THE EDGE · ARYON
01 / 08
VI
Scroll or use the arrow keys
The Edge · Themed talk series

AI andFinance & Accounting

◆Milestones & keywords ◆Data before & after AI ◆Real-world use cases ◆Future trends
01 · Keyword

The milestones of AI

Two eras, two sets of keywords that shape how finance & accounting work with AI.

2022 – 2024
The content-generation era
LLM Large language model Generative AI AI that creates content Prompt Engineering The craft of instructing models
02 · Data

Data: before & after AI

The most fundamental shift: before AI, finance was a data-processing job — after AI, it becomes a strategic decision-making job. The classic three-step process doesn't disappear; it gets wrapped in a loop with human oversight.

Before AI · A linear process
1

Collect

Gather data by hand from vouchers, invoices and systems.

2

Analyze

Reconcile, consolidate and build reports by hand.

3

Use

Make decisions based on periodic reports.

After AI · Human in the loop
Human oversee · approve
Collect Analyze Use

AI runs all three steps continuously; the human sits at the center of the loop to control quality and make the final call.

02 · Data — Input

Change in Input

Before AI, the input was almost entirely hand-keyed structured data; unstructured data was ignored because people simply couldn't process that volume.

Before AI · Structured, hand-keyed
  • Financial statements — keyed in by hand from PDFs
  • Stock prices, exchange rates — download a CSV, then process
  • Macro data — looked up by hand from GSO, World Bank
  • Industry data — buy reports, read them manually
After AI · Everything as before + a new layer
  • NLP: real-time news, social-media sentiment, board-meeting transcripts
  • OCR + LLM: emails, contracts, legal documents — read & extracted automatically
  • Alternative data: satellite imagery, credit-card data, Google search trends
  • Speed: from daily/weekly data → real-time by the second
02 · Data — Analyze

Change in Analyze

From backward-looking — explaining the quarter that just ended, to forward-looking — forecasting and recommending actions ahead of time. Finance shifts from reporter to strategic advisor.

Analysis stepBefore AIAfter AI
CollectHand-keyed, copy-pasteAutomated: API + OCR + scraping
CleanA few days, manualA few minutes, automated
AnalyzeRatios & trends in ExcelML finds hidden patterns across millions of data points
ForecastLinear extrapolationMultivariate, non-linear ML/DL
RiskSampling checksScan 100% of transactions in real time
What-ifEach scenario computed by handThousands of Monte Carlo scenarios in seconds
ReportWritten by hand, fixed templateNarrative auto-generated from the data
Before AILittle data + slow processing + looking at the past→Report
After AILots of data + real-time + looking ahead→Decision-making

People change roles too: less collecting & computing — more time spent asking the right questions and interpreting the AI's output.

03 · Use case

Real-world applications

Three products that show AI doing real work in finance & accounting.

Use case · A

Double A

Automated accounting
  • Automatically reads and posts vouchers and invoices to the ledger.
  • Reconciles the books and flags discrepancies in real time.
  • Accountants shift into a review-and-approve role.
Use case · B

Finance Scraper

Financial data collection
  • Automatically gathers market data, reports and financial news.
  • Normalizes it into clean, analysis-ready data.
  • A continuous input source for models and management reports.
Use case · C

Legal Scraper

Tracking legal documents
  • Automatically collects laws, decrees and circulars on tax & accounting.
  • Detects regulatory changes and alerts on clauses that have impact.
  • Extracts and summarizes legal content with an LLM.
04 · Future

What comes next

Once AI touches financial data, trust becomes the biggest asset.

Data Private

Finance & accounting data stays within the business's control: models deployed privately, data never leaves the internal infrastructure.

Privacy by default

Data Secure

End-to-end encryption, access controls and audit logs for every interaction between the AI and the books — safety designed in from the ground up.

Security by design

Automation Health Check

AI scans all operational data for a periodic business "health check": cash flow, receivables/payables, process efficiency — catching weak spots early, before they become risks.

Diagnose early, act early
Summary

AI runs the work.
People take the wheel.

From Generative AI to Agentic AI, the value of finance & accounting lies in putting people in exactly the right place within the data loop.

The Edge — Aryon · Thank you for listening