
External Support (Optional)
| Item | Estimated Cost |
|---|---|
| Technical review or architecture consultation (10 to 20 hours) | $3,000 to $8,000 |
| Prompt engineering specialist (if internal skills are limited) | $5,000 to $15,000 |
| Legal review of AI use and data handling (if needed) | $2,000 to $5,000 |
| External support subtotal | $0 to $28,000 |
Sprint Phase Total
| Category | Low Estimate | High Estimate |
|---|---|---|
| Personnel | $60,000 | $80,000 |
| Infrastructure | $2,000 | $6,000 |
| Tooling | $0 | $4,000 |
| External support | $0 | $28,000 |
| Contingency (15%) | $9,300 | $17,700 |
| Total | $71,300 | $135,700 |
Planning guidance: Most B2B SaaS companies in the 50 to 200 employee range should budget $75,000 to $110,000 for a single well-scoped sprint. Budgets below $60,000 are structurally insufficient for a sprint that produces a production-quality system with adequate evaluation infrastructure. Budgets above $150,000 for a single sprint typically indicate either scope that is too broad or external consulting overhead that should be questioned.
Template 2: Mid-Term Phase Budget (Months 4 to 12)
The mid-term budget covers the operating cost of the first production system and the investment in building the second use case. It is structured as an annual budget with monthly breakdowns for the operating components.
Ongoing Personnel Allocation
| Role | Ongoing Allocation | Monthly Loaded Cost | Annual Cost |
|---|---|---|---|
| AI lead engineer (operating + development) | 60% | $14,000 | $100,800 |
| Supporting engineer (rotating, per active project) | 30% | $12,000 | $43,200 |
| Product manager (AI portfolio) | 20% | $11,000 | $26,400 |
| Data analyst or data engineer (if hired) | 100% | $10,000 | $120,000 |
| CS calibration support (ongoing, decreasing over time) | 10% | $9,000 | $10,800 |
| Personnel subtotal (mid-term, months 4-12) | $301,200 |
Note: The data analyst or data engineer is the highest-value mid-term hire and significantly affects the budget. If this role is filled by an existing team member in a partial allocation, the cost is lower but the capacity constraint is real.
Infrastructure Cost Growth
Production AI systems serving real users at scale have higher infrastructure costs than sprint systems serving beta users. Model the cost growth against your usage growth trajectory.
| Component | Month 4 | Month 8 | Month 12 |
|---|---|---|---|
| LLM API (production volume) | $800 | $1,400 | $2,200 |
| Embedding and retrieval infrastructure | $200 | $350 | $500 |
| Monitoring and observability | $100 | $200 | $300 |
| Additional infrastructure | $200 | $400 | $600 |
| Monthly infrastructure | $1,300 | $2,350 | $3,600 |
Annual infrastructure estimate (months 4 to 12): $22,000 to $32,000 depending on usage growth rate.
Second Use Case Development (within mid-term)
Budget for the second sprint within the mid-term phase. This sprint is typically more efficient than the first because infrastructure and methodology are already established, but it still requires protected time.
| Item | Estimate |
|---|---|
| Engineering time (incremental, 6 to 8 weeks focused effort) | $15,000 to $25,000 |
| Infrastructure additions specific to second use case | $1,000 to $5,000 |
| Evaluation framework extension | $2,000 to $5,000 |
| Second use case development subtotal | $18,000 to $35,000 |
Mid-Term Phase Annual Total
| Category | Annual Estimate |
|---|---|
| Personnel (months 4-12) | $180,000 to $320,000 |
| Infrastructure | $22,000 to $32,000 |
| Second use case development | $18,000 to $35,000 |
| Governance and program management | $5,000 to $15,000 |
| Contingency (10%) | $22,500 to $40,200 |
| Annual mid-term total | $247,500 to $442,200 |
Planning guidance: The wide range in the mid-term budget reflects the significant variable of whether a dedicated data engineer is hired in this phase. The $180,000 to $220,000 personnel range assumes no new hire; the $280,000 to $320,000 range assumes a dedicated data engineer is added in months 6 to 8. The data engineer hire is almost always the right investment once the first system is in production and the team is spending meaningful time on manual evaluation and data processing tasks.
Template 3: Year-Two AI Function Budget
The year-two budget represents the annualized cost of a formal, scaled AI function operating a portfolio of three to five AI systems, with dedicated ownership of AI systems engineering, data and evaluation, and AI product management.
Personnel: Formal AI Function
| Role | FTE | Monthly Loaded Cost | Annual Cost |
|---|---|---|---|
| Head of AI Systems (senior engineer) | 1.0 | $18,000 | $216,000 |
| AI systems engineer | 2.0 | $14,000 | $336,000 |
| Data and evaluation engineer | 1.0 | $12,000 | $144,000 |
| AI product manager (dedicated or 50% allocation) | 0.5 | $13,000 | $78,000 |
| AI function personnel subtotal | 4.5 FTE | $774,000 |
This represents a fully built-out AI function. Many organizations will phase this over months 13 to 24 rather than hiring all roles at the start of year two.
Infrastructure at Scale (Three to Five AI Systems in Production)
| Component | Monthly | Annual |
|---|---|---|
| LLM API (production volume, multiple systems) | $3,500 to $8,000 | $42,000 to $96,000 |
| Vector database and embedding infrastructure | $600 to $1,500 | $7,200 to $18,000 |
| Evaluation and observability platform | $500 to $2,000 | $6,000 to $24,000 |
| Data pipeline infrastructure | $300 to $800 | $3,600 to $9,600 |
| Additional cloud compute | $500 to $1,500 | $6,000 to $18,000 |
| Infrastructure subtotal | $5,400 to $13,800 | $64,800 to $165,600 |
Year-Two Program Costs
| Item | Annual Estimate |
|---|---|
| New use case development (2 to 3 sprints) | $40,000 to $80,000 |
| External benchmarking and program assessment | $10,000 to $25,000 |
| Training and skill development | $15,000 to $30,000 |
| Customer adoption support programs | $10,000 to $20,000 |
| Governance program management | $8,000 to $15,000 |
| Program costs subtotal | $83,000 to $170,000 |
Year-Two Total
| Category | Annual Estimate |
|---|---|
| AI function personnel | $500,000 to $800,000 |
| Infrastructure | $65,000 to $166,000 |
| Program costs | $83,000 to $170,000 |
| Contingency (10%) | $64,800 to $113,600 |
| Year-two total | $712,800 to $1,249,600 |
Planning guidance: The year-two range is wide because it encompasses organizations at different stages of AI function maturity. Organizations that built strong foundations in year one and have a clear use case roadmap are toward the lower end of the range. Organizations that are catching up on foundation work while also trying to scale are toward the higher end. The $700K to $800K range represents a well-scaled AI function for a B2B SaaS company in the $10M to $20M ARR range; this is 7 to 8% of ARR, which is at the high end of what most boards will approve and requires a strong ROI narrative to defend.
Budget ROI Framework
The budget templates above are only defensible if the program produces measurable business outcomes that justify the investment. Use this framework to construct the ROI narrative the program needs to secure and maintain funding.
Primary ROI Metrics by Use Case Type
Customer Success AI systems (response generation, knowledge retrieval): - Reduction in average response time: measure in hours, translate to CS capacity freed - Increase in CS-to-customer ratio: additional accounts per rep at constant cost - Reduction in churn attributable to slow support: translate using average contract value and average churn reason distribution
Onboarding AI systems: - Reduction in time to first value: measure in days, translate to CS hour savings and correlation with early retention - Reduction in manual onboarding steps: translate to CS capacity freed
Product AI features (customer-facing): - Feature adoption rate among customers who have access - Deal win rate change when AI features are included in demos - NPS correlation with AI feature usage depth
Constructing the Business Case
For each AI system in the portfolio, define:
- Baseline metric: The specific measurement before the AI system was deployed (e.g., average response time: 2.8 days; CS ratio: 23 accounts per rep).
- Target metric: The specific measurement the AI system is designed to produce (e.g., average response time: 1.4 days; CS ratio: 31 accounts per rep).
- Achieved metric: The actual measurement after the system has been in production for at least ninety days.
- Business value translation: The dollar value of the metric improvement, using a consistent methodology that the finance team has reviewed. For CS capacity improvements, use the loaded cost per CS hour. For churn improvements, use the average contract value and the churn rate delta.
- Attribution methodology: The comparison that supports attributing the metric improvement to the AI system rather than to other factors (comparable period without the system, cohort comparison, or regression analysis if the analytical infrastructure supports it).
A portfolio of three AI systems with documented, credible business value translations is the most effective board-level narrative for continued AI investment. The precise dollar amounts matter less than the methodology: boards that trust the measurement trust the investment decision.