Skip to content
JATIN SACHDEVA

Case Studies (Demonstrations)

Three educational projects showing how AI could be applied. Each one is a demonstration, not a real client engagement, and none report measured results.

Contents

Each demonstration follows the same seven-part format.

AI-powered financial reporting

How a finance team could draft monthly management commentary faster, with a qualified person approving every figure.

Demonstration project. Not a real client engagement.

1. Business problem

Assembling the monthly management pack takes several days, mostly spent copying numbers and writing the same kind of commentary each month.

2. Existing process

  • Export trial balance and management accounts from the accounting system
  • Copy figures into a reporting template by hand
  • Write variance commentary from scratch each month
  • Circulate for review and revise

3. AI solution

An AI assistant drafts the variance commentary from approved figures. The finance lead checks the numbers and the reasons before anything is sent.

4. Tools used

  • Spreadsheet
  • A general AI assistant
  • Checklist for review

5. Implementation workflow

  1. 1.Lock the approved figures in the spreadsheet
  2. 2.Paste only the approved figures, with last month's figures, into the assistant
  3. 3.Ask for a draft commentary using the rule 'do not give reasons that are not in the data'
  4. 4.Mark anything the assistant could not explain as [CHECK]
  5. 5.The finance lead confirms each reason with the budget owner
  6. 6.Save the final version and record who approved it

6. Results and limitations

  • No results have been measured. This is a demonstration of the approach.
  • Limitation: the assistant may suggest a reason that sounds plausible but is not supported by the data.
  • Limitation: commentary is only as good as the figures and the explanations behind them.

7. Quality controls

  • Totals in the draft are checked against the ledger
  • Every reason is confirmed by the budget owner
  • A named person signs off each version
  • Prompts and outputs are kept with the pack for audit

AI invoice processing and accounts payable

How an accounts payable team could use AI to prepare invoices for checking, while keeping approval with people.

Demonstration project. Not a real client engagement.

1. Business problem

Invoices arrive in many formats. Staff re-type details, match them to purchase orders and chase missing information.

2. Existing process

  • Invoices arrive by email and in shared folders
  • Staff type supplier name, amount, date and reference into the system
  • Match each invoice to a purchase order by hand
  • Chase suppliers for missing details

3. AI solution

An AI step pulls the key fields from each invoice into a table, flags mismatches and drafts a request to a supplier for missing details. A person approves every payment.

4. Tools used

  • Invoice inbox or shared folder
  • Spreadsheet or accounts system
  • Automation tool (such as n8n)
  • A general AI assistant

5. Implementation workflow

  1. 1.New invoice is saved to a shared folder
  2. 2.The AI step reads the invoice and fills in supplier, date, amount and reference
  3. 3.The step compares the result with the purchase order list
  4. 4.Mismatches are placed in a review list with a short reason
  5. 5.A draft request is written for any missing detail
  6. 6.An accounts payable clerk reviews and approves each item

6. Results and limitations

  • No results have been measured. This is a demonstration of the approach.
  • Limitation: poor scans and handwritten notes can lead to wrong fields.
  • Limitation: bank details must never be changed by AI alone. Verify changes by a separate call-back.

7. Quality controls

  • Every payment needs a human approval
  • Bank detail changes require a separate verification step
  • Extracted fields are checked against the source image for a sample each week
  • Errors are logged so the process can be improved

AI-powered investment research

How an analyst could use AI to scan company updates and flag what has changed, while forming their own view.

Demonstration project. Not a real client engagement.

1. Business problem

Analysts must read many filings, news items and earnings updates. Important changes are easy to miss in a long list.

2. Existing process

  • Collect company filings and news for a list of companies
  • Read each document and note key points by hand
  • Compare with the previous period from memory or old notes
  • Write a summary for the team

3. AI solution

An AI step summarises each new document and compares it with the last one, listing changes with a link to the source. The analyst reads the flagged items and decides what matters.

4. Tools used

  • Official filings and public news sources
  • A research assistant such as Perplexity or Claude
  • A notes document

5. Implementation workflow

  1. 1.Collect the newest filing or update for each company
  2. 2.Ask the AI for a summary of key points, with the exact section each point came from
  3. 3.Ask it to compare with the previous period and list only the changes
  4. 4.The analyst opens each source and confirms every flagged change
  5. 5.The analyst writes the view in their own words

6. Results and limitations

  • No results have been measured. This is a demonstration of the approach.
  • Limitation: summaries can drop a caveat or misread a table, so the source must be read.
  • Limitation: this is research support only. It is not investment advice.

7. Quality controls

  • Every point must link to a source section
  • The analyst confirms every flagged change in the original document
  • Opinions and recommendations are written by people, not generated
  • The team keeps a log of what was checked