AI agent implementation, starting with one business process.
A Lummy implementer helps you choose a workflow, connect your systems and test your AI agent with real cases. Your team defines what it should handle and reviews the results. You decide when it goes live.
Test cases · Orders and deliveriesExampleIllustrative test run · Draft version 8
Invoice number instead of order number: correct order found.
Order without a delivery route: no delivery date promised.
Credit request: handed to the sales team.
Review test results before publishing.
Resolve the decisions that hold up an AI project.
- A manageable first workflow
Start with one task, such as answering order-status questions, and agree on how to measure the pilot.
- Access to the right data
Connect Odoo or your documented API and enable only the operations the agent needs.
- A way to check answers
Test real questions before launch and review the activity log to see what the agent queried or changed.
From discovery to a live AI agent in four steps.
- 01Define the pilot
Choose the task, the systems involved, the actions that require confirmation and the cases that should go to a person.
- 02Connect your systems
Your implementer configures the channels and access to Odoo or your API. Someone on your team who knows the system helps provide access.
- 03Test real cases
Try normal requests, missing information and out-of-scope questions. Agree on a minimum test threshold before publishing.
- 04Launch and improve
Supervise conversations and actions from Lummy. Flag answers for correction; each correction becomes a test case for later versions.
- Start with one workflow and expand after reviewing results
- Pause an agent and request corrections
- Review versions, conversations and system actions
- Act outside the permissions you define
- Publish a version that fails its agreed tests
- Replace your team's decisions on out-of-policy cases
Who does what during the pilot.
You bring knowledge of the process and access to the relevant systems. Lummy handles the agent configuration and testing with you.
- Your Lummy implementer
Configures connections, prepares tests and applies corrections. The implementer's access has an end date.
- Your supervisors
Review conversations and results, identify answers that need work and decide when the agent is ready.
- Your administrator
Invites team members, manages connections, pauses agents and reviews billing.
What should we prepare before implementing an AI agent?
Bring a description of one repetitive task, examples of real requests and the name of the system that holds the answers. For a custom integration, share available OpenAPI documentation, a Postman collection or working cURL calls.
How long does AI agent implementation take?
Timing depends on the workflow and how your systems connect. We estimate it during discovery after reviewing your setup; there is no fixed timeline for every project.
Do we need developers on our team?
A person who understands your system helps us arrange access. Your Lummy implementer configures the agent. If your system needs a new API, we review that requirement during discovery.
How much does implementation cost?
Implementation is included in the pilot quote after discovery. Ongoing pricing combines a monthly fee and usage, with a budget and alerts. International customers are billed in US dollars.
What happens after the pilot?
Review the results in the dashboard before adding workflows or agents. Your team can request corrections, and changes are tested before a new version is published.
BOOK A DISCOVERY CALLThat backlog of work
already has a next step.
Tell us which business process you want to automate first and which systems it uses.
We start with what your business needs solved.