Sales-tax compliance is one of the few back-office workloads whose size can be traced to a specific date. Before June 2018 a seller generally collected tax only in states where it had a physical presence, so a company shipping nationwide from one warehouse filed in a handful of states. The Supreme Court's decision in South Dakota v. Wayfair, Inc. changed the rule: a state may now require an out-of-state seller to collect tax based on economic activity alone, measured by sales volume or transaction count. Forty-five states plus the District of Columbia have since set such thresholds. The same shipment of goods that once created a filing obligation in two or three states can now create one in twenty, each with its own forms, rates, and due dates, which is why the work has grown while the tools used to do it have not.
That growth has not been matched by automation. A 2017 Wakefield Research survey commissioned by Avalara, of more than 400 US finance and accounting professionals at companies with annual revenues of $10 million or more, found that the average company dedicated six accounting professionals to sales and use tax, that the work took thirty-nine hours in a typical month, and that eighty-eight percent of companies did not use cloud-based automation to file and remit. The same survey found that seventy-five percent of the professionals believed a state auditor would find mistakes in their returns. Those two findings sit together for a reason. A process that consumes most of a week of senior accounting time every month, and that the people running it expect to be wrong, has had both its effort and its accuracy drift away from the judgment it nominally requires.
Where the hours actually go
It is tempting to assume the hard part of sales tax is the rate. Avalara maintains more than 900,000 tax rules across more than 12,000 US sales-and-use-tax jurisdictions, and its 2017 report with Wakefield Research counted close to 14,000 changes to US rates and jurisdictions in that year alone. The numbers are real, and the variation is genuine: a single delivery address can sit inside a state, a county, a city, and a special district that each levy their own rate, and whether a given product is taxable can turn on its category and on the destination. But rate determination is the part of the workload that already submits to a lookup. Once an address is resolved to its jurisdictions and a product is mapped to a taxability category, the rate is a table read, and tables can be maintained centrally and applied at the point of sale.
The labor lives downstream, in the filing. A company with economic nexus in twenty states is not filing twenty identical returns on one schedule. It is tracking a fragmented calendar where one state wants a return monthly and another quarterly, where the same state can move a high-volume seller from quarterly to monthly filing, where due dates fall on the twentieth in some states and the last day of the month in others, and where each return has its own form, its own way of breaking revenue down by local jurisdiction, and its own rules for prepayments and credits. The thirty-nine hours a month are spent assembling that breakdown, reconciling collected tax to the general ledger, populating each state's form, and not missing a date. None of that is rate lookup.
What the redesign mechanizes
The redesigned workflow does not start by replacing the accountant. It starts by separating the mechanical majority of the work from the small share that needs a person, and it runs in four stages.
The first is taxability and rate determination by jurisdiction. Each transaction's destination address is resolved to its full stack of jurisdictions, the product or service is matched to a taxability category, and the applicable rate is applied. This is the stage most existing tools already address, so the redesign's contribution is mainly to supply it cleanly from the order and invoice systems rather than from a monthly export.
The second is return preparation. For each state where the company has nexus, the agent assembles the return the way that state expects it, aggregating taxable and exempt sales and breaking the tax down to the local jurisdictions the form requires. Because order systems describe the same sale in inconsistent ways, this stage runs into the long tail of document and record variability that defeats brittle rules, which is the reason a model that reads context, rather than a fixed mapping, does the assembly.
The third is the filing calendar. The agent maintains, per state, the filing frequency, the due date, and the prepayment rules, and it tracks the frequency changes states impose as a seller's volume crosses their internal bands. This is the stage that consumes the most human attention today and the one where mechanization pays back most directly, because missing a schedule produces a penalty rather than an accounting correction.
The fourth is exception routing. When a return will not reconcile, when a new state crosses its nexus threshold, when a customer's exemption certificate is missing or expired, or when a product's taxability is genuinely ambiguous, the agent stops and routes a short, ranked list to a tax professional with the supporting detail already assembled. The same pattern governs a redesigned month-end close, where the agent does the mechanical bulk, surfaces the few items that need a person, and leaves the judgment with the person who can sign for it.
Why a person stays responsible
A tax position is a legal claim about what is owed, and responsibility for it cannot be delegated to software. Whether a particular product is taxable in a particular state, whether a marketplace-facilitator rule shifts the collection duty to a platform, whether a transaction qualifies for an exemption: these are judgments a tax professional signs, and the Wakefield finding that three quarters of the professionals surveyed expect a state auditor to find mistakes is a reason to keep that signature rather than remove it. What the redesign changes is the share of the professional's time spent on the positions rather than on the assembly. The agent does the work that is the same every month and brings the exceptions forward; the professional decides the exceptions and owns the filing. This is the structure of an aligned engagement in general, where the human keeps the accountable decisions and the agent absorbs the volume.
Filing also requires getting the return into each state's system, and many state portals offer no usable programmatic interface, which puts this work in the same category as writing back into a legacy system with no API: the agent must operate the state's own filing surface the way a person would, under the same controls.
A reasonable counter
A reasonable counter is that sales-tax software already exists, that the large vendors have sold rate engines and return-filing services for years, and that the thirty-nine hours reflect companies that have not bought those tools rather than a gap the tools cannot close. There is something to this. The rate-determination stage in particular is well served by existing engines, and a company still doing rate lookup by hand has an easier problem than it thinks. But the survey that produced the thirty-nine-hour figure also found that eighty-eight percent of companies were not using cloud-based automation to file and remit, years after those products shipped, which suggests the binding constraint is not the absence of a rate table. It is the work of fitting a general tool to one company's order systems, product catalog, and exemption records, then keeping the fragmented filing calendar current as nexus spreads. That fitting and upkeep is the part that has stayed manual, and it is the part the redesign is built to take on.