The work should not exist
STOP or simplify before automating unnecessary reports, duplicate entry or obsolete approvals.
Coral Shark Lab · Seven Stop Conditions
When Should You Not Automate a Business Process?
Direct answer
Do not automate work that should disappear, changes every time, hides too many unusual cases, makes important decisions or is difficult to detect and reverse when wrong.
Seven Stop Conditions
“Automatable” is the wrong question
Almost any digital process can be automated in some way. The useful question is: what part of this work can move without losing control?
The objective is not maximum software involvement. It is better flow with explicit human decisions, exception handling and recovery paths.
Seven Stop Conditions
The seven stop conditions
The normal process is not stable
HOLD or narrow when every case is “it depends” and the dependencies are not explicit.
The exception rate is unknown
Measure representative cases; exceptions often determine whether automation removes work or relocates it into rescue.
A consequential decision is being handed over
Separate preparation from price, rights, disputes, commitments and other material judgement.
Failure is hard to see or reverse
STOP autonomous paths with unacceptable consequence; HOLD where fallback/rollback is weak.
A new system is added too early
Check stop/simplify, existing features and configuration before another provider or custom build.
Success or ownership is unclear
Define baseline, acceptance criteria, owner, exception process, ongoing cost and maintenance.
Seven Stop Conditions
Every HOLD needs an Evidence-to-Change-HOLD statement
MISSING: exception rate for non-standard quotes.
WHY IT MATTERS: scope depends on whether standard cases can be identified reliably.
MEASURE: classify recent quotes as standard/exception; record reason and detectability.
DECISION CAN CHANGE IF: routine preparation is consistent enough to separate from pricing judgement and exceptions stop before commitment.
“Not enough information” without this next step is not a useful recommendation.
Seven Stop Conditions
Consequence and reversibility are separate variables
A high-frequency task can still be a bad automation candidate if one failure creates a serious contractual, financial, safety or relationship problem.
Low consequence / reversibleGood place for controlled learning.
Low consequence / hard to reverseImprove recovery before autonomy.
High consequence / reversibleStrong approval and evidence still required.
High consequence / hard to reverseKeep decision/action human or stop.
Seven Stop Conditions
Synthetic invoice-chasing example
Routine overdue reminders may be appropriate when due/paid state and cadence are clear. Disputed invoices, payment plans, relationship-sensitive accounts and legal escalation are not standard reminders.
Outcome: NARROW / HOLD. Measure exceptions, define stop flags, inspect accounting-system built-in reminders and keep disputes/recovery escalation outside the routine automation.
Seven Stop Conditions
Synthetic enquiry-follow-up example
Acknowledge receipt can often be deterministic. AI may help interpret free text. Completeness can be checked by rules. Price, unusual scope and consequential promises should remain controlled.
Outcome: NARROW. Fast acknowledgement can move without handing the whole customer conversation to an agent.
Seven Stop Conditions
When “do nothing” is the best automation decision
Not automating can prevent unnecessary subscriptions, fragile integrations, hidden rescue work, custom code, false confidence and migrations the business never needed. “Do nothing” should still be evidence-based: the burden may be small, exceptions may dominate, risk may outweigh benefit or an existing feature may already solve it.
Evidence and claim boundary
What this page is — and is not.
Seven Stop Conditions is a Coral Shark decision framework, not a claim of scientific or industry consensus. Synthetic examples are illustrative, not customer proof. Where a material external fact is added later, it should be sourced to a current primary source and dated.
Unknown stays unknown. No industry-average saving, automation percentage or customer outcome is invented to make the case more impressive.
Related reading
Need this applied to one real job?
The Owner’s Brief applies the same discipline to one recurring job: what you told us, your estimates, our interpretation and what is still unknown stay separate. The answer can be stop, wait, start smaller or proceed.
Apply this to one job · £249 + VAT