General Purpose

The pitch

When the accounting and finance function strains, the answer arrives in one shape: a bigger system. One platform to hold the ledger, the reporting, the payroll, the grants, the board packet, and everything between. The pitch is completeness. The whole function, one vendor. Twenty years ago the shape was called an ERP. This year the shape is called AI, pointed at the whole function at once, and the pitch has not changed a word.

What it costs

Gartner's own research predicts that by 2027 more than seventy percent of recently implemented ERP initiatives will fail to fully meet their original business case goals, and as many as a quarter of those will fail catastrophically. That is the vendor-neutral read on the completeness pitch. The implementation runs long, the budget runs over, the organization bends its processes to fit the system, and the system is revisited the next time the organization outgrows it. The cost is not the license. The cost is the years spent conforming to a general-purpose design, paid again at every upgrade cycle.

The same pitch, new engine

A model with the whole function in its window and no assigned duty answers everything and accounts for nothing. What it read, what it changed, and what it relied on are not stated. The books leave the building to be reasoned about somewhere else. The output arrives fluent, unsourced, and unchecked. None of that is a flaw in the technology. It is a flaw in the scope.

Where the profession already ruled

Accounting settled this argument centuries before software existed and named the settlement segregation of duties. No one person holds cash, records, and reconciliation, not because people are bad, but because unlimited scope defeats the check. Authority is split so that error has somewhere to surface. The principle was never about people. It is about scope, and it applies to software with no modification at all.

The lens

A tool deserves the same terms a new hire gets. One duty, stated. Sources cited, so the work can be tied back. A boundary on what it sees, so confidentiality is a property of the design rather than a promise. A second look built in, because the check is the control. An instrument built this way is narrow on purpose. It carries one duty completely, states where every figure came from, sits beside the ledger it does not touch, and can be checked by anyone with the source in hand. As the technology advances, the function absorbs it the same way it absorbs any capable junior: one assigned duty at a time, supervised, cited, and checkable. The people running it stay in charge of it, and no part of it requires a programmer.

Where this goes

AI is not a threat to this design. It is the labor the design was waiting for. When the models improve, an instrument puts the improvement to work inside its one duty: the commentary reads sharper, the flag arrives earlier, a duty that needed a person yesterday runs on its own tomorrow. The duty, the sources, and the check do not move. Nothing gets reimplemented and nobody retrains. A general-purpose system delivers the same advance as another project. The organizations that gain the most from what is coming will be the ones whose function was already arranged for it: the ledger theirs, the duties encoded one at a time, every figure cited, every output checked. That arrangement is Native Constraint.

The four questions

Four questions expose the design of any tool in front of you. What single duty does it carry. What does it cite. What does it see. Who checks it, and against what. A general-purpose system has no clean answer to any of the four. A well-built instrument answers all four in a sentence. The future of the function is not one system that does everything. It is a ledger that stays yours, surrounded by narrow instruments that each do one thing to a standard, and the standard is the point.

Source: Gartner, Enterprise Resource Planning insights, gartner.com, read July 2026.

← The Casebook