CVOCA

The Forward Deployed CA

Karan Shah November 1, 2026 Other ⏱️ 9 min read

A recent YouTube video, Untapped Goldmines for Indian Engineers in the AI Era, makes a simple argument. The Indian IT industry was built mostly on writing software as per the client’s specifications, and AI is now getting very good at writing code. So now the industry has to move to the other half of the work: understanding the client’s real problem, sitting with them, and owning the result.

Swap “writing software” for “preparing returns and reports”, and the argument describes our profession just as well.

What AI is doing to Indian IT

For decades, Indian IT grew in a straight line: more work meant more engineers. That link is breaking. TCS’s revenue per employee rose 3.4% in FY26 even as its workforce shrank by about 2%, and HCLTech’s CEO said the company had grown 4 to 5% without adding headcount. The industry’s answer is a role with an odd name, the forward deployed engineer. Hiring for it in India grew about 130% in a year.

What is a forward deployed engineer?

A software engineer works from a specification: someone else has already decided what to build, a bit like preparing a return from the papers a client sends. A forward deployed engineer starts earlier, from a messy business problem, the kind a client describes as “our cash is always tight”. They work inside the client’s organisation, learn how things really run, deal with the people and the politics, and push back when what the client asks for is not what the business needs. Then they build the fix.

If three clients have the same problem, a good forward deployed engineer doesn’t fix it three times. They take what they learnt back to the company, which builds the fix into the product for everyone. Without that, it becomes just another custom project.

This role began at Palantir, a US data company. Microsoft, Amazon Web Services and OpenAI now send engineers like this to work inside their clients’ businesses too. Recruiters say the people most in demand are those with deep knowledge of the client’s industry.

Most of that description fits a good CA: sitting with the client, understanding how the business really runs, and telling them what they need rather than what they asked for. That is why the role may suit a chartered accountant better than it suits an engineer.

Why this is our story too

Start with the work itself, because compliance is to us what code is to IT. A salaried taxpayer’s income tax return now arrives largely pre-filled. Salary, TDS, interest and dividends come in from what employers, banks and others report to the department (visible in Form 26AS and the Annual Information Statement), and capital gains increasingly do too. ClearTax says its AI assistant can file a return in under three minutes over a WhatsApp chat. An article on BDO India’s website in June put it plainly:

“Most salaried individuals with simple income sources can file their ITR without hiring a CA.”

GST is going the same way. From the July 2025 tax period, the tax liability in GSTR-3B is filled in from GSTR-1, GSTR-1A and IFF, and cannot be edited in the return; corrections go through GSTR-1A instead. More and more of the return prepares itself.

That matters because production is where most of our day, and most of our fee, goes. Look honestly at how a small practice spends its time: returns, reconciliations, schedules, reports. Many of us spend more effort making the audit report than doing the audit. If a tool can produce the report, clients will sooner or later stop paying for the hours that went into it.

The client will go to whoever uses AI to get closest to their problem. Software companies are already hiring for this. FloQast and Sumary, both in London, have advertised for “forward deployed accountants” who work with clients’ finance teams, often in their offices, and set up AI agents in their processes. The one who gets there first could be a tech startup, or the CA firm down the road.

The video also says the role is hard to fill because it needs technical skill, consulting skill and judgment together, and that it is best learned by apprenticeship. Articleship is that apprenticeship. We already have the judgment and the client’s trust, which is the hard part. Building the tools is the part AI now makes easy.

Four things a CA can start now

1. Automate your own routine work first

Start at home. Ask AI to help you write a tool, say a Python script that prepares a debtor’s ageing report from a Tally export. Test it on dummy data, then run it on client data on your own computer. The client’s data never goes into AI, and the same tool works for every client.

Take the audit report. Automate the report, and the hours go back into the audit itself. AI can help there too, by going through the full ledger, picking samples and flagging unusual entries, while the judgment and the signature stay with you.

You don’t need to know Python. AI writes the code; you check that it works. I have seen CA firms prepare TDS returns in minutes, reconcile GST for clients automatically, and send MIS reports that update themselves. ICAI’s AI site has dozens more, demonstrated by the CAs who built them. It is worth an evening of browsing.

Start small, with one tool for one task. And decide early who will look after it, because a tool that only one person understands is a risk of its own.

2. Supercharge your articles and staff

Many CA firms grew the way Indian IT did, on large teams of articles and employees doing routine work. The way out is to make each person more capable. An article or employee who spends the year on data entry learns data entry. One who builds a reconciliation tool and uses it on ten clients learns how ten businesses work.

Good people are already hard to hire and harder to keep, and a team like that eases the problem, since you need fewer new hands to take on more work. But watch where the saved time goes. If efficiency only goes into cutting fees, we will repeat what parts of IT are going through now, handing the gains back to clients as lower prices. The time saved has to go into work clients value more: the audit itself instead of the report, the client’s real problem instead of the form.

3. Fix one part of the client’s business at a time

Some virtual CFOs I know already work this way. They don’t take over the whole finance function at once. Like the big consulting firms, they pick one area (receivables, inventory, payroll, vendor payments, or a part of the business itself), sit with the team, fix it, and move to the next.

What they leave behind is different.

A consultant leaves a report; a forward deployed CA leaves a process that works.

You don’t need a desk in the client’s office for this. You need to know how their work actually gets done.

For members in industry it is simpler still. You are already inside the business, so you can be its forward deployed CA.

4. Build it once, use it for every client

Learning AI won’t set anyone apart for long. ICAI says it has trained over 50,000 members in AI, which is good for the profession and also means that knowing AI will soon be common. What will set a CA apart is what they build with it, and how often they can use it again.

Say a client gets a GST notice about input tax credit, or a virtual CFO takes on a client’s vendor payments. Under GST, credit must be reversed, with interest, if the supplier is not paid within 180 days of the invoice, and it can be claimed again once paid. So, you build a report that shows how many days each invoice took to be paid.

Build it for one client and you have answered one notice. Build it to work for any client, from any Tally export with nothing specific to one business, and you have a tool for every client you serve. The next notice takes minutes instead of days. This is the three-clients idea from earlier, in CA terms, and it is the step most firms miss. Over time you build a library of tools drawn from real client problems, which a competitor who has only learnt AI cannot copy.

Track productivity alongside revenue

Every year we check whether revenue has grown. It is worth adding one more: productivity, meaning how much work each person handled, compared with last year.

Revenue can grow through fee increases or new hires. Productivity shows how much AI is adding in your firm. Indian IT watches the same measure as its revenue grows without more people. With AI, the number should jump, not inch up. If it moves only a little, look at which work the AI is being used for.

So next year, along with how much you earned, ask how much more your team could do, and where the time went.

The video calls this an untapped goldmine for Indian engineers. For chartered accountants, it may be even closer to home.

Further reading

Sources

[The author can be reached at karan@shandar.in]

Post Views: 12
Share on:

Related Posts

Other
Understanding Insider Trading and Governance in India
The SEBI (Prohibition of Insider Trading) Regulations, 2015, serves as...
Other
Peer Review Decoded: What Every CA Firm Must Know Today
Imagine investing years building your CA firm’s reputation — only...
Other
ESG: Where Sustainability Meets Financial Reality
Introduction Over the last few years, businesses in India and...
Scroll to Top