Every tool your company runs is a filtered view of the same thing: the data that moves the business. We build the layer underneath — systems that run on their own, surface the one decision that actually needs you, and take it from there.
Part of Acutant Services IndiaTwenty systems, Bengaluru
Different problems, identical structure: the data moves on its own, the person is called in once, and the work goes back to the machine to finish.
For thirty years we built a screen for every operation a business performs, because every operation needed a person standing at it. That constraint is gone, and almost nobody has redesigned around its absence.
An invoice is a row. An approval is a field. A dashboard is a query somebody froze into a page and put a login in front of. Underneath every tool a company runs is one body of data, and the tool is a filtered view of it — useful, but never the thing itself.
The system runs unattended. When it reaches something only a person can settle, it surfaces exactly that — one decision, in a view built for that decision and nothing else — then takes the answer and finishes the job. The screen shrinks to the size of the judgment.
Read them bottom to top and they are one architecture: get the data out of the paper, move it to where the business keeps its record, give people a surface for the part that needs them, and run the whole thing somewhere that does not fall over.
Invoices, bank statements, ledgers, scans. None of them are documents in any sense that matters — they are records that happen to have been printed. We take the costume off: structured fields, a confidence score on every one, and a flag on the handful a person should actually look at. Not a black box with an accuracy claim.
Zoho to Tally. Airtable to Tally. A form to the agent that provisions the tenant. Everyone builds the interesting middle; almost nobody builds the boring end, which is where the data actually has to land.
“Where are we on this?” is a question the data can already answer. A front door is a query with identity and permissions attached — not a second copy of the record, and not a sync waiting to go wrong.
A system that only works while a person is watching it is a person’s job with extra steps. Containers, deployments, identity, monitoring, the asset and drive registers nobody thinks about until a migration — and something that notices when a thing stops, before a client does.
This list used to be thirty-seven, and about half of it was scaffolding — backup scripts, config repositories, a forked OSS tool, the same idea attempted twice. Counting those as delivered systems flattered us and told you nothing. What is left is work we would put in front of you. Each one has a write-up; open any of them.
Every application a company runs is a filtered view of one thing: the data that moves the business. An invoice is a row. An approval is a field. A dashboard is a query somebody froze into a page and put a login in front of.
For thirty years we built those views and called them systems. They were not systems. They were windows — and we built a new one for every operation, because every operation needed a person standing at it.
That was the real constraint, and it was so total that nobody thought to name it. A person had to read the invoice, so somebody built a screen for reading invoices. A person had to route the approval, so somebody built a queue. A person had to notice the exception, so somebody built a report, then a dashboard, then an alert about the dashboard. The sum of those surfaces is what the industry sells as software. It was never the work. It was scaffolding around the work, shaped entirely by the fact that only humans could do any of it.
That constraint is gone, and almost nobody has redesigned around its absence.
The work can now happen where it always belonged — in the data, continuously, unattended. What is left for a person is what was always genuinely theirs. Knowing that this vendor is in dispute. Deciding that this month is an exception. Carrying the context that lives in one head and no database.
So we build the other way round. The system runs. When it reaches something only a person can settle, it surfaces exactly that — one decision, in a view built for that decision and nothing else — then takes the answer and finishes the job. No inbox. No module. No thirty-field form where four fields mattered.
The screen shrinks to the size of the judgment.
Which means the loop is what we are designing, and the interface is only one step of it.
A position is cheap. These are the four places it costs us something.
A conventional tool is finished when a person can perform every step in it. Ours is finished when a person has to perform only the steps that need them — and when the handoff in both directions carries enough context to be trusted. Varta exists because we could not find a good example of that second part, so we had to draw one.
If the machine does the work, the interface has one job left: make the person’s part legible and their decision reversible. So every write-up in the archive carries a line reserved for naming where a person is still accountable — and audit, in the console at the bottom of this page, counts how many actually fill it and lists the ones that do not. Currently one does not.
The failure mode of an automated bridge is not an error, it is a file that imports cleanly into the wrong ledger and is found at a reconciliation three months later. So our bridges stop. An unmapped ledger or an unbalanced voucher is a hard refusal with the row named, and a person resolves it once, permanently.
Three stages of the Extractor / LedgerLink path with sample data, including the moment the work stops and waits for a person. Runs entirely in this browser tab — no model is called and no accounting system is touched.
Acutant Labs is the product arm of Acutant Services India Pvt. Ltd., an accounting and compliance practice in Bengaluru. That lineage is the whole advantage. We are not technologists guessing at finance — we are the people who did the work by hand, which is why we can tell you which parts of it never needed a person at all.
Nobody sat down and decided software was obsolete. We just kept building tools, noticing that the good ones had almost no interface, and that the part people valued was never the screen — it was the moment the thing came back finished. After about twenty of those, the pattern was hard to keep calling a coincidence.
Meet Acutant ServicesAku is the lab’s agent collaborator, and the lab’s own work moves through exactly the loop we sell: an agent picks it up, a person is pulled in for the calls only they can make, and it goes back. The Instrument is the board that runs it. If the model did not survive contact with our own Tuesday, we would not be putting it on a website.
What we commit to on every engagement, stated plainly enough to be held to.
Client data lives in client-controlled infrastructure wherever the engagement allows it. Where a model provider is involved we name which one, for which step, and why — and where the documents are not permitted to leave the building, we build the same pipeline against a model that runs inside it.
No invented logos, no borrowed case studies, no uptime figures. The archive is open instead. The write-ups behind each system are ours, not extracts from internal documents — and the console on this page runs an audit that grades them against the standards above and prints where they fall short.
You will talk to the people building the thing. That is a genuine constraint on how much we take on, and we would rather say no than staff a project we cannot do well.
We will not automate a decision because automating it would be convenient. Some processes are broken rather than slow, some are too rare to be worth a system, and some end in a number a regulator may ask about — which means it needs a name attached, and the name has to be a person’s.
A workflow that eats a day a week. A tool that will not talk to the tool next to it. A process where somebody re-keys what another system already knows. Or an idea you want pressure-tested before you spend real money on it. Write it down and we will take it from there.