01
Volume
How many times does the task happen each day, week, or customer cycle?
Process selectionDiscovery guide
How to Identify the Best Business Process to Automate
The best first automation is rarely the flashiest idea. It is usually a bounded handoff that happens frequently, follows a recognizable pattern, and creates a visible cost when it is delayed or missed.
Decision snapshot
Inventory repeated work, then rank each candidate on volume, rule clarity, exception rate, business consequence, data access, and ownership. Start with a high-frequency process that is easy to reverse and easy to measure—not the process with the most impressive AI demo.
Frequency
High
The task repeats enough for improvement to compound.
Rules
Visible
People can explain how the normal path works.
Risk
Bounded
Mistakes are detectable, reversible, or approval-gated.
Outcome
Measurable
The team can compare speed, effort, quality, or completion.
Working framework
The VALUE scoring model
Give each candidate one to five points on these dimensions. The score starts a conversation; it does not replace a risk review.
02
Agreement
Does the team agree on the correct normal path and definition of done?
03
Loss from delay
What happens when the handoff waits, fails, or depends on memory?
04
Usable inputs
Are the required messages, fields, records, or files available consistently?
05
Exception control
Can uncertain or sensitive cases be detected and routed to a human?
01
Build an inventory from the work, not the software
Do not begin with a list of tools you want to use. Begin with moments where information changes hands: a form becomes a lead, an email becomes a task, a signed contract becomes a project, or a meeting becomes a set of assigned actions.
Ask each team member what they copy, chase, reformat, check, or remind others about. Those verbs reveal the operational work that is easy to overlook in a process diagram.
- Copying information between systems
- Checking whether required information has arrived
- Routing requests to an owner
- Drafting similar replies or summaries
- Creating the same folders, projects, tasks, or reports
- Following up because a status has not changed
02
Separate automation value from AI value
Many useful workflows do not need AI. If a rule can reliably map a form field to a CRM owner, use a rule. Add AI only when the step involves unstructured language or limited judgment, such as extracting intent from an email, summarizing a call, or drafting a response.
This separation reduces cost and makes the workflow easier to test. Deterministic steps should remain deterministic; AI-assisted steps need confidence thresholds, examples, and a review path.
A strong design uses the least complex mechanism that reliably completes each step.
03
Choose a thin slice of an important process
A first project should matter, but it should not attempt to automate an entire department. Choose one trigger, one normal path, and one measurable outcome. For lead operations, that could mean acknowledging web inquiries, creating a clean CRM record, and assigning an owner—without automating sales conversations.
This thin-slice approach exposes real data and exception issues while keeping failure recoverable. The next phase can expand only after the first workflow behaves as expected.
Candidate interview questions
Use these questions with the person who performs or owns the work today.
- 1What event tells you this process should start?
- 2How often does it happen and when does volume spike?
- 3Which steps are always the same?
- 4Which decisions require context or judgment?
- 5What information is commonly missing?
- 6Where do mistakes or delays become expensive?
- 7Which systems contain the source of truth?
- 8What would a successful automated result look like?
- 9Who should be notified when the workflow cannot continue?
- 10How could the team stop or reverse the workflow safely?
Continue from here
Choose the next useful step.
Move from understanding the problem to evaluating the business case or scoping a concrete implementation.
FAQ
Important questions, answered directly.
Should we automate the task that takes the most time?
Not automatically. A long task may be rare, ambiguous, or high-risk. Frequency, consistency, data access, and recoverability can make a smaller task a better first project.
Is data entry always a good automation candidate?
Structured transfer between known systems often is. Data entry becomes harder when the source is inconsistent, duplicate records are common, or the target system has no reliable unique identifier.
Should we automate a broken process?
Document and simplify it first. Automation will make unclear rules run faster, not make them correct.
How do we compare two similarly scored workflows?
Prefer the workflow with lower downside, stronger ownership, cleaner data, and a shorter path to measurement. Learning value matters in a first project.
Bring one workflow, not a transformation pitch.
Describe the trigger, tools, repeated work, and where the handoff breaks. You will get a direct view on feasibility and the most sensible next step.
Request a workflow audit