AI in Accounting: What Should Firms Automate First in 2027?

Accounting firms are hearing about new AI tools almost every week, but adopting more technology does not automatically create better results. The harder questions are which problems are worth solving, which workflows are ready for automation, and how firms can measure whether AI is actually creating value. 

To explore those questions, Ace Cloud Hosting spoke with Lavesh Chattlani, President of Scalene AI Inc. Lavesh has spent more than a decade working at the intersection of accounting and technology, including leadership roles at Intuit and Sage.

Today, he helps accounting, finance, and professional services teams approach AI with a practical focus on strategy, governance, process improvement, and measurable outcomes. 

In this conversation, Lavesh explains why firms should start with the business problem rather than the tool, which tasks are best suited for automation, what needs to be standardized first, how to measure ROI, and where human review and data security must remain firmly in place. 

Accounting firms are being presented with new AI tools almost every week. How should they decide which ones are actually worth their attention?  

Start with the problem, not the tool. I know that sounds obvious, but you’d be surprised how often I have to keep repeating it. We’ve bought technology this way for decades, and for some reason with AI, we tend to put the tool first.  

Before you sit through any demo, write one sentence that includes: who has the problem, what it is, why it matters, and how you’ll measure success. If you can’t do that, you’re not ready to buy anything.  

Then ask the vendor a few honest questions. Where does our client data go, and is it used to train your model? And where is the product heading, not just what it does today?  

One last thing. Most firms are not as far ahead as you think. In 2025, MIT found that 95% of generative AI pilots showed no measurable impact on the bottom line. So there’s no prize for buying first. There is a prize for buying right.  

What types of accounting or finance tasks are the best candidates for automation today, and which ones should firms leave alone?  

The best candidates are tasks that are high volume, low judgment, and follow the same rules every time. Think data entry from receipts and invoices, bank feed matching, moving files between systems, sending reminders, or pulling together a first draft of a client email. If a task happens daily or weekly, has clear rules, and is prone to keying errors, it belongs on your list.  

Agentic AI is where things get interesting. I worked with a not-for-profit where one month-end process was taking someone seven to eight hours every month. We got it down to 30 minutes, and that freed them up to finally work through a backlog they’d been sitting on.  

What should you leave alone? Anything where the final call is professional judgment, like a tax position or an audit opinion. AI and automation can help you research and draft, but the output should never be the final product for sign-off. 

Before automating a workflow, what should a firm fix or standardize first so it does not simply automate a bad process?  

Sit down and watch the work being done by the people on the front line. Talk to the bookkeeper or controller who actually lives the pain, map where the time, errors, and rework pile up, and keep asking why until you hit the root cause.  

I’ll give you an example. A client wanted to use Copilot to automate a spreadsheet an analyst ran every morning to split out Quebec data from the rest of Canada. When I watched him do it, it turned out a pivot table that refreshes when you open the workbook would solve it. No AI needed. He’d just never had any formal Excel training.  

So, before you automate, standardize. Make sure everyone does the steps the same way, the data comes in a consistent format, and someone owns the process. If three people do the same task three different ways, pick one and write it down first. Skip this and no technology will save you. You’ll just automate a mess, only faster.  

How should firms measure whether an AI or automation project is genuinely saving time or generating ROI?  

Measure business outcomes, not activity. I’ve heard of organizations measuring AI success by how many tokens they consume. I have never heard anything more ridiculous. Usage tells you people are logging in. It tells you nothing about value.  

Set your baseline before you start. If month-end close takes 10 business days and hundreds of staff hours, write that down. Then track it. If one process drops from seven hours to 30 minutes, and the close moves from nine days to eight, you know you’re trending the right way. 

Pick one or two numbers that matter to the business: hours saved, error rates, days to close, turnaround on client requests. Put a dollar value on those hours where you can. And be honest about editing time. When ChatGPT first came out, a lot of people spent longer fixing the AI’s email than it would have taken to just write it themselves. That’s a loss, not a saving, and it should show up in your numbers. 

As firms build more AI into their workflows, how can they balance faster automation with human review, data security, and professional judgment?  

My philosophy is simple. Technology scales our capacity, but human judgment scales our impact. AI should speed up the work, not take your name off it.  

I think about it in three parts.  

• Policies: decide what data can go into which tools. Free consumer tools can use what you type to train their models, so client data never goes there. If you’re on Microsoft 365, bookmark https://m365.cloud.microsoft so staff doesn’t wander into the consumer version by accident.  
• People: train your team and give them approved tools; otherwise they’ll find their own.  
• Process: build review points into every workflow. Treat AI output as a first draft, check citations at the source, and tell the AI to leave a placeholder instead of guessing.  

And remember, you can’t hold AI accountable. Air Canada tried to argue its chatbot was responsible, and they lost (obviously). Your signature still promises due care, and that hasn’t changed.  

Build Automation Around Better Work 

The firms that get the most from AI will be the ones that treat automation as an operating decision, not a software purchase. Clear ownership, consistent processes, measurable outcomes, and defined review points give teams a stronger foundation for deciding where AI actually belongs. 

Used well, automation can reduce time spent on repetitive work and create more capacity for the work that still depends on people. The priority is not removing humans from the process, but making sure their time is used where judgment and experience matter most. 

As firms introduce more AI into everyday workflows, the environment supporting their core accounting applications matters too. Ace Cloud Hosting helps CPA and accounting firms explore AI without exposing local devices or giving agents unrestricted access to sensitive client data. Automate workflows, process documents, summarize information, and work with hosted accounting applications inside an isolated, managed workspace.