Install Now Assist for Creator
Refer to the installation instructions in the docs site (docs.servicenow.com) on how to install and configure Data Visualization generation.
Once installed and the skill activated in Now Assist Admin Center, users will require the now.assist.creator role. Users with this role, will now see the "Create a Data Visualization" topic in the Now Assist panel.
The language of describing data
Decision making in work comes from a combination of intuition and information gathering. In turn, the impacts and externalities of our decisions are measured, generating more data to use in our next decision. As workers, becoming more data savvy, we've slowly been developing our own ways to describe and talk about data. Academic terms used in research, mathematics and statistics to describe the collection, transformation and interpretation of data have been adopted or substituted with more comfortable, approachable language.
While you can read books and take training on data communication, most people learn this language in the same way other languages are learned: exposure and repetition. The way we structure data sentences closely resembles adjective ordering. Many languages have specific ordering of adjectives, in English the order goes: opinion, size, age or shape, colour, origin, material, purpose. Entities in data sentences follow the general order of: aggregate, state or categorization, data source, condition, grouping, trend, presentation.
Let's break down a few example data sentences:
Count of active incidents assigned to me grouped by priority as a pie chart
Spoiler
- Aggregate: Count of
- State: active
- Data source: incidents
- Additional condition: assigned to me
- Grouping: grouped by priority
- Presentation: as a pie chart
Average duration of software incidents resolved this month
Spoiler
- Aggregate: Average duration of
- Categorization: software
- Data source: incidents
- Additional condition: resolved this month
- presentation (inferred): single score
Percentage of successful critical changes per month as a column chart
Spoiler
- Aggregate: Percentage of
- State: successful
- Categorization: critical
- Data source: changes
- Trend: per month
- Presentation: as a column chart
Prompting tips for generating Data Visualizations
Entering into the Create a Data Visualization topic will start the conversation with the following message:
Certainly. What would you like to visualize? For better results, please review the guidelines for analytics generation.
Clicking on the guidelines link will launch a popover with some good guidance on the types of information Now Assist will look for in your data sentences.
Metrics or counts
This part of the sentence represents how the data should be aggregated. Are you after a simple count of records, or do you want to compute an aggregation of a field? e.g. "average reassignment count"
Trends over time
Trending is attached to the sentence using "by" or "per" statements e.g. "count of incidents opened per month". A datefield is required to be the source of the trend. Common date fields used are: created/opened, closed, scheduled start,resolved. A time frame also needs to be declared: "per week", "by quarter"
Sorting and grouping
Some data visualization types require grouping to be defined (if not, Now Assist may ask you!) A pie chart, for example, represents the size of a particular group or category, relative to other elements in the group. Use "grouped by"phrase as a signal to Now Assist to group your data. e.g. Open incidents grouped by category
Dates
Relative to the current time, words which refer to past present and future can be used to create additional time frame based conditions e.g. "Changes scheduled in the next 7 days", "Incidents resolved last month", "Total hours billed last quarter" Just like trends, time frame conditions need a date/time source field and a timeframe which can be from seconds to years.
Type of visualization
Currently Data Visualization generation supports: single score, line, vertical bar, pie chart and list. Be specific about which visualization type you want by using "as a" followed by a type e.g. "count of open incidents grouped by priority as a pie chart". The visualization type can also come at the beginning of the sentence e.g. "List of open incidents", "bar chart showing count of incidents created per day".
Filtering and conditions
Include filtering statements for your data by using operators such as: is, is one of, is not, more than, less than, starts with. Join statements with and / or terms. Use where to indicate the start of conditions.
E.g. Incidents resolved last month where priority is critical, Incidents where priority is critical and category is software
Other useful phrases
Instead of typing out whole conditions, some short-hands exist for select common phrases. These often align to records in the Dynamic Filter Options (sys_filter_option_dynamic_list) table. For example when a condition is being created on a reference to the sys_user field, the operator "is (dynamic)" is available with an option of "Me" e.g. "Incident's assigned to me"
Other short-hands may be expanded by the LLM e.g. P1 (priority=1), unassigned (assigned_to=)
Share your prompt tips below
Start simple and build up your data sentences into more complex prompts.
If you've found prompts or phrases that help you get better accuracy out of Now Assists, let us know in the comments.
