Skip to content
ChamaPro is live. View the product

The problem

Repetitive checking, matching, summarising or routing consumes time, but errors still need accountable review.

This is a fit when

  • • The task repeats
  • • Examples and review criteria exist
  • • Failure can be detected and handled

Call the boundary early

This is not the right engagement when

  • Errors cannot be reviewed or corrected
  • The desired decision cannot be explained
  • The task is rare enough that manual work is safer

What the scope contains

01Classification and routing
02Drafting and summarisation
03Matching and anomaly review
04Rule-based operational automation

Deliverables

  • • Automation opportunity and risk map
  • • Evaluation examples and acceptance thresholds
  • • Human-review workflow
  • • Monitored automation inside the product

Explicit exclusions

  • • Unbounded autonomous decisions
  • • Guaranteed model accuracy
  • • Automation without an accountable fallback

How the work runs

Artifacts before activity.

The method is judged by the decisions and working records it produces, not by a generic phase label.

View the complete method
frame

Risk and opportunity map

model

Evaluation set

prototype

Review prototype

build

Fallback and monitoring plan

Technology and stack

  • • Use rules before models where rules are sufficient
  • • Keep provider choices replaceable
  • • Measure the workflow outcome, not model novelty

Specific technologies are named only after the operating constraints and approved public stack are verified.

Quality and operation

  • • Human review at consequential decisions
  • • Traceable inputs and outputs
  • • Defined failure and fallback paths
  • • Evaluation against representative examples

Integrations

Integrations are scoped against your existing systems, provider requirements and operational controls.

Questions

Questions about ai and automation

Where should a person stay in the loop?

At decisions with meaningful financial, legal, safety or relationship consequences, and wherever confidence or evidence is insufficient.

Do all automations need AI?

No. Deterministic rules are often easier to understand and operate. Models belong where they solve a bounded task better.

Next record

Scope a automation engagement without guessing.

We start by reviewing the process and confirming fit. If the work is clear, the next step is a written scope and quote.