Notes from the ledger

Every few weeks someone asks us to name the best accounting software. They usually expect a product name. What they get instead is a question back: best for what? That is not evasion. At small and mid-sized business scale, the leading platforms all do invoicing, bills, bank reconciliation and a profit and loss competently. Feature checklists will not separate them. Four other things will.

Statutory fit comes first, and it eliminates products outright

Before any discussion of features or price, ask whether the platform natively handles the compliance your entity is bound by. In India that means GST returns, e-invoicing with IRN generation, and e-way bills. In the UK it means HMRC recognition for Making Tax Digital. In the Gulf it increasingly means e-invoicing through an accredited provider. A platform that lacks these is not a contender with a weakness — it is not a contender. This single filter usually removes a third of the shortlist, including some very well-known names that simply do not operate in your market.

Compare the cost model, not the price

Headline monthly prices are almost meaningless on their own. One product charges per user tier and caps seats; another includes unlimited users; a third is a one-time perpetual licence with a small annual renewal; a fourth is per user per month across an entire ERP suite. These scale in completely different directions. Run the numbers at your real seat count over three years, not at the entry tier for one month. The ranking will often flip.

A five-user business and a fifty-user business rank the same products in opposite orders. The price list never tells you that.

Where does the data need to go?

Accounting software does not live alone. It has to exchange data with billing, payroll, point of sale, banking, expenses and often an inventory or CRM system. The quality of the integration surface — and whether there is a real, documented API behind it — matters more over three years than almost any in-product feature. A platform that is excellent on its own and isolated from everything else will cost you in re-keying, reconciliation and errors.

What happens when you need to leave?

Nobody asks this at selection time and everybody wishes they had. Vendors withdraw from markets, get acquired, change pricing, or simply stop fitting a business that has grown. Ask how your data comes out, in what format, and whether a competitor can import it. Treat data portability as a selection criterion, because one day it will be the only one that matters.

So what is the answer?

For a single-location Indian business that is cost-sensitive, a perpetual desktop licence with native GST is very hard to beat on total cost. For the same business with multiple locations or a remote finance team, a cloud platform with native Indian compliance wins. For a growing US business above five users, the unlimited-user model changes the maths entirely. For anything with real inventory, manufacturing or multiple entities, an ERP is the right shape — provided it is budgeted as an implementation, not a subscription.

The best accounting software is the one that clears your statutory gate, fits your cost model at your real size, connects to the systems you already run, and lets you leave cleanly. Name those four things and the product name usually follows on its own.

An Enrolled Agent is licensed by the US Treasury to represent taxpayers before the IRS — the same representation rights as a CPA or an attorney, earned through a tax-specific examination rather than a general accounting one. For a long time EAs sat inside larger firms doing the tax work while others owned the client relationship. That has changed. A steadily growing share of EAs now run independent practices, and the reasons behind that shift say a great deal about the profession as a whole.

Why practitioners are going independent

Three things converged. Cloud accounting and tax software removed the need for a firm's infrastructure — a sole practitioner can now run a book of clients on the same tools a mid-sized firm uses, at a fraction of the fixed cost. Remote work normalised the idea that your accountant need not be in your town. And clients themselves began to value a direct relationship with the person actually doing the work over the reassurance of a large firm's name on the letterhead.

The result is a practitioner who is more accessible, often more specialised, and free of the utilisation targets that push large firms toward standardised, high-volume work.

The ceiling every independent hits

Independence has a well-known problem: the practitioner is the product. Every hour spent on bookkeeping, reconciliations, month-end close and chasing documents is an hour not spent on the advisory and representation work that clients actually pay a premium for. Most independent practices plateau not because they lack clients but because the founder runs out of hours.

The practices that grow are the ones that stop doing their own back office.

Where the profession is heading

Put these together and the shape of the next decade becomes clear. Compliance work — the transaction processing, the reconciliations, the standard close — is becoming a commodity delivered by specialist operations teams with heavy automation. The value, and the fees, are concentrating in judgement: tax strategy, representation, business advice, and the interpretation of numbers rather than their production.

Independent EAs are well placed for this because they already sit at the judgement end. What they typically lack is a reliable, scalable operations layer underneath them. Increasingly they are getting it not by hiring but by partnering — with outsourced accounting operations that produce a clean, reconciled ledger on a fixed calendar, so the EA opens a trial balance rather than building one.

What this means for a client

If you work with an independent practitioner, ask how they handle the operations layer. A practice that has solved it will give you faster closes, cleaner books and more of their attention on your actual questions. A practice that has not will, sooner or later, be too busy to take your call in March.

The future of accounting services is not fewer accountants. It is accountants spending their time on the part of the job that needs a human, with the rest handled by people and systems built for exactly that.

Most industries have to work hard to find a use for AI. Accounting does not. The profession runs on structured data, explicit rules, and enormous volumes of repetitive decisions that are individually trivial and collectively exhausting. That is precisely the profile of work where machine learning performs best. The match is real. The question is what to hand over and what to keep.

Where AI is already doing the work

Categorisation. Deciding which account a transaction belongs to, thousands of times a month, was once a junior's entire job. Models trained on your own history now do it with high accuracy and flag only the genuinely ambiguous cases for a human.

Matching and reconciliation. Pairing bank lines to invoices, payments to bills, and inter-company entries to each other is pattern recognition at scale. It is the single largest time sink in most closes and it is almost entirely automatable.

Document capture. Reading a supplier invoice, extracting the vendor, amount, tax and line items, and drafting the bill — without anyone typing.

Anomaly detection. A duplicate payment, an expense that does not fit the pattern, a margin that moved when nothing else did. Machines are better at noticing these than tired humans at month-end.

Answering questions. This is the newest and, we think, the most important. Ask the ledger a question in plain language — why did operating costs rise, which customers are slowest to pay, what is driving the margin change — and get an answer with the numbers behind it, rather than a report to go and read.

The teams adopting this well spend their time on judgement. The teams that do not keep paying for keystrokes.

Where it should not be trusted alone

Judgement calls about revenue recognition, capitalisation, provisioning and the treatment of unusual transactions are not pattern-matching problems. They need an accountant who understands the standard, the business and the consequences. Controls — who can approve what, segregation of duties, the audit trail — are also not something to automate away; they are the thing that makes the automated output trustworthy.

And there is a quieter risk. AI applied to a messy ledger with a badly designed chart of accounts produces wrong answers faster and with more confidence. The technology amplifies whatever it sits on.

The order of operations

This is why we insist on a sequence. Get the chart of accounts right for reporting, not just for posting. Integrate the systems so data arrives once, correctly. Establish the controls. Then automate the processing, and only then put a natural-language layer on top. Done in that order, AI in accounting is not a novelty — it is simply how a modern finance function runs. Done in the other order, it is an expensive way to be wrong.

We are often brought in after a migration that did not deliver. The business moved from one accounting platform to another — usually a more capable one — and the close is no faster, the reports are no more useful, and the team is, if anything, more frustrated. The software was not the problem. It was never going to be the solution either.

What functionality actually buys you

Functionality is real and it matters at the margins. A platform with native e-invoicing saves a manual step. One with batch tracking suits a business that needs it. One with unlimited users removes a cost. These are genuine differences and they are why selection matters. But functionality is the floor, not the ceiling. Two teams on the same product with the same features routinely get completely different results from it.

What knowledge transfers — and why it is the scarcer asset

Double-entry does not change between vendors. Neither does the logic of a chart of accounts, the discipline of a close calendar, the purpose of a reconciliation, or the meaning of a trial balance that does not tie. Someone who understands these can sit down at an unfamiliar platform and be productive within days, because they know what they are trying to make the software do. Someone who has only learned where the buttons are on one product is lost the moment the buttons move.

Know the accounting and you can learn any software. Know one software and you are one migration away from starting over.

There is a second kind of knowledge that matters just as much and is rarer still: an understanding of how systems fit together. Where does the data originate, which system is the master for which record, what happens at each hand-off, and what breaks when one of them changes. This is not accounting knowledge and it is not product knowledge. It is the thing in between, and almost every failed implementation we see failed for lack of it.

So where should a business invest?

Pick a platform that clears your statutory and cost requirements — that is the functionality decision, and it is bounded. Then spend the real effort on the process: a chart of accounts designed for the reports you need, a close calendar people actually follow, controls that hold, and integrations that move data once. And put people on it who understand accounting and systems, not just the product.

A capable team on modest software will outperform a weak team on the best software in the market, every time. That is not an argument against good tools. It is an argument for being clear about what the tools can and cannot do for you.