Interactive calculator for estimating the business value of Solace Agent Mesh relative to framework-only approaches (LangChain, OpenAI SDK, Bedrock, etc.).
Live: solacese.github.io/sam-roi
- 6 value cards with live formulas
- Conservative / Base / Aggressive preset toggle
- Editable advanced assumptions
- Stack selector (adapts "Today:" text per competitor)
- Region presets for engineer cost
- Diminishing returns at scale (agents > 50)
- Export as PDF (single landscape page)
- Copy summary to clipboard
- Shareable URL (inputs encoded as query params)
| Assumption | Base Value | Source |
|---|---|---|
| Effort reduction per new scenario | 40% | Field-observed across 5+ enterprise deployments |
| Token savings | 20% | BBVA production data (25% measured, 20% conservative) |
| Cost per system outage | $250K | Gartner avg downtime cost for enterprise SoR |
| Cost per additional jurisdiction | $90K | Platform team build cost (6 eng-months equivalent) |
| Pilot failure burn | 4 months x 3 engineers | Industry avg failed AI pilot duration |
| Time-to-production (today) | 16 weeks | Customer interviews, pre-SAM baseline |
| Time-to-production (SAM) | 6 weeks | Measured across early adopters |
Validated proof points:
- BBVA: 25% token savings, 400+ agents in production
- British Telecom: 1,000+ agents monitored
- Discovery call: Open the calculator, plug in the prospect's numbers live. Start with Base preset.
- Follow-up email: Use "Copy Summary" to paste into email with the shareable URL.
- Business case deck: Export as PDF, include as an appendix.
- Technical deep-dive: Open Advanced Assumptions, walk through each formula to build credibility.
- Executive sponsor: Use Aggressive preset for upside scenario, Conservative for floor.
Tips:
- Always acknowledge these are estimates. Say "directional" not "guaranteed."
- Card 3 (systems of record) is risk-based. Frame it as insurance, not savings.
- The diminishing returns function prevents absurd numbers at scale. If a prospect has 500+ agents, the calculator stays credible.
Single HTML file, no build step. Open index.html in a browser.
# Serve locally
python3 -m http.server 8000
open http://localhost:8000- Edit
index.htmldirectly - Test all 6 cards with default and extreme inputs
- Verify PDF export renders on one page
- Push to
main- GitHub Pages deploys automatically
MIT