If you look back at how much accounting work has changed, the clearest picture is not what the software looks like. It is where the data lives. In the first era, the data sat on a single computer, and anyone who wanted to change a number had to sit at that machine. In the next era, the data moved to the cloud, and the whole team could open it at the same time from anywhere. Now, that same data has started to answer questions.
I am Sutthinee Theppariyapol (Yim), Head of Product at FlowAccount. I walked through these three eras on the stage of DAC 2026 (FlowAccount's Digital Accounting Conference), and the Thai media outlet Future Trends has already summarised that talk. So this article is not a second summary. It is our view, as the team that builds accounting software, on the question that sat underneath the whole talk: where does the cloud-plus-AI era put the accountant?
Part 1Three eras of accounting, as seen by the people who build the software
People who build accounting software see each shift a little differently from the people who use it. We look at where the accountant's time went in each era, and where the next era moved that time.
The desktop era. The software was installed on the machine, and the work was tied to your own desk. Most of the time went on keying in data, then on sending files back and forth inside the team. One client's data lived on one computer. Anyone who wanted to look had to wait for whoever owned that machine.
The cloud era. The data moved to one shared place. The team could work at the same time and open the books from wherever they were. The time that used to disappear into waiting and merging files almost vanished. This is also the era when accounting data started to gather somewhere a system could reach at any moment. The closest example for us is our own system: over the past year, more than 213 million documents were created in FlowAccount (figure from FlowAccount's DAC 2026 summary).
The cloud-plus-AI era. The system no longer just stores the data for you. You can ask it questions and have it help with the analysis, and it gives time back to the people doing the work.
The first two eras gave accountants time back by cutting manual work: less keying, less waiting. The third era gives time back in a different way. It shortens the distance between a question and its answer. Before, if you wanted to know which debtors had been outstanding for a long time, you opened a report and worked through the numbers yourself. Now you can type the question in plain language and get an answer back.
Once a question becomes the door into the data, the advantage no longer goes to whoever keys fastest. It goes to whoever knows what to ask.
Part 2Why is this era different from the two before it?
Because this time the tools are already good enough. What AI lacks is not intelligence. It lacks two things that accountants hold: context, and fresh data.
Most conversations about AI and accounting still circle around how good the model is, how many questions it can answer, how often it gets things wrong. For those of us who build the systems, the question that actually decides the outcome is what the AI gets to read. I said it in one line on stage, and Future Trends picked that line up in its summary of the talk: "What AI needs is context and fresh, real-time data, and accountants already hold both." (Translated from Thai.)
The first is context. Here that means everything that is not in the numbers but gives the numbers their meaning. AI knows accounting principles better than most people. What it does not know is what your client sells, whether they opened a new branch this year, or whether the big debtor who pays late does so out of habit or because they are in trouble. If sales drop sharply this month, AI can tell you they dropped. It cannot tell you whether that is because the owner closed the shop for a holiday, or because a regular customer has started buying somewhere else. The accountant who has looked after this client all year knows. Sometimes they know before the numbers even come out.
The second is fresh data. That means data that moves with the real business: a transaction entered today is visible to a question asked today. Not a set of accounts closed three months ago. However smart the AI is, if it is given old accounts to read, it can only tell you about the day those accounts were closed. The questions business owners really want answered are usually about this week.
In my view, fresh data is not only something accountants hold. It is also one of their duties. The IFRS Conceptual Framework for Financial Reporting sets out several qualities that make financial information useful, and two of them are relevance and timeliness. Relevant information is capable of making a difference to the decisions people make. Timely information reaches decision-makers in time to influence those decisions, and generally, the older the information, the less useful it is.
But day-to-day work tends to pull accountants' attention towards getting the numbers right and filing financial statements and tax returns as required, so timeliness and relevance can slip out of view. With AI in the picture, the expectation of timely, relevant information only rises. Accountants have to adapt and find ways to cut the time bookkeeping takes, so that what comes out at the end is fresh data that decisions can rest on.
What people outside the profession tend to miss is that fresh, trustworthy data does not appear on its own. It comes from accountants entering transactions, reconciling balances and chasing documents from clients every single day. The thing AI needs most has been the accountant's work from the start.
So the cloud-plus-AI era does not pull power away from accountants. It moves power towards the people who hold both context and fresh data at once.
Part 3What accountants have not asked for, because it used to be impossible
Every year we talk to accounting firms, and almost everything we hear back falls into three requests: close the books faster, serve clients better, and make running the firm itself easier. All three are fair requests, and we have been working on them all along.
But there is a fourth, one that nobody asks for because it used to be impossible: being able to ask your own data a question directly. No firm has ever written to us saying "we would like to talk to our trial balance", because there has never been a world where a trial balance could talk back.
People who build products run into this every time. The thing that changes the game is rarely on the list customers ask for, because customers can only ask for what they can picture being possible. The job of the people who build tools is to hear every one of those requests, and also to see what has only just become possible this year.
FlowAccount AI Connector is our answer to that fourth request. Picture a standard plug: once it is plugged in, the AI you already use, such as Claude or ChatGPT, can talk to your data in FlowAccount. The plug uses a standard called MCP (Model Context Protocol), which is how AI plugs into other systems. Notice the word Context in the middle of the name. That is because the standard exists to hand context to AI directly.
Part 4AI can answer questions now. Can it do the work yet?
Today AI can already answer questions about accounting data, and read and analyse it. Doing repetitive work on your behalf is the next step, and to be straight with you, it is not fully here yet.
On stage we showed three groups of people using the same data but asking about different things (Future Trends describes all three cases in its summary of the talk). The business owner asked the question that nags at them every month: how much profit this year, and is the cash on hand enough to cover the next three months of expenses? The accountant asked for a half-year profit estimate. The auditor asked for a materiality calculation. Materiality is the threshold for how far a number can be off before it starts to affect the decisions of the people reading the accounts.
AI answered all three. The more interesting part is that the three questions came from three kinds of people, each of whom knew what they needed from the numbers. The auditor starts from materiality, because materiality sets how deep the audit goes. The accountant knows that by mid-year, the business owner needs to see certain things before planning the second half. Good questions come from people who know the context, and AI can only answer as well as the question it is given.
The examples we prepared for the stage were written purely in the accountant's voice. At closing, for instance: which debtors are more than 60 days overdue, and what is the total? Or: give me a summary page of the trial balance, with outstanding receivables and payables, for one company. People outside the profession would not think to ask these, because they do not know which numbers you look at first when closing the books.
The next step is automation. Month-end work that repeats every cycle, such as bank reconciliation or recurring adjusting entries, can be set up once and left to the system. On stage we told the room plainly that this part is still Coming Soon. So that day started with asking, reading and analysing.
In my view, that is the right order. Any work AI does on your behalf needs someone who can tell whether the result is correct. Until you can ask the right question and tell whether the answer holds up, AI should not act on its own.
Part 5Where does the accountant's position get stronger?
It gets stronger in the part AI cannot do for you: expertise, experience (what Thai speakers call "flight hours"), and professional judgement. I used those three terms at the same event, and FlowAccount carried them into its DAC 2026 summary.
I am not saying this to make anyone feel better. The reason is plain supply and demand. The easier answers are to get, the more valuable it is to have someone who can tell which answers hold up and which only sound good.
Now tie those three back to context and fresh data. Expertise is what tells you which numbers are unusual for a given kind of business. Flight hours are the context you build up from seeing similar cases many times over, until you know how this sort of thing usually ends. Judgement is deciding which numbers to trust, and whether to let a number leave your hands and reach anyone else. Together, these three are what turn fresh data into data that can be trusted, and what make the question you hand to AI the right question.
So the accountant's future does not depend on racing the machine for speed. It depends on how much of the time the machine gives back goes into these three things. If you want to see which skills gain value in real day-to-day work, read on at Which accounting tasks have already moved to AI, and which have not.
Part 6We build the tools. Accountants ask the questions
As the people who build the tools, our job is to make sure the data accountants look after can answer questions quickly and accurately. Choosing what to ask, when to ask it, and how far to trust the answer is the accountant's work, and we have no intention of taking that work out of anyone's hands.
If you take one thing back to work tomorrow, make it this question. Go through the list of clients at your firm and ask yourself: whose data is already fresh enough to ask AI about, and whose still has to wait for month-end?
The first group is where you can start right away. The second group is your homework: how to get those clients' documents to you sooner. That homework is the timeliness duty I described in Part 2. The answer to this question will tell you more clearly than any article how ready your firm is for this era.
AI will answer faster every year. But the questions worth asking will still have to come from the people who know each client best.
Sources and references
DAC 2026
- The author's talk on the DAC 2026 stage, titled Future of Work: using AI and technology to build the accounting firm of the future (given in Thai; covers the three eras · the fourth request nobody has made · the accountant's example questions · automation still marked Coming Soon)
- FlowAccount's DAC 2026 summary (its section 4, and the 213 million documents figure) https://flowaccount.com/blog/flowaccount-dac-2026-key-takeaways/
- Future Trends summary of the talk, compiled by ธนพนธ์ หัสกรรัตน์ https://www.facebook.com/photo/?fbid=1496469962514130&set=a.630751835752618 (source of the three demo cases in Part 4 and the line quoted in Part 2; the text of this article is written in our own words)
FlowAccount AI Connector
- Launch article https://flowaccount.com/blog/flowaccount-ai-connector-mcp/
- Help centre https://flowaccount.com/help-center/category/ai-connector-mcp
IFRS Conceptual Framework
- IFRS Conceptual Framework for Financial Reporting (2018), paragraphs 2.6 (relevance) and 2.33 (timeliness) https://www.ifrs.org/content/dam/ifrs/publications/pdf-standards/english/2021/issued/part-a/conceptual-framework-for-financial-reporting.pdf
Read next
- Which accounting tasks have already moved to AI, and which have not https://productize.life/blog/ai-replace-accountants/en
- About the author https://productize.life/en/profile