Pick a business outcome and boundary
Do not audit “the whole company” as one process. Choose a useful boundary such as lead intake through assigned follow-up, invoice receipt through approved record, customer request through resolution, or recurring report through review.
Name the beginning, the finish, the people involved, and the business outcome. This prevents the exercise from becoming a software inventory with no connection to completed work.
Observe the real path
Interview the people doing the job, watch representative examples, and inspect the tools where the work actually happens. Include email folders, personal spreadsheets, copied notes, browser tabs, text threads, and manual checks. The unofficial path often contains the most important rules.
For each step, capture the trigger, input, action, system, decision, output, wait state, owner, and exception. Mark where information is retyped, searched for, reconciled, or lost.
Separate rules from judgment
Rules can be stated and tested: required fields, approved sources, service areas, routing logic, dates, status changes, and permissions. Judgment depends on context, negotiation, trust, or authority. A good system can prepare a judgment without making it.
Identify actions that are reversible and actions that are not. Drafting a reply is different from sending it. Preparing a payment record is different from moving money. These distinctions shape safe automation.
Rank readiness, not excitement
A promising workflow has a stable enough path, available inputs, a clear owner, a known source of truth, and an observable finish. A frustrating task may still be a poor first candidate if nobody agrees on the process or the required data is unreliable.
Recommend the simplest intervention that solves the problem. That may be better instructions, a form, a website edit, a database cleanup, a connection between tools, a fixed automation, or an agent with approved decisions.
Deliver an operating plan
The audit output should state the current map, proposed design, systems touched, permissions, exception handling, human approvals, test plan, maintenance owner, and sequence. It should also state what not to automate yet and why.
Ridgeway uses this approach across business use cases and can continue into a one-time build or ongoing operation. A useful audit creates clarity before it creates a technology project.
Collect evidence, not only opinions
Use representative work items to check the interview map. Follow a recent lead, invoice, customer request, or report from trigger to finish. Record elapsed waits, handoffs, re-entry of data, searches, corrections, and tools used. Sample ordinary work and difficult exceptions; either one alone gives a distorted picture.
Separate observed facts from estimates. If nobody can report volume, rework, or unresolved items reliably, note the measurement gap rather than inventing a baseline. The first intervention may be a status field or exception queue that makes the process visible enough to improve.
Score candidates with a common rubric
Evaluate each workflow on business importance, frequency, clarity of rules, input quality, system access, reversibility, exception rate, security sensitivity, and strength of ownership. Use the rubric to compare opportunities, not to produce a false precision. Written reasons matter more than a total score.
A strong first candidate has a visible finish and a safe fallback. A high-value process with unclear authority may require policy work first. A repetitive task with unreliable inputs may require data cleanup. A rare but painful task may fit a checklist or one-time tool instead of ongoing automation.
Turn the map into a testable scope
Define normal cases, edge cases, forbidden actions, required approvals, and the evidence that proves completion. Include what happens when an integration is unavailable, a record cannot be matched, required information is missing, or a person rejects a proposed action. These examples become acceptance tests instead of surprises after launch.
Sequence the work around dependencies. Fix ownership and source data before automating downstream messages. Connect one channel before coordinating several. Run in shadow or review mode before expanding authority. Assign a decision owner and a review date to every recommendation. The audit is complete when the team can decide what to build, what to change first, and who will keep the result correct.
Use the browser-based AI Automation Fit Check for a quick first screen, then review the AI Opportunity Audit when the business needs a documented scorecard, ranked opportunities, and action sequence.