Implementation Readiness · April 14, 2026 · 9 min read

What To Fix Before You Automate A Business Process

Automation enforces a process, it does not fix one. The seven pre-automation checks owner-led companies should clear before buying business process automation, plus what not to automate.

TL;DR

Automation does not fix a process. It enforces one. If the trigger, owner, evidence, exception path, and metric are unclear, automation makes the mess move faster. The best process to automate is repeated, visible, evidence-backed, owned, exception-aware, and measurable. Before buying business process automation, run the seven pre-automation checks. If a process fails them, fix the process first; automating it will scale the problem, not solve it.

What should we fix before automating a business process?

Run every candidate process through seven checks. Each one should have a clear, written answer:

1. Trigger: what event starts this process, and is it unambiguous? 2. Input and evidence: what information is required, and is it complete and current? 3. Owner: who is accountable for the outcome, not just the steps? 4. Output: what does a correct result look like? 5. Exception path: what happens when the input is missing, wrong, or unusual? 6. System of record: where does the truth live, and what may be changed? 7. Metric: what number tells us this process is working?

If a process cannot answer all seven, it is not ready for automation. It is ready for cleanup.

Why does automation fail in small and growing companies?

Automation fails because it is applied to a process the business has never actually defined. The process lives in someone's head, varies by who runs it, and depends on informal judgment. Automating that does not remove the variation; it freezes one version of it and runs it faster, including the errors.

Growing companies are especially exposed because processes that worked at low volume by individual effort break when they are encoded and scaled without an owner or an exception path.

When is simple automation enough?

If the rules are stable, the inputs are clean, and the work does not require judgment, simple automation is enough and AI is not required. Routing a form to the right inbox, creating a task when a stage changes, or sending a templated acknowledgment are rules, not reasoning. Do not add AI where a rule already works.

When does AI add value?

AI adds value when the work requires classification, summarization, drafting, or routing that a fixed rule cannot express well, and there is a human review point before anything customer-facing, financial, legal, or record-changing happens. AI prepares; the owner decides.

When does cleanup come first?

Cleanup comes first when the owner is unclear, the inputs are scattered or stale, or the process changes every time it runs. Automating an unowned process with bad inputs produces fast, confident, wrong output. Define the owner, consolidate the source of truth, and stabilize the steps before any tool is involved.

The pre-automation audit

Score the process against each line. Any "fail" means stop and fix that before automating.

Owner: is one person accountable for the outcome? Fail means do not automate yet; assign accountability first.
Inputs: is the evidence complete, current, and in one source of truth? Fail means clean up the evidence and system of record first.
Stability: do the rules hold from run to run? If stable and no judgment is needed, simple automation may be enough and AI is optional.
Work type: does it need classification, summarization, drafting, or routing? If yes and the process otherwise passes, an AI workflow may help, with a review point.
Risk: does it touch customers, money, contracts, or records? If yes, a human review point is required before the action regardless of the other scores.

A process that passes owner, inputs, and stability is ready for automation. A process that fails any of the three is ready for cleanup, not a tool.

Examples by workflow

Lead capture: automate only after the source of truth and routing owner are defined; otherwise leads are routed fast to the wrong place.
Proposals: drafting and compliance can use AI with an approval point; do not automate sending.
Onboarding: checklist tracking automates well once the steps and owner are fixed; it fails if onboarding differs per client with no baseline.
Reporting: recurring reports automate cleanly when the metric and data source are agreed; ambiguous metrics produce confident, misleading reports.

What not to automate

Do not automate a process with no accountable owner.
Do not automate decisions that change pricing, contracts, credits, or legal language without review.
Do not automate on top of scattered or stale inputs.
Do not automate a process that changes every time it runs.
Do not automate customer-visible commitments without a human approval point.

Implementation review checklist

Before buying automation, confirm in writing: the trigger, the required evidence and its source of truth, the accountable owner, the definition of a correct output, the exception path, and the metric that proves it works. If all six exist, automation will enforce a good process. If any are missing, fix that first.

Related workflow pages

Related field reports

Where to go next

The business process automation service page explains the workflow-first method, and the AI workflow automation and AI implementation services pages show how ADA scopes the first workflow. The AI readiness assessment helps you work through the seven checks. To run the pre-automation audit against a specific process, request an implementation review.

FAQ

Does automation fix a broken process?

No. Automation enforces and accelerates a process. If the process is unclear or unowned, automation scales the problem.

What makes a process safe to automate?

It is repeated, visible, backed by reliable evidence, owned by an accountable person, aware of exceptions, and measurable.

When should we use AI instead of simple automation?

When the work needs classification, summarization, drafting, or routing that a fixed rule cannot express well, with a human review point before risky actions.

What is the first thing to check before automating?

The owner. If no one is accountable for the outcome, fix that before any tool is selected.

What should never be automated without review?

Pricing, contracts, credits, refunds, legal language, and customer-visible commitments. AI can prepare these; a person should approve them.

References

NIST AI Risk Management Framework Core: Govern, Map, Measure, Manage: https://airc.nist.gov/airmf-resources/airmf/5-sec-core/
McKinsey State of AI 2025: https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
McKinsey: how organizations are rewiring to capture value from AI: https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-how-organizations-are-rewiring-to-capture-value

Research Standard

AI Deployment Authority briefings are built to help operators make deployment decisions, not to summarize the AI conversation.

For new briefings and major updates, we review the search landscape around the topic: current results, common vendor claims, buyer objections, related workflows, and the practical questions the top pages often leave unanswered. We then compare the topic against ADA's workflow framework: trigger, evidence, owner, review point, risk boundary, stop rule, and measurable result.

Some pages are more mature than others. We update the library as better examples, stronger source material, and clearer operating patterns become available.