The ESSA principle

ESSA stands for Eliminate, Simplify, Standardize, Automate. It is a four-gate sequencing discipline for preparing any business process before technology touches it, and the order is not a menu. It is a strict dependency chain: each gate must be passed before the next is allowed.

The governing idea fits in one sentence: you cannot automate what is not standardized, and you cannot apply AI to what is not automated. Most failed transformations break this rule. They reach for the tool first, a bot, a workflow engine, a model, and bolt it onto a process that is wasteful, inconsistent, and undocumented. The result is automating chaos: the bad process now runs faster and costs more to change. Automation and AI amplify whatever they touch, excellence and chaos alike, so the foundation has to come first.

Remove non-value-adding steps, approvals, handoffs, and reports outright. This is the most overlooked gate. Organizations routinely try to automate processes, or steps within them, that should not exist at all. Value is defined from the customer's perspective; anything that does not add it is a candidate for removal. The discipline is blunt: do not optimize what you should delete.

Reduce what survives. Combine tasks, cut decision points, collapse redundant reviews, and strip out the complexity that crept in over years of patching. Simplification comes only after elimination, so you never spend effort streamlining a step that should have been deleted.

Document the simplified process as the current best-known way of working: SOPs, standard work, visual checklists, explicit decision rules. This is the critical gate. Without a standard, automation has nothing to follow and simply locks in whatever inconsistency it finds.

The principle traces to Taiichi Ohno: where there is no standard, there can be no improvement. The digital-age corollary follows directly. Where there is no standard, there can be no successful automation.

Now, and only now, apply technology. Because the process is clean and consistent, automation runs reliably. Just as important, it generates structured, timestamped data as a byproduct, and that clean data is the fuel AI needs. Garbage in still produces garbage out, only at machine speed.

Two warnings hold the whole sequence together. Never automate a step you should have eliminated. Never standardize a process you should have simplified.

Consider a typical invoice-approval process inside a finance team. A walk of the actual work, not the documented version, finds eleven steps where the SOP claims three, duplicate approvals on low-value purchases, the same data re-keyed into three systems, and a true lead time of two weeks against a stated two-day target.

Run it through the gates. Eliminate: drop the duplicate approvals that protect against a risk smaller than the cost of the review itself, and retire a confirmation report nobody reads. Simplify: combine the three data-entry steps into a single capture and collapse two sequential reviews into one. Standardize: document the shortened process as standard work, with explicit rules for what gets auto-approved versus escalated. Automate: only here does a bot enter, routing standardized approvals and reading consistent fields. It works on the first try, because there is now exactly one documented way the work is done.

The payoff order matters. Systematic waste removal delivers meaningful efficiency gains before any technology is deployed. Automation layered onto the cleaned, standardized process then multiplies the result instead of cementing the mess.