Sampling was a constraint, not a preference. We run testing across the full population — journal entries, duplicates, Benford's Law, trend analytics — and a CPA still signs the opinion.
Request an Audit ProposalThe Sampling Problem
For most of the history of the profession, testing a population meant pulling a few dozen items out of it and hoping they were representative.
Sampling was never a preference. It was a constraint. Nobody could read 180,000 journal entries by hand, so the standards were built around selecting enough items to draw a reasonable conclusion. That constraint produced a real limitation everyone quietly accepted: the item that mattered might simply not be in the sample.
Analytics removed the constraint. Reading the entire population is now a computational problem rather than a staffing one — which means the interesting question is no longer which items did you pick but what did the whole set look like.
On engagements where the data supports it, we run our testing across every transaction rather than a selection:
The Platform
We run engagements on Audit Systems (auditsystems.app), a San Antonio-built audit engagement platform.
Planning through issuance runs as phase-gated workflow. Work cannot skip ahead of the procedures it depends on, which is exactly the discipline peer reviewers look for.
Business, nonprofit, Single Audit, 401(k) and employee benefit plan, government and municipal, PCAOB, SOC 2 and more — each with its own checklist rather than a generic one bent to fit.
Compliance checklists tied to GAAS, PCAOB standards and Uniform Guidance, so the requirement and the evidence sit next to each other in the file.
Preparer, reviewer and partner sign-offs, with a full activity log. When someone asks who checked a number and when, the file answers.
Working papers organized, client documents collected, every file versioned and cross-referenced. No more chasing the current version of a schedule through email.
Encryption, row-level access controls, two-factor authentication and activity logging. Client data stays inside the platform.
What Doesn't Change
This is the part firms selling AI audits tend to skip, and it is the part your board should care most about.
The AICPA's position is that AI does not absolve a practitioner of professional responsibility. The right way to think about an AI tool in an audit is as staff work: useful, fast, and requiring review before anyone relies on it. Professional skepticism still applies, and automation bias — accepting a machine's answer because it came from a machine — is treated as a risk to be managed, not a feature.
So the analytics do not replace the audit. They change which items a person spends time on. Instead of examining forty entries chosen by a formula, the engagement team examines the entries the analysis actually flagged, and has evidence about the rest of the population instead of an inference.
What that means in practice on your engagement:
If you are engaging us for an audit, we do not also build or maintain your accounting records. That is not a preference, it is what independence requires.