The first time AI actually scared me

For years, my view on AI was simple.

Useful? Yes.
Reliable enough for serious financial work? Not quite.

In our profession, trust is everything.

A small mistake in a tax position or financial statement is not just another mistake… it can mean penalties, litigation, or reputational damage.

So most of us (atleast the sane ones) used AI for lighter tasks: drafting emails, summarising documents, or speeding up small bits of research.

Helpful tools, but nothing that changed how we actually worked.

That perception shifted for me recently because I saw how dramatically it can reduce the preparation work behind financial analysis.

Let me share the moment when that clicked.

Section 1 – The Moment AI Started Feeling Different

Recently, I came across Claude Cowork.

This is not a sponsored mention by any means — just sharing something I experimented with in my practice.

What makes it different is that it can work directly with folders of documents, instead of you feeding one file at a time.

Anyone who deals with tax or compliance work knows what that means.

Our work rarely involves one document. It usually involves a chain of notices, orders, replies, assessments and submissions across multiple years.

Piecing together that history itself takes time.

So I decided to test it.

I uploaded an entire folder containing notices and orders from a tax litigation case we were handling. Multiple years, multiple communications.

Then I asked a simple question:

“Analyse the case history and suggest the possible way forward.”

The model I was using was Sonnet 4.6, which is not even their strongest version.

Within about 10 minutes, it produced:

Now, let me be very clear.

It was not the final output. No responsible professional should treat AI output as final advice.

But the accuracy was easily around 85%.

And that is where the real impact lies.

Not perfection.

Time saved.

The kind of groundwork that would normally take my team a few hours was prepared in minutes.

That is when I realised something important.

"AI is not replacing the professional judgment.

But it is dramatically compressing the time required to reach that judgment."

Section 2 – The Real Example (Tax Litigation Case)

Let me explain a practical use case.

In litigation matters, one of the most time-consuming tasks is reconstructing the story of the case.

You may have:

These documents often span several years.

Before we even start drafting our strategy, someone from the team has to:

This groundwork is essential but it takes time.

When I uploaded the folder to the AI system, it did exactly that groundwork.

It created a timeline of the case, identified the key points of dispute, and suggested possible legal responses supported by case laws.

Was it perfect? No.

But it got us 90% of the way there.

From there, our role shifted from collecting and organising information to
reviewing, refining and applying professional judgment.

And that is where the real value lies. What used to take 80 hours of team work now takes 8 hours!

Section 3 – What AI Can Actually Do in Finance Today

Many founders hear about AI and assume it is mostly for marketing or coding.

But it is quietly becoming very useful in finance and compliance work as well.

Let me share a few practical examples I have been experimenting with.

If you provide a properly formatted trial balance, AI tools can help convert it into:

It can automatically group accounts into assets, liabilities, revenue and expenses.

Again — you must verify the classification — but the initial structuring work becomes extremely fast.

Disclosure requirements are expanding every year.

AI can scan financial statements and highlight potential gaps such as:

Think of it as a second pair of eyes during review.

One of the common issues in financial reporting is inconsistency between statements.

For example:

AI tools can check these linkages quickly and flag inconsistencies.

This kind of cross-checking is tedious manually but easy for machines.

As I saw in the litigation example, AI can draft responses supported by:

The draft still requires professional review, but the research groundwork becomes faster.

Tax professionals often deal with massive document trails.

AI can scan entire folders and answer questions like:

What issues are repeated across years?
What arguments have already been used?
Which notices remain unanswered?

This kind of document review used to take hours.
Now it can be done in minutes.

Auditors and finance teams spend significant time preparing working papers.

AI can help create:

It reduces the administrative load and allows professionals to focus on analysis rather than formatting.

A Balanced Perspective (And an Important One)

Before anyone gets carried away, let me say this clearly.

AI outputs must always be reviewed by a professional.

These tools do not carry accountability.

They do not sign audit reports.
They do not appear before tax authorities.
They do not bear professional liability.

That responsibility remains human.

What AI changes is something else.

It reduces preparation work.

And when preparation becomes faster, the importance of expertise actually increases.

Because the real value is no longer in collecting information.

It is in interpreting it correctly.

There is one important catch in all of this.

AI is only as good as the instructions you give it.

Ask vague questions, you get vague answers.
Give structured prompts, you get surprisingly strong results.

This is why people talk about prompt engineering.

It simply means learning how to:

Provide clear context
Structure your questions properly
Ask the AI to reason step by step

The good news?

Almost all this knowledge is freely available online.

Anyone willing to spend a few hours learning can significantly improve the output they get.

The Takeaway

Professionals who adopt such tools move ahead much faster than those who resist them.

AI will not replace good professionals.
But one thing is becoming clear: Professionals who use AI will replace those who don’t.

For founders, the message is simple.

Start experimenting.

Use these tools for document analysis, financial reviews, research, and internal reporting.

Ofcourse, AI makes the most silly mistakes ever to exist. It can use calculus properly but will mess up 2+2 (quite literally!)

So do not ever let AI replace your work.

Even if they are not perfect, the time saved can be enormous.
And time, as every entrepreneur knows, is the most valuable resource you have.

As always…
To all those with a mission in life,
VijayBhava!

P.S. If you found this useful, do share it with fellow founders who might benefit from understanding how finance is evolving on the ground.