You can enable an AI assistant for
Maximo Manage,
Maximo Health, Operational Dashboard, and asset dashboards. The assistant
uses the nl2oslc model template.
Before you begin
Ensure that Maximo AI
Service is available and running. For more
information, see Maximo AI Service.
About this task
The following table describes what Maximo Assistant can do according to what
Maximo Application Suite version you are using. If you are using the 9.2 feature channel, you
can enable the assistant to access any object. For more information, see
Step 4: Configure what data Maximo Assistant can access.
Table 1. Maximo Assistant capabilities
| Maximo Application Suite version |
Maximo Assistant capabilities |
|
9.1
|
The assistant can retrieve data for assets, work orders, and service requests. In the Object
Structures application, the Use for assistant checkbox is selected in the
object structures that are related to these data types. If you do not want the assistant to answer
questions about those data types, you can clear the checkboxes. The checkbox is available for all
object structures, but the assistant can use only those object structures that have the checkbox
selected by default.
|
|
9.2 feature channel
|
The assistant can retrieve data for any object, including custom
objects. By default, the assistant can answer questions about work orders, assets, service requests,
meters, preventive maintenance, job plans, work logs, assignments, and persons. For more information
about enabling the assistant to access more data, see Step 4: Configure
what data Maximo Assistant can access.
|
|
|
|
After a configuration is activated, to change settings, you must deactivate it. In the AI
configuration, click
, edit the
configuration and then activate the configuration again. The quality of the assistant's answers is
directly dependent on the quality of the data that it uses.
Procedure
- In
Maximo Manage, open the AI configuration
application and then in the table, click the ASSISTANT AI
configuration.
If you delete the default configuration and then add another configuration
for the assistant, ensure that you use the nl2oslc model template and that you select the latest
version of the model.
- Optional: Review the content in the Edit AI
configuration dialog.
- Click
.
- In the Template version field, select the latest version of the
model. To view all versions, click the Lookup icon.
As of 25
November 2025,
IBM®
watsonx™ deprecated its IBM Granite 3.2 8b Instruct
model. If you are already using
Granite™ 3.2 8B
Instruct, you must move to the
gpt-oss-120b model. For more information, see
Changing to gpt-oss-120b models.
model version. If you are using
Maximo AI
Service on-premises, you must have
adequate hardware to run the model.
Table 2. Hardware requirements for GPT
model
| Template version |
Requirements for on-premises Maximo AI
Service
|
|
number-gpt
|
The gpt-oss-120b model fosters better performance, including faster response
times and greater accuracy. However, it requires a minimum of one H100 GPU.
Note: If you already configured the AI assistant but want to change the template version to the
gpt-oss-120b model, changing the template version will delete any set arguments. Make sure that you
record any current argument settings before you upgrade so that you can set the arguments after you
change the template version.
|
- Edit the quick starters.
Quick starters are suggested requests or
questions that appear in the welcome message when the assistant is opened by one of your users.
Quick starters can enable users to explore the assistant and its features more efficiently. You can
use the default quick starters or specify your own. Quick starters cannot be customized to certain
roles or users. Specify requests or questions that most of your users might want and that the
assistant can answer based on the selected object structures.
- Click OK.
- Set up arguments.
Arguments control some aspects of the models output
and how it processes data.
- Click . Some earlier versions of the assistant do not have arguments
that you can set. If the Actions menu does not contain a Set
arguments action, you do not set arguments.
- Review the values for the arguments.
Table 3. Arguments for assistant model
template
| Argument key |
Description |
Type |
| use_domain_rule_filter |
Indicate whether to limit what prompt tuning prompts are sent to the model. If set to true,
before answering a user prompt, the model completes a hybrid rag search on the prompt tuning prompts
based on the user's prompt. It extracts and reviews only the most relevant domain rules. It can
extract up to 50 rules. Setting to true might increase relevancy and accuracy of responses.
|
string |
| use_explanation |
Indicate whether responses include the query that was used to retrieve data. If set to true,
which means the query is included, performance can be affected.
|
string |
| use_router_filter |
Indicate whether prompts are screened for relevance before the assistant processes them. By
default and as a best practice, this argument is set to false. If set to true, meaning the prompts
are screened, performance can be affected.
|
string |
| use_schema_filter |
Indicate whether the model reviews only the most relevant business objects and attributes. The
model determines, based on the user's prompt, what the most relevant business objects and attributes
are. Setting to true might improve response time.
|
string |
- Click Save.
- If you are using the feature channel, you can also
configure what data Maximo Assistant can access.
- On the Searchable data tab, review the default object
structures that the assistant has access to.
- In the Enabled column, disable any object structures that you
do not want the assistant to access.
- To add an object structure, click Add and then select the
object structure and default query template.
When the assistant returns a data table,
the query template determines the table column headers. You create query templates in the Object
Structures application. For an example of a template, choose a query template for one of the default
object structures, locate that object structure in the Object Structures application, and in the
More Actions menu, click Query Template and locate the
query template.
- Click
.
Activating the AI configuration indicates that the AI configuration is prepared.
- From the
Creating the model enables the assistant on the
user interface. The assistant is available for
Maximo Manage,
Maximo Health, Operational Dashboard, and asset dashboards.
The
assistant appears in the lower right of the interface. You can drag and drop the icon to different
areas on the user interface, but the assistant chat window always opens on the right side of the
page.
Note: If you are using the Maximo Application Suite customer-managed feature
channel, the AI assistant is launched from an icon in the page header in
Maximo Manage, Maximo Health, Operational
Dashboard, and asset dashboards. The chat window opens on the right side of the page. If the
assistant requires time to respond to a question or request, you can close the chat window. The
assistant continues working, and the icon notifies you when the work is
done.
- Grant users access to the assistant.
Access to the AI assistant chat window is
managed through security groups, but if users do not have access to data, the assistant cannot
answer questions. Users can only receive answers that contain data that they have access to.
- In Security Groups, open a security group and on the
Applications tab, in the Applications table, in the
Description column, search for
assistant.
- In the Options for assistant table, as a system administrator, you can grant or remove
access to the Maximo Assistant option.
What to do next
If you want to edit the quick starters or other configuration settings, you must deactivate the
model first. In the AI configuration, click
.
Optionally, you can complete prompt tuning. Prompt tuning is a process in which you
specify parameters, also known as prompts, for the model to consider. Each prompt is context
that the assistant uses when answering questions or requests. Complete prompt tuning only if you
want to alter how the assistant responds to certain questions.
If you are enabling your organization's first deployment of the AI assistant, you might skip
prompt tuning until you complete testing of the assistant. If the assistant can answer questions
accurately for your organization, you do not need to complete prompt tuning. For more information,
see Prompt tuning for the Maximo Assistant.
The assistant does not currently include a feedback mechanism for answers. Conversation history
cannot be saved. The assistant does not support integrations with external systems.
For more information about using the assistant, see Using the assistant.