The Common Decision Problem
Every energy sector AI project starts with the same three questions: Can we justify the cost? Will our infrastructure support it? Do we have the internal capability to execute? I've watched utilities spend six months on vendor evaluations only to realize they skipped the foundational work — quantifying current state, sizing systems correctly, and assessing organizational readiness.
EthosPower's tool suite addresses this gap. We built these calculators after deploying n8n, Qdrant, and Ollama in air-gapped substations and watching teams struggle with the same sizing and justification questions. The tools fall into three categories: financial (ROI and cost modeling), engineering (battery system sizing), and AI capability (assessments and prompt optimization). They're designed to work as a workflow, not standalone widgets.
This article walks through each tool, what it does, who uses it, and how they chain together into a coherent evaluation and deployment process. I'll end with the minimum toolkit for three reader types: financial decision-makers, facility engineers, and AI implementation teams.
Financial Tools: Building the Business Case
SaaS vs Sovereign ROI Calculator
The flagship tool at ethospower.ai/roi-calculator does one thing: compute the true total cost of ownership for SaaS versus self-hosted open-source over 1, 3, and 5-year horizons. I use this in every initial client conversation because it forces honesty about hidden costs on both sides.
Inputs are straightforward: number of tools you're replacing, user count, per-seat SaaS monthly cost, one-time server hardware investment, and estimated implementation hours. The calculator doesn't accept vague estimates — you need actual numbers. For a typical utility replacing Salesforce, Jira, and a proprietary data lake with ERPNext, n8n, and Qdrant, you're looking at three tools, 50 users, $90/user/month SaaS ($135k/year), $25k in server hardware, and 480 implementation hours at $150/hour.
Outputs: break-even month (usually 14-18 months for mid-sized deployments), cumulative 5-year savings (often $400k-$600k), and a sovereignty score that quantifies data control and vendor independence. The sovereignty score isn't marketing fluff — it's a weighted metric based on data residency, API lock-in risk, and compliance surface area. For NERC CIP environments, this score matters as much as dollar savings.
I run this calculator before writing any proposal. If break-even is past 24 months, the project needs scope adjustment or the client isn't a good fit for open-source. No point pretending otherwise.
AI Implementation Cost Calculator
Available at /tools?tab=cost-calc, this tool estimates total project cost by scope (pilot/departmental/enterprise), team size (1-3 / 4-8 / 9+ people), and stack complexity (simple/moderate/complex). It breaks costs into discovery, build, integration, and 12-month maintenance.
The value here is forcing scope discipline. A "simple" stack means Ollama with a single LLM and basic RAG on existing PDFs. "Complex" means multi-model routing, vector database clustering, custom MCP integrations, and n8n workflows that touch six different OT systems. The cost difference is 5x, and most clients underestimate which category they're actually in.
For a 50-person operations team deploying a moderate stack (Qdrant + Ollama + AnythingLLM for procedure lookup), the calculator estimates $85k discovery and build, $40k integration, $30k first-year maintenance — $155k total. Compare that to the SaaS ROI calculator showing $135k annual SaaS burn, and you've got your justification narrative in two screenshots.
ROI Calculator and Package Comparison
These toolbar tools are lighter-weight versions for quick sanity checks. The ROI calculator focuses on productivity gains versus implementation investment — useful when you're justifying internal tooling for a single department. Package Comparison does side-by-side service tier analysis, which I use when clients ask about our managed hosting versus full DIY.
They're not replacements for the detailed calculators, but they're fast enough to run during a meeting when someone asks "what's this actually cost?"
Engineering Tools: Battery System Sizing
The eBatt suite at ebatt.ethospower.ai serves battery professionals sizing installations for BESS, UPS, and substation applications. Two calculators, both solving compliance and safety calculations that you'd otherwise do by hand or in Excel.
IEEE 485 Battery Sizing Calculator
This calculator determines the required ampere-hour capacity and cell count for a DC battery bank. Inputs: load current (amps), discharge time (minutes), temperature correction factor (typically 1.11 for 25°C rated cells in varying temps), and aging factor (1.25 is standard for 80% end-of-life capacity).
Example: You're sizing a battery bank for a 200A load that must sustain for 120 minutes. Temperature correction 1.11, aging factor 1.25. The calculator outputs 555 Ah required capacity and recommends a specific cell count based on system voltage (48V/125V/250V).
I've used this for substation UPS sizing where the alternative was a $15k engineering study. The IEEE 485 standard is well-established; the calculator just automates the math and ensures you don't miss correction factors. Output feeds directly into procurement specs.
EN 50272-2 Hydrogen Venting Calculator
This addresses the safety requirement for battery room ventilation to prevent hydrogen accumulation during charging. Inputs: number of cells, maximum charge current (amps), and calculation method (conservative or operational).
Output is ventilation rate in cubic meters per hour and recommended air changes per hour. For a 100-cell lead-acid bank charging at 50A, the calculator specifies 425 m³/h ventilation with 6 air changes per hour using the conservative method.
This matters for NERC CIP physical security compliance — you need documented ventilation calculations for battery rooms in critical facilities. The calculator generates numbers you can hand directly to your HVAC contractor and your compliance auditor.
AI Capability Tools: Readiness and Optimization
AI Readiness Assessment
The assessment at /tools?tab=assessment is a 5-question diagnostic covering data management maturity, AI familiarity, security policies, IT infrastructure capability, and leadership appetite for change. Each question scores 1-4, total range 5-20.
Scoring tiers: Foundation (5-9), Developing (10-14), Advanced (15-18), Leader (19-20). A Foundation score means you need data governance and infrastructure work before touching AI. Developing means pilot-ready. Advanced means you can deploy at departmental scale. Leader means enterprise rollout is viable.
I run every prospect through this before proposal work. A Foundation score with executive pressure for immediate AI deployment is a red flag — the project will fail or turn into a science experiment. The assessment includes next-step recommendations tailored to score tier, which gives us a shared roadmap.
Example output for a score of 12 (Developing): "Your data is catalogued but not centralized. Pilot an AI use case in a single department with clean data access. Focus on structured knowledge retrieval before generative tasks. Budget 3-6 months for infrastructure stabilization parallel to the pilot."
PromptCraft and Prompt Optimizer
Available at /promptcraft and /tools/promptworker, these tools implement the RSTCC framework: Role, Step-by-step instructions, Task definition, Context, and Constraints. You paste a prompt, the analyzer scores quality across those five dimensions, identifies weaknesses, and generates an improved version.
Freemium model: three free optimizations, then paid. I use this when clients hand me a 50-word vague instruction and wonder why their Ollama outputs are inconsistent. The optimizer forces structure — specifying role ("You are a NERC CIP compliance specialist"), breaking tasks into steps, providing context ("This procedure applies to BES Cyber Systems at High impact sites"), and defining constraints ("Responses must cite specific CIP standard sections").
Before/after example: Original prompt "Summarize this substation maintenance log." Optimized prompt: "You are an electrical substation maintenance analyst. Step 1: Identify all equipment mentioned and classify by voltage class. Step 2: Extract anomalies or deviations from normal parameters. Step 3: Flag any entries requiring immediate follow-up. Context: This log covers weekly inspections per IEEE C37.2 standards. Constraints: Output as bullet list, cite log entry timestamps, do not infer root causes."
The optimized version produces structured, auditable output. The original produces a paragraph of vague summary. The difference matters when the output informs a switching order or a compliance filing.
EthosAI Chat and Multi-Agent Demo
The chat interface at /chat is a RAG-powered assistant with knowledge of energy sector operations and our platform stack. Four agent personas handle routing: General, Tech Advisor, Business Analyst, Technical Writer. Ask about NERC CIP compliance and you get routed to Tech Advisor with citations. Ask about ROI and you get Business Analyst with calculator links.
The Multi-Agent Demo in the toolbar shows live handoff behavior — watch a query get routed from General to Tech Advisor, handed to Business Analyst for cost modeling, then to Technical Writer for documentation. It's a working prototype of the orchestration layer we deploy in client n8n workflows.
I use the chat for quick sanity checks on technical questions ("What's the recommended Qdrant collection configuration for 500k procedure documents?") and send clients to the demo when they ask how multi-agent systems work in practice.
How the Tools Work Together: A Typical Workflow
Here's how I use these tools in sequence for a new utility client evaluating AI deployment:
- AI Readiness Assessment (/tools?tab=assessment) — Establish baseline capability and identify gaps. Score determines whether we proceed to pilot scoping or infrastructure remediation.
- SaaS vs Sovereign ROI Calculator (ethospower.ai/roi-calculator) — Quantify the financial case for open-source versus continuing current SaaS spend. Output feeds board-level justification.
- AI Implementation Cost Calculator (/tools?tab=cost-calc) — Estimate project cost based on assessment score and desired scope. Cross-check against ROI calculator break-even timeline.
- PromptCraft (/promptcraft) — If proceeding to pilot, optimize initial use case prompts to ensure quality output from day one. Poorly structured prompts torpedo early demos.
- eBatt Calculators (ebatt.ethospower.ai) — If infrastructure work includes battery system upgrades (common for air-gapped edge deployments), size installations correctly. Bad battery sizing causes downtime during grid events, which kills AI project credibility.
- EthosAI Chat (/chat) — Ongoing technical Q&A during implementation. Faster than documentation search for edge cases.
This workflow takes a prospect from "should we do this?" to "here's the sized system and project plan" in 2-3 weeks instead of 2-3 months of vendor evaluation theater.
Recommended Starting Points by Role
Financial Decision-Makers (CFO, Finance Director)
Start with SaaS vs Sovereign ROI Calculator and AI Implementation Cost Calculator. Your job is justifying spend and managing risk. These tools give you defendable numbers for board presentations. Run sensitivity analysis — what if implementation takes 20% longer? What if SaaS costs increase 15%/year (they will)? The ROI calculator handles these scenarios.
Ignore the engineering tools unless you're evaluating infrastructure CapEx. Glance at the AI Readiness Assessment to understand organizational gaps, but delegate execution to your CTO.
Facility and Electrical Engineers
Start with eBatt calculators. You're sizing real systems that must meet code and perform during outages. IEEE 485 and EN 50272-2 compliance isn't optional, and these calculators eliminate manual errors that cause undersized installations or ventilation failures.
Use the ROI calculators when justifying capital equipment purchases to management — showing that a $40k battery bank investment enables a $200k AI deployment with 18-month payback gets budget approval.
AI Implementation Teams (IT Directors, Data Scientists, Solutions Architects)
Start with AI Readiness Assessment, then PromptCraft, then EthosAI Chat. The assessment tells you if your organization can actually execute. PromptCraft ensures your demos don't embarrass you with low-quality LLM output. EthosAI Chat becomes your technical reference during build.
Use the cost calculators to sanity-check vendor proposals and internal project estimates. If a vendor quotes $400k for a "simple" stack deployment, the calculator will show you they're either inflating scope or padding hours. Either way, you have negotiating leverage.
The Verdict
The minimum toolkit depends on your immediate decision:
Evaluating whether to pursue open-source AI at all? Run the AI Readiness Assessment and SaaS vs Sovereign ROI Calculator. 30 minutes of input, clear go/no-go answer. If readiness score is below 10 and ROI break-even is past 30 months, stop here and fix foundational issues first.
Already committed, now scoping the project? Add AI Implementation Cost Calculator and PromptCraft. You need cost boundaries and quality output from day one. Budget 2-3 hours to run scenarios and optimize your core use case prompts.
Deploying infrastructure in parallel? Add eBatt calculators if battery systems are in scope. Undersized UPS or failed ventilation compliance will delay your AI project by months. Spend the 20 minutes to size correctly now.
I've watched teams skip these tools and learn the hard way. The utility that deployed Ollama without running the ROI calculator discovered at month 6 that their server undersizing meant constant performance issues — they eventually spent more on hardware upgrades than they saved in SaaS costs. The substation modernization that skipped IEEE 485 sizing had a battery failure during a test event, which killed executive confidence in the entire AI initiative.
The tools are free and take minutes to run. Use them. They encode 30 years of practitioner mistakes so you don't have to repeat them.