Automation Scope Triage (Onboarding)
A checklist with 7 steps: Classify the client workflow by rule density before naming any platform.
By InnovaAI ResearchPublished
What are the steps?
Automation Scope Triage (Onboarding)
- 01
Classify the client workflow by rule density before naming any platform
Score each candidate process on how many decisions follow a written rule versus human judgment. Anything below roughly 80 percent rule-based goes to a discovery backlog, not a bot build.
- 02
Map every system the workflow touches and whether an API exists
UI-level automation is the fallback, not the default. Where a documented API covers the step, price the integration; where it does not, budget for selector maintenance on that surface.
- 03
Pick one platform tier for the engagement and write the choice into the statement of work
Compliance-heavy clients with audit trails and role governance point toward an enterprise suite such as Automation Anywhere. SMB work with narrow, prompt-shaped tasks fits lighter tools like BrowserAct or Warmwind. Straddling both tiers in one retainer splits maintenance attention and erodes margin.
- 04
Set a minimum billable threshold per automated process
Estimate hours saved per month multiplied by the client's loaded labor rate, then compare against build plus 12 months of upkeep. Processes under the threshold become a bundled add-on rather than a standalone line item.
- 05
Document the human fallback for each step the bot owns
Name the person, the trigger, and the manual path when the automation fails. Client staff are already requesting exemptions from AI-assisted work, so a written human-only route prevents a delivery stall.
- 06
Scope permissions to the minimum each bot needs to complete its defined task
Read-only where possible, no shared admin credentials, and a named owner for every connected account. Agents that inherit broad access are the documented failure mode behind recent containment breaches.
- 07
Agree the success metric and the review date in the kickoff document
One primary metric (cycle time, error rate, or hours reclaimed) plus a 30-day checkpoint. Without a dated review, automation engagements drift into unbilled support.