● AI Study Case

Token Saving in SME Financial Statement Automation

Comparing AI Agent repetition vs Detached System with Smart Routing for 100 SME bank statement sets.

AI Study Case

100 SME Bank Statement Sets → Financial Statements

100 SMEs upload 100 different sets of bank statements. The goal is to generate accurate financial statements for all SMEs.

100

SMEs

Small & Medium Enterprises

100

Bank Statement Sets

Different Bank Formats

100

Financial Statements

Generated Reports

The problem is not whether AI can do the job.
The real question is how the system should be designed.
Approach 1

AI Agent Processes 100 Reports One by One

The AI agent reads, understands, classifies, calculates, validates and generates each report individually repeated 100 times.

REPEATED FULL WORKFLOW ×100
Input
Bank Statements
Full AI
Reasoning
Classify &
Calculate
Validate &
Generate
↻ REPEAT ×100 • FULL CONTEXT RELOAD
Total Tokens
~7.5M
GPU Energy
~55 kWh
CO₂ Emissions
~24 kg
Context Reload
100%
Key problems: Full reasoning repeated 100× • 50–100k tokens reloaded every time • No knowledge reuse
Approach 2

Detached System + Smart Routing

AI is used once to build a reusable detached system. Bank statements are segmented and intelligently routed with minimal context.

SMART SEGMENTATION + ROUTING
Input
Bank Statements
Segmentation
By Bank/Format
Smart Router
Intelligent Decision
⚡ Rules Engine
1.2k–2.8k
🤖 Selective AI
4k–9k
👤 Human Review
~12k (rare)
Total Tokens
~1.0M
87% reduction
GPU Energy
~7.3 kWh
87% reduction
CO₂ Emissions
~3.2 kg
87% reduction
Context Reload
~10%
90% reuse
Key advantages: Segmentation by bank type • Rules first • AI only when needed • Persistent small context
Token Efficiency

The Difference: 87% Token Reduction

Fewer tokens. Same outcome. Much smarter system design.

🔁

AI Agent One-by-One

7.5M
tokens (baseline)
Per SME50k – 100k
GPU Energy~55 kWh
API cost impact~$6,000+
🧭

Detached + Smart Routing

1.0M
tokens (87% saved)
Per SME5k – 15k
GPU Energy~7.3 kWh
API cost impact~$800
Estimated Savings (100 statements)
~6.5M tokens saved
~$5,200 API savings
47.7 kWh energy
~85% faster
Methodology

Assumptions & Sensitivity

Transparent model inputs for investor due diligence. Validates the RM 288K/SME/year unit economics claim in the pitch deck.

Parameter Base Case Best Case Worst Case
SMEs processed 100 100 100
Statements per SME 12 / year 12 / year 12 / year
Avg tokens / statement (Approach 1) ~8,500 ~6,000 ~12,000
Avg tokens / statement (Approach 2) ~1,100 ~800 ~1,600
Token price (external API) $0.002 / 1K $0.0015 / 1K $0.003 / 1K
Annual savings (100 SMEs) ~$5,200 ~$7,800 ~$3,100

Study date: July 2026 · Local inference API fee assumed RM 0/token (GPU infrastructure amortization tracked separately)

Validates the pitch deck unit economics

This study is supporting evidence for the RM 2M investor round. See the full financial model, cap table and growth roadmap.

* Internal validation based on live platform data. Independent audit available on investor request.

View Pitch Deck Request Investor Meeting
Sustainability & Real Impact

From Token Saving to Real-World Impact

87%
Less energy per document

Lower GPU compute and data center load

~21 kg
CO₂ emissions avoided

Significant during heatwaves when cooling demand spikes

Affordable
AI for SMEs

Structured systems make automation viable for smaller businesses

Auditable
+ Guardrails

Rules + selective AI + human review layer

“The smartest AI system is not the one that uses the most tokens.
It is the one that knows when not to use them.”
AINNA
CLICK ME

Site Sections

No section data available yet.

Sites with documented sections will appear here.

🔒 Inference traffic secured via VPN, encryption, access control, and network allowlisting, with no public endpoint exposure under proper private deployment.

© 2026 AINNA NeuralOps Token Saving Study. All rights reserved.

Made with ❤️ in Melaka, Malaysia