Choosing between rules and an LLM
Stable complete rules should remain deterministic.
See the decision applied to real work
Each scenario explains the expected disposition, preferred pattern, rejected alternatives, controls, and reassessment triggers.
Drafting an internal email
A low-stakes draft is easy for the sender to review before sending.
Generating marketing variations
Draft variations are reversible and straightforward for a marketer to review.
Summarizing supplied meeting notes
The supplied notes are the complete source and the summary is easy to compare.
Answering questions about current policy
Policy answers need the current authoritative version and traceable passages.
Answering questions over internal documents
Answers must be grounded in access-controlled internal sources.
Building a recurring report
A fixed pipeline can collect validated fields and use AI only for bounded narrative synthesis.
Calculating invoice totals
Validated arithmetic and tax rules are deterministic.
Classifying support requests
A bounded classifier can suggest routing while exceptions remain reviewable.
Creating medical or financial guidance
Personalized high-consequence guidance needs a qualified responsible professional and a narrower support task.
Extracting structured data from documents
A schema and deterministic validation can bound extraction.
Producing customer-facing claims
Public claims need substantiation and accountable publication review.
Publishing scheduled social posts
Drafting and scheduling can be separated so publication remains an explicit human-approved action.
Reconciling duplicate records
A bounded agent may compare a preselected queue but merges require preview and approval.
Researching current market information
Current claims require traceable sources and observation dates.
Reviewing a contract
AI may assist issue spotting only after the task is reframed away from legal decision ownership.
Running a multi-step browser workflow
Adaptive navigation may justify an agent only inside strict action and stop boundaries.
Screening job applicants
The v1 product boundary excludes AI influence over employment screening.
Summarizing sensitive support records
The task can proceed only in an approved environment with minimized records and qualified review.
Updating a CRM
A narrow tool can preview and write selected reversible fields.