Where to start with AI on IFS
Which use case first, how to roll out, what to measure.
The short answer
Start with questions, prove the answers, then widen by role.
Where does AI help in IFS?
Where IFS holds the answer but people struggle to reach it.
Access
People who rarely use IFS ask its live data in plain words.
Capture
Records are created where and when the event happens.
Explanation
How a process is meant to work, without finding a super user.
Automation
Defined processes run on events and schedules: that is NgageFlow.
Which use case comes first?
A read-only question that someone answers today by interrupting a colleague. Stock balances and order status are the usual choices: asked often, quick to check against IFS, and no write risk.
Why not start with the best demo?
The first phase is not there to impress anyone. It is there to show the answers are right. People trust a system that has been right about small things.
How soon can that use case run?
On IFS, a use case can be live in hours, depending on access to your test environment and the scope. Bring one real question from your own IFS and watch NgageChat work it out.
How should a rollout run?
In four phases, each turned on for a whole role.
| Phase | Scope | Move on when |
|---|---|---|
| 1. Read | Stock, order status, the work order backlog | Answers match what IFS shows |
| 2. Low risk writes | Fault reports, safety incidents, meter readings | Records arrive complete, on the day |
| 3. Operational writes | Work order updates, requisitions, date changes | People use the approval step well |
| 4. Actions | Release, approve, complete | Ongoing: refusals show permissions working |
What should we measure?
The behaviour you wanted to change. Count interruptions avoided, records created at the source, and the time from an event to its IFS record. Message counts show use, not improvement.
What usually goes wrong?
Three things. Turning on writes before people trust the answers. Rolling out a thin slice of every department, so nobody gets a whole job done. And expecting chat to fix master data.
Will it fix our master data?
No. A conversational interface makes bad master data easier to find, not better. If part descriptions are inconsistent, people will see it in week two, which is useful, and not what they expected.
More questions from IFS teams
How soon do we see value?
Answers come from the first day on a test setup. On IFS, a use case can be live in hours, depending on access to your test environment and the scope. Changing how people record work takes longer, because it is habit.
Do we need to clean our data first?
No, but expect the assistant to show you where it is messy. Starting with questions only is a cheap way to find problems before they matter.
Who should own it?
Whoever owns IFS adoption. It is an access and adoption programme with AI inside it, not the other way round.
Where do NgageFlow and NgageCode fit?
Ask first with NgageChat. Automate what repeats with NgageFlow, which runs agents on events and schedules. Build what needs a screen with NgageCode, which builds web and mobile apps on IBM Maximo, IFS and SAP.
See NgageChat answer a question from your IFS
Send us a task you have not prepared for. We run it with you, live, on a test system.