CtrlSpend

Table of Contents

1. Mission & Vision
2. What We Do (and Don't Do)
3. The Problem We Solve
4. Market Opportunity (TAM / SAM / SOM)
5. Product: MVP Overview
6. Core User Flows
7. Ideal Customer Profile (ICP)
8. Competitive Landscape
9. Go-to-Market Strategy
10. Finding Your First Customer
11. Next Steps: From MVP to Production

1. Mission & Vision

Mission Statement

CtrlSpend gives engineering and finance teams a single pane of glass to understand, forecast, and govern every dollar of AI and cloud usage spend — before it becomes a surprise on the invoice.

We believe that as AI becomes operational infrastructure, usage-based pricing (tokens, API calls, Flex Credits, compute units) will replace seat-based SaaS as the dominant cost structure for technology companies. CtrlSpend is the financial operating system for that world.

Vision

To be the default spend intelligence layer for any company running AI at scale — the way Datadog is for observability and Rippling is for HR.

What CtrlSpend Is Not

Understanding scope is as important as the mission itself. CtrlSpend explicitly does not:

• Replace your cloud billing dashboard. We read from AWS Cost Explorer, Azure Cost Management, and GCP Billing — we don't replicate them.
• Manage your procurement contracts. We don't negotiate, sign, or store vendor agreements (though we store your negotiated rates for cost derivation).
• Compete with FinOps tools like Apptio or CloudHealth. Those tools optimize Reserved Instance coverage and EC2 rightsizing. We focus on the AI/ML usage layer: tokens, agentic credits, model calls.
• Replace developer observability tools. Langfuse, Helicone, and Weights & Biases are great for debugging prompts. CtrlSpend is for finance and VP Engineering audiences asking "what are we spending and is it on track?"
• Build the underlying AI infrastructure. We are a management and visibility layer, not an inference provider.

2. What We Do (and Don't Do)

3. The Problem We Solve

The AI Spend Explosion

Enterprise AI API spend reached $8.4 billion in mid-2025, up from $3.5B just six months prior — a 140% increase in under a year. The average company now spends $85,521/month on AI services, growing at 36% year-over-year.

This growth is outpacing every financial control mechanism companies have:

• 85% of companies misestimate their AI costs by more than 10%
• 25% are off by more than 50%
• 80% miss their AI spend forecasts by more than 25%
• 40% of companies now spend over $10M/year on AI

The Three-Layer Problem

The Salesforce Agentforce Timing Opportunity

Salesforce Agentforce is the fastest-growing source of unexpected AI spend in the enterprise:

• 9,500 paid Agentforce customers as of early 2026
• $540M ARR with 330% year-over-year growth
• Only 8% of Salesforce's 150,000+ customer base has adopted Agentforce — meaning the install base will grow 10x+ in the next 2–3 years
• Flex Credits are opaque — companies routinely run out mid-month with no warning
• No existing tool (not CloudZero, not Vantage, not Apptio) provides Agentforce-level attribution with dollar derivation

4. Market Opportunity (TAM / SAM / SOM)

Total Addressable Market

Combined TAM for AI usage spend management: ~$4–6B by 2028, growing to $12–15B by 2032 as AI spend becomes the dominant enterprise IT cost category.

Serviceable Addressable Market (SAM)

Focusing on: US and EU companies with >$50K/year AI spend, using 2+ AI vendors, with a dedicated engineering or FinOps function.

• Estimated 180,000 companies globally meet this criteria today
• Average ACV target: $24,000/year
• SAM: ~$4.3B

Serviceable Obtainable Market (SOM) — Year 1–3

Year 1 is not about revenue — it is about proving the ICP and building the integration depth that makes CtrlSpend a must-have.

5. Product: MVP Overview

The current MVP is a fully functional browser-side demo built in Next.js 16 + Tailwind CSS v4, demonstrating all core product surfaces. All data is currently mock/static — the UI/UX is complete and ready for real integration development.

The 7 Pages

Key Technical Decisions

6. Core User Flows

Sign up (email / Google SSO)
→ Onboarding wizard: "What AI tools does your team use?"
→ Connect first integration (recommended: OpenAI or Salesforce)
    → OAuth authorize
    → Set contract rate (if applicable)
    → First sync runs (15–30 seconds)
→ Dashboard populates with real data
→ "You have 3 unattributed resources — review them?" → Settings > Needs Review
→ Assign owners to top 5 cost drivers
→ Set first budget threshold alert
→ Activation complete ✓

Land on Dashboard
→ See MTD spend vs. last month + forecast
→ Check Events feed for new anomalies (badge count in nav)
    → If anomaly: click Investigate → review root cause → Notify owner or Resolve
→ Review any pending Needs Review items in Settings
→ Done (typical: 3–5 minutes)

Finance partner opens Vendors page
→ Switch to "30d" view
→ Export (planned) or screenshot for budget reconciliation
→ Compare Forecasted vs. Budget column for each vendor
→ Identify over-budget vendors → click through to Integration Detail
→ Pull audit trail + agent breakdown for cost explanation
→ Prepare variance memo for leadership

Integrations page → "Add Integration"
→ Browse catalog or search by name
→ Select vendor
    → If coming soon: click "Request early access" (email capture)
    → If available: begin OAuth flow
        → Authorize with vendor
        → Confirm org/account details
        → Set contract rate (for rate × units vendors)
        → "Finish setup" → redirect to integration detail page
→ Integration appears in active cards with real data

Slack alert: "🚨 OpenAI spend up 340% vs. yesterday — $12,400 projected impact"
→ Click link → Events detail page
→ See: spike at 2pm, model gpt-4o-mini, team: backend-infra
→ CtrlSpend Insight: "Correlates with deployment 'search-v2' pushed at 1:48pm"
→ Click "Notify" → Slack DM sent to owner
→ Owner replies: "Known — we're load testing. Ignore for 48h."
→ Update status to "Ignored" with note
→ Auto-re-open if spend doesn't normalize within 48h

7. Ideal Customer Profile (ICP)

Primary ICP: Series B–D SaaS or AI-native company

Primary Buyer Personas

Secondary ICP: Enterprise Teams at Large Companies

Large enterprises (Fortune 1000) with dedicated FinOps teams who have adopted Salesforce Agentforce and are hitting the limits of what Salesforce's own reporting provides.

• Decision maker: VP of IT Finance or Director of FinOps
• Pain: Agentforce Flex Credits are a black box; they need credit-level attribution by department
• Deal size: $50K–$200K ACV; longer sales cycle but much higher LTV

Anti-ICP (Who NOT to Target)

• Early-stage startups (<$10K/month AI spend) — the problem isn't big enough yet
• Companies with only one AI vendor and <$50K/month — billing console is sufficient
• Companies using only open-source models (Llama, Mistral self-hosted) — no billing API to read from
• Traditional enterprise with no AI vendors yet — they're not in pain yet

8. Competitive Landscape

Direct Competitors

CtrlSpend's 5 Defensible Differentiators

1. Cross-vendor unification — the only tool that shows LLM tokens + Salesforce Flex Credits + cloud AI in one budget view
2. Derivation transparency — we show the formula, not just the number (8,125 credits × $0.08 = $650)
3. Agent-level attribution — for Salesforce Agentforce, costs attributed down to individual agents
4. CtrlSpend Insight layer — we separate raw vendor data from our own correlations and label them distinctly
5. Speed to value — connect first integration and see real attributed spend in under 5 minutes

9. Go-to-Market Strategy

Phase 1: Design Partner Program (Now → Month 6)

Goal: 10–25 committed design partners who help shape the product in exchange for early access and a significant discount (50–80% off future pricing).

Who to recruit: Series B/C SaaS companies using OpenAI + Salesforce, with a VP Engineering or Head of FinOps as champion.

Commitment ask from partners:

• 30-minute weekly check-in for the first 8 weeks
• Introduce to 2 other potential customers if they find value
• Written testimonial or case study by month 6

Phase 2: Product-Led Growth Foundation (Month 3–9)

Goal: Self-serve onboarding funnel so users can activate without a sales call.

Key PLG motions:

• Free tier: up to 2 integrations, 90-day history, 1 seat (enough to prove value)
• Upgrade triggers: 3rd integration, 6-month history, team seats, Slack alerting
• Activation email sequence: "You connected OpenAI — here are the 3 things to check first"
• Weekly spend digest email (keeps users coming back even without an emergency)

Phase 3: Outbound Sales (Month 6–18)

Target accounts: 100–500 employees, B2B SaaS, using Salesforce CRM + 2+ AI APIs.

Outbound signals to watch for:

• Job postings: "FinOps engineer," "AI cost optimization," "ML platform engineer"
• LinkedIn activity: VP Engineering posting about AI infrastructure costs
• Conference attendance: FinOps X, Dreamforce, AI Engineer World's Fair
• Tech stack signals (G2 or BuiltWith): Salesforce + OpenAI in their stack

Example outbound message:

Hi [Name], saw your post about managing AI costs at [Company]. We're building CtrlSpend — it's the first tool that shows Salesforce Agentforce Flex Credits and OpenAI spend in the same budget dashboard, with agent-level attribution. We're working with a handful of design partners right now and would love to show you a 15-minute demo. Worth a look?

Pricing Model

10. Finding Your First Customer: A Practical Playbook

Week 1: Research & List Building

Build a spreadsheet of 50 target companies using these criteria:

• Series B or C (check Crunchbase)
• B2B SaaS product, 50–300 employees
• Salesforce CRM in their tech stack (check BuiltWith or G2)
• At least one AI API in use (GPT-4, Claude, etc.)
• Have a "VP Engineering" or "CTO" on LinkedIn

For each company, find: VP Engineering (primary), Head of Finance (secondary), Platform/FinOps Engineer (champion).

Week 2: Warm Introduction Outreach

• Your own network: LinkedIn connections at matching companies — ask for a "15-minute conversation about how your company manages AI spend" (research, not a sales call)
• Investor network: Ask for 3 introductions each to portfolio companies matching the ICP
• Community channels: Heavybit, SaaStr Insider, Rands Leadership Slack, FinOps Foundation Slack, Salesforce Trailblazer Community

Weeks 3–4: Discovery Call Questions

Goal of the first call: learn, not sell.

1. "How do you currently track your AI API spend across vendors?"
2. "What's your monthly AI spend — ballpark?"
3. "Have you ever had a surprise on an AI invoice?"
4. "How do you attribute AI costs to different teams or products?"
5. "What does your month-end process look like for AI cost review?"
6. "If you had a tool that showed all of this in one place, who would use it — you, your finance team, both?"

High-signal responses to listen for: specific invoice surprise stories, Salesforce Agentforce cost concerns, finance team involvement in AI cost reviews, "We've tried X but it doesn't cover Y."

Weeks 4–8: Convert Best 5 to Design Partners

After 20–30 discovery calls, identify best 5 based on: high AI spend (>$20K/month), experienced at least one invoice surprise, has both engineering and finance stakeholders, expressed genuine interest.

Design partner offer: Free access for 6 months, weekly 30-min product feedback calls, you'll build integrations for their top vendors, reference customer status at launch.

Ask them to: connect real integrations, share actual spend numbers (anonymized if needed), introduce you to 2 other potential customers.

11. Next Steps: From MVP to Production

1. Authentication & Multi-Tenancy

• Implement Clerk for authentication (email, Google SSO, future Salesforce SSO)
• Add Supabase (Postgres) with Row Level Security to isolate tenant data
• Session management and route protection via Next.js middleware

Effort: 1–2 weeks. Must be first — nothing else matters until customer data is private and secure.

2. Real Integration Layer

Start with OpenAI + Anthropic (API key, no OAuth). Validate within a few days. Then tackle Salesforce.

3. Anomaly Detection Engine

• Baseline: rolling 7-day average spend per vendor
• Alert trigger: current day spend > 2× rolling average
• V2: per-hour rate monitoring for intraday spikes

Effort: 1 week after real data pipeline is live.

4. Notification Layer

1. Email (Resend or SendGrid) — daily digest + threshold breach
2. Slack (Slack App + incoming webhooks) — real-time anomaly alerts
3. PagerDuty (future) — for critical severity events

Effort: 3–5 days for email + Slack.

5. CtrlSpend Insight Layer (AI-Powered)

When an anomaly is detected, fetch deploy history (GitHub/Linear webhook), audit trail changes (Salesforce), and resource creation events. Send to Claude API with context: "We detected a 340% spend spike starting at [time]. Here are the events in the 2 hours before: [list]. What is the most likely cause?" Display the result in the labeled blue insight box.

Cost: ~$0.01–0.05 per insight generation. Effort: 1 week after anomaly detection is live.

Go-Live Readiness Checklist