What it means
We begin by listening to your description of what is frustrating, time‑consuming, or error‑prone about the current way of working. The goal is to grasp the outcome you want to achieve, not to jump to a presumed solution.
RIGHT HAND SERVICES × AI
How the Approach Works
A step‑by‑step overview of the proposed method for understanding a problem, designing a workflow, and implementing it with human review and authorization at every critical stage.
AI recommends. JP decides.
STEP 1
Understand the problem
What you may provide
A clear description of the pain point, examples of where the current process falls short, and any goals you have for improvement.
What would be reviewed before proceeding
We will confirm that we have captured the problem accurately and that it is something we can help explore—without claiming we can solve it outright.
STEP 2
Identify the inputs and information sources
What it means
We map out what information or materials are needed to start the workflow and where they come from. This includes both the data you provide and any external sources we might need to access (with your authorization).
What you may provide
Details about the inputs you currently have—such as forms, emails, documents, or database fields—and any known external sources (e.g., a CRM, a property records website).
What would be reviewed before proceeding
We will verify that the inputs are sufficient to begin the workflow and that any external sources we plan to connect to are accessible and authorized for use.
STEP 3
Map the workflow
What it means
We break the work into discrete steps: trigger → inputs → processing → review/approval → authorized action → record of outcome. For each step we note whether it is rule‑based or AI‑assisted, and where human judgment is required.
What you may provide
Your own description of how you currently perform the work, including any unofficial shortcuts or checkpoints you rely on.
What would be reviewed before proceeding
We will ensure the workflow is complete, that all steps are logically ordered, and that review points are placed where mistakes could cause problems downstream.
STEP 4
Define permissions, limits, and approval points
What it means
For each step that involves an external system or data, we specify exactly what permissions are needed (e.g., read‑only access to a contact list, permission to draft but not send an email). We also set limits on what the AI assistant may do (e.g., summarize but not alter the original text) and mark where human review is required.
What you may provide
Any known constraints on data usage, security requirements, or preferences about which steps should be fully manual versus assisted.
What would be reviewed before proceeding
We will confirm that the proposed permissions and limits align with your policies and that all approval points are clearly marked before any action that could have external effects.
STEP 5
Develop and test a scoped implementation
What it means
We build a minimal version of the workflow that covers the core steps and uses mock data or sandbox connections where possible. This lets us verify the logic, the review points, and the flow of information without affecting live systems.
What you may provide
Access to test data, sandbox environments, or example inputs and outputs that we can use to validate the implementation.
What would be reviewed before proceeding
We will check that the implementation follows the mapped workflow, that all review points are functional, and that the AI assistant’s output is presented clearly for human judgment.
STEP 6
Review evidence and outputs
What it means
We run the scoped implementation with real (or realistic) data and examine the results. We look for accuracy, completeness, and whether the output matches what a human would expect given the inputs.
What you may provide
Real‑world examples of inputs and the expected outputs, or access to a pilot run in a non‑production environment.
What would be reviewed before proceeding
We will confirm that the output is useful, that any uncertainties are noted, and that the human review points are positioned to catch errors before they propagate.
STEP 7
Authorize appropriate use
What it means
Once we are satisfied that the workflow works as intended and the output is reliable, we present it for your formal authorization. This includes a clear description of what the system will do, what it will not do, and where your approval is required for each external action.
What you may provide
Your formal approval to proceed with the implementation in a live environment, along with any specific conditions or limitations you wish to impose.
What would be reviewed before proceeding
We will ensure that the authorization covers all steps that modify data or send messages, and that the limits and permissions are clearly documented.
STEP 8
Refine based on observed results
What it means
After the workflow has been running for a period, we collect feedback on its performance: Are the outputs still useful? Are any bottlenecks appearing? Have the inputs or external systems changed? We then adjust the workflow accordingly—always keeping human review and authorization at the core.
What you may provide
Observations from the live run, any new requirements that have emerged, and notes on what is working well and what could be improved.
What would be reviewed before proceeding
We will confirm that any changes to the workflow are evaluated for impact on the review points and authorization steps, and that the refined version still meets the core principles of human‑guided AI assistance.
LET'S TALK
Ready to explore how this approach could work for you?
Describe what you’d like to improve. We can start with the problem, not a sales pitch.