Assess one real AI workload before committing to architecture or infrastructure. PulseAI evaluates the workload and provides a deterministic, provider-neutral placement and sizing view across Cloud, On-Premises and Hybrid. R99 per workload. It does not affect the PulseAI Score.
Deterministic, provider-neutral, one workload per assessment. Already have an account? Sign in
| Question | Answer |
|---|---|
| What is assessed? | One real AI workload |
| Price | R99 per workload |
| Placement options | Cloud, On-Premises and Hybrid |
| Method | Deterministic rules and formulas |
| Vendor bias | Provider-neutral |
| PulseAI Score impact | None |
| Main outputs | Placement, sizing, economics where supported, confidence and validation requirements |
Objective. Users, requests, concurrency. Model and data characteristics.
Sovereignty, network, site, skills, and power or cooling where relevant.
Cloud, On-Premises and Hybrid, with infrastructure sizing and economics where supported.
Placement recommendation, confidence, assumptions, validation requirements and next steps.
| Stage | What happens |
|---|---|
| 01 Describe workload | Name the workload and describe what it needs to achieve |
| 02 Measure demand | Users, requests, concurrency, latency and availability expectations |
| 03 Model and data | Model characteristics, context, working set, retention and growth |
| 04 Constraints | Sovereignty, data movement, network, skills, site, power and cooling |
| 05 Size infrastructure | CPU, RAM, storage, and accelerator class, count and topology where justified |
| 06 Compare placement | Cloud, On-Premises and Hybrid each scored independently on the evidence |
| 07 Economics | TCO and break-even where they can be calculated reliably |
| 08 Validation | Benchmark and engineering review flagged wherever the evidence requires it |
| 09 Result | Placement, sizing, assumptions, confidence, validation requirements and next steps |
| Output | What it tells you |
|---|---|
| Placement | Cloud, On-Premises or Hybrid |
| Confidence | How strongly the evidence supports the result |
| Sizing | Server and accelerator requirements where justified |
| Economics | TCO or break-even where evidence permits calculation |
| Assumptions | What had to be assumed |
| Validation | What still needs benchmarking or engineering review |
| Next step | What should happen before committing |
A useful assessment should distinguish between what the evidence supports and what still needs validation. PulseAI does not turn incomplete inputs into a false production specification.
R99 per workload · deterministic · provider-neutral · does not affect your PulseAI Score.