Questions
Frequently asked questions
Short, plain answers on security, data, the AI account, timing and what a person can and cannot do.
How secure is it?
Every question runs under the asking person's own JD Edwards security. Writes use the sanctioned JD Edwards routes and are re-read from the table. The platform never holds a copy of anyone's corporate password. An administrator can stop all AI activity with a kill switch, and can set rate limits and spend budgets.
Where does my data live?
Each customer has a dedicated Azure subscription, in a region chosen during onboarding to match your JD Edwards geography. Where AI processing happens depends on the route you choose: Anthropic directly, or AWS Bedrock in your own AWS account.
Who owns the AI key?
You do, always. You create the credential, hold the billing relationship with the provider, set your own spending limits and rotate it. Fusion5 does not supply model credentials.
Which JD Edwards versions are supported?
Onboarding needs your exact JD Edwards tools release, because some services vary by release. Ask us to confirm your release before you commit.
Does it work with JD Edwards in the cloud or in your own data centre?
Either. If your JD Edwards web services are reachable over HTTPS, the platform connects directly. If not, a small outbound-only relay connects without exposing anything. The relay is the more common answer. The platform itself is a managed cloud service.
How long does onboarding take?
Two to four weeks is typical. The technical work is about a week; the rest is your approvals. The fastest onboardings book the JD Edwards administrator session on day one.
What does it cost to run?
Fusion5 looks after hosting, monitoring and updates. You pay your own AI usage directly to your provider, and it is visible per conversation and per scheduled run, with optional daily and monthly budgets. Talk to us for commercial terms.
What can a person do, and what can they not?
Only what their own JD Edwards security allows. Assistants and agents can write, and each can be set to read only. Agents are limited to the tool families granted to them, and anything triggered by email should run read only.
How accurate is it, and how is it checked?
Data is read live from JD Edwards, not generated. Writes are re-read, each run carries a trust badge, and arithmetic and rules can run as code. Every chat says answers are written by AI and can be wrong, so key figures should still be checked.
What happens to my data?
JD Edwards stays your system of record, and the platform keeps no replica or warehouse of it. Conversation history and logs hold what was shown in a conversation. An optional retention job deletes logs after a period you set. It is off by default.
Can my own people build assistants and agents?
Yes, without code. People can build from a wizard, a description or a template, write rules as skills, and move builds from one environment to another as a bundle. Training tracks cover end users, builders and administrators.
Do I have to move off JD Edwards or change it?
No. No data is copied out, no JD Edwards objects are changed, and the same security model applies.