Financial and Operational Computations using AIRecord inserted or updated successfully.
AI & Accounting AI & Audit

Financial and Operational Computations using AI

Author: Aakash Raju Udasi

Automated Data Extraction (Step 1):

  1. Using Gen AI technology mixed with Optical Character recognition and RAG, extract required fields in tabular structure from underlying source PDF or scanned image. Depending on your use case, the underlying source will vary.

Examples include Lease Agreements in case of Ind AS 116, Understanding documents for Brokerage computation for Asset Management Company, Discount structure for Discount computations, Dealer wise rate card for revenue reperformance and examples are many where an organisation has to compute its financial or operational numbers based on some unstructured documents such as PDF or scanned image files.

Data Validation (Step 2):

  1. As LLM models or Gen AI technology is not 100% accurate, it is wise to back your input data extracted from previous step through some validations. In this step, one can use rule-based data validation approach. Alternatively, one can use non-traditional approach where Unsupervised outlier detection technique such as Isolation Forest can be used to identify outlier in your data.
  2. In this step, one may use Python and its Data science library to run Isolation Forest algorithm.
  3. To generate Isolation Forest python script, one can very well use Gen AI such as ChatGPT, LLAMA, Gemini, etc.
  4. Please note that some basic understanding of Python and Isolation Forest working will be essential to perform this step. Alternatively, one can stick to traditional rule-based data validation.

Financial Computations using AI (Step 3):

  1. The cleansed data from Step 2 can be further processed to compute financial or operational number in excel. This step can be automated by generating the VBA script using tools like perplexity or chatgpt or any other LLM.
  2. To properly generate the VBA script, it is essential to write detailed prompt specifying the computations one needs to perform.

Accounting Entries Generation (Step 4):

  1. Last step is to generate accounting entries to feed into your ERP. The computed excel sheet in Step 3 can be used as the basis to generate accounting entries using ChatGPT or any other LLM.
  2. Please note that to achieve better accuracy, you need to back the computed excel sheet with Chart of accounts information.
  3. As a generic rule of thumb, Gen AI or LLMs are still evolving. This is where they are not 100% accurate while still helping us achieve a lot of speed in our day-to-day work or for client delivery. This is where our experience, as Chartered Accountants, comes into play to make sure that we assess and correct any potential error in above steps

Importance of topic

  1. In the above use case, i attempted to present a generic approach which has many more use cases than what i have identified above.
  2. It has use cases across various sectors such as Telecom, Retail, OEMs, Financial service industry, Manufacturing, Hospitality, etc.
  3. Moreover, the stakeholders that can leverage this use case are Finance & Accounts function, Ops team, Legal & Compliance team, Auditors (both Statutory and Internal auditors)

How much impact will AI make in your Use Case?

  1. As highlighted above and quite clear from the attached presentation, AI was the heart of everything that was done. While everything done above is very possible through rule based or manual approach, however the time investment, it will take, will be significantly higher than the AI based approach.

Alternate AI tools for your Use Case. (If Any)

  1. As stated above, this AI use case is a generic approach. How one may want to use it is up to one’s imagination and the problem statement they are trying to solve.