How does Cognocient classify AI spend as investment or waste?

Every AI API call flowing through Cognocient is automatically classified as Investment (generating business value) or Waste (over-engineered, redundant, or inefficient). The split appears live in the Engineering Dashboard and Executive Overview — no configuration required.

How auto-classification works

Classification uses a three-tier priority system. Manual overrides always win. When no manual classification exists, the system uses keyword matching and then falls back to a token-count heuristic.

Tier 1 — Manual classification (highest priority) You or your team explicitly classify a feature via the dashboard or API. This classification persists indefinitely and overrides all automatic logic.

Tier 2 — Keyword heuristic Feature names matching known production patterns (e.g., "chat", "search", "summariser", "extractor") are classified as Investment. Names matching waste patterns (e.g., "test", "dev", "playground", "debug") are classified as Waste.

Tier 3 — Token-count heuristic (lowest priority) Features with fewer than 1,000 total tokens in the last 30 days are classified as Waste (zero production impact). Features generating over 10,000 tokens/day are classified as Investment.

Manual classification via API

Classify features programmatically to integrate with your deployment pipelines — for example, auto-classify new features as "Investment" when they pass production readiness review.

# Classify a feature
curl -X POST https://api.cognocient.com/api/features/chatbot/classify \
  -H "Authorization: Bearer sk-cog-YOUR-PROXY-KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "classification": "investment",
    "reason": "Support cost reduced 34% since launch. $12 saved per ticket resolved."
  }'

import requests
 
def classify_feature(feature_name: str, classification: str, reason: str):
    """
    classification: "investment" | "waste" | "unknown"
    reason: human-readable justification stored in audit log
    """
    response = requests.post(
        f"https://api.cognocient.com/api/features/{feature_name}/classify",
        headers={
            "Authorization": "Bearer sk-cog-YOUR-PROXY-KEY",
            "Content-Type": "application/json",
        },
        json={
            "classification": classification,
            "reason": reason,
        },
    )
    response.raise_for_status()
    return response.json()
 
# Example: mark a new production feature as Investment
classify_feature(
    feature_name="contract-generator",
    classification="investment",
    reason="Accelerates contract close time by 40%. Measured via CRM deal velocity.",
)

// Node.js — classify on deploy
async function classifyFeature(
  featureName: string,
  classification: 'investment' | 'waste' | 'unknown',
  reason: string
) {
  const response = await fetch(
    `https://api.cognocient.com/api/features/${featureName}/classify`,
    {
      method: 'POST',
      headers: {
        Authorization: 'Bearer sk-cog-YOUR-PROXY-KEY',
        'Content-Type': 'application/json',
      },
      body: JSON.stringify({ classification, reason }),
    }
  );
  if (!response.ok) throw new Error(await response.text());
  return response.json();
}
 
// In your CI/CD pipeline
await classifyFeature('pdf-summariser', 'investment', 'Reduces manual review time by 2h/day');

The classification dashboard

The Cost Intelligence dashboard shows a three-column panel with your classified features ranked by spend within each category.

Click any feature to reclassify it, view its waste detection flags, and see the AI recommendation for that feature. The classification appears in the FOCUS 1.1 export as a custom tag so your FinOps tool can filter on it.

AI Investment ROI panel

The Engineering Dashboard shows a live AI Investment ROI card below the FinOps Maturity Score. It pulls directly from the waste-classification data and shows three numbers:

Below the metrics, a Board summary sentence is generated automatically: "$8,200 of AI spend is driving measurable value. $4,200 (34%) is recoverable waste — eliminating it improves AI efficiency by 34 percentage points."

This is the sentence your CFO can read in 10 seconds. No report generation needed.

Integration with Recommendations

Classification data feeds directly into the Recommendations engine. Features classified as "Waste" receive higher-priority recommendations, and the estimated savings on the Recommendations page shows the investment/waste split so you know whether you're cutting waste or trimming value.

Example recommendation card:

Next steps: Waste Detection · AI Advisor · Budget Enforcement