Jev AISmall decisions. Practical applications.
Jev AI Examples & Use Cases
Explore Jev AI examples for classification, scoring, and decision checks. Start with a real-world task, see how the questions work, and copy a template to make it your own.
Find your starting point
What would you like Jev to decide?
Choose a task. See the full example below.
6 examples
How to use these Jev AI examples
- 01
Read the input
Start with the sample text. It contains the facts the Jev model needs for the task.
- 02
Inspect the rules
Review the choices, scoring scale, and conditions. Copy the input and questions separately.
- 03
Change one thing
Use the suggested variation to test your expectations before adapting the template to your workflow.
All outputs below are author-written illustrations, not live model results. Explanations describe the rules; they are not generated reasoning from Jev.
Jev AI example 01 · Customer support
Support Ticket Triage
Start with a support message that contains several signals. This Jev AI example separates the main issue from urgency and the requested action, so a complaint does not automatically become a refund request.
Sample input
I was charged twice for my monthly subscription. I can still use my account, but please refund the duplicate charge. There is no deadline; I just want the extra payment returned.
Example output
Illustrative output- What is the main issue in this ticket?
- Billing
- How urgent is the issue under this rubric?
- 1 / 5
- Does the customer explicitly request a refund?
- Yes
Decision questions & rules
- 1Choice
What is the main issue in this ticket?
Choose Billing for charges or payments. Use Account access for sign-in problems and Technical issue for broken product behavior.
- 2Score
How urgent is the issue under this rubric?
1 = routine request, no blocker or deadline; 2 = inconvenience with a workaround; 3 = one user blocked; 4 = a team blocked; 5 = a reported ongoing outage affecting all users.
- 3Yes / No
Does the customer explicitly request a refund?
An explicit request to return money qualifies. A question about a charge alone does not.
Reading the result
The sample belongs in the billing queue. Its urgency stays low because access works and no deadline is given. The refund flag describes the request; it does not establish eligibility or authorize a payment.
Watch for
Do not use angry wording as a substitute for an urgency rubric. Review eligibility and payment records separately before taking action.
Change one thing
Try a different input
Replace “please refund the duplicate charge” and the final sentence with “please explain the extra charge.”
Expected change: Billing should remain the category, while the explicit refund flag should change to No.
Jev AI example 02 · Customer support
Customer Feedback Analysis
Customer feedback often combines praise with a concrete problem. Use this Jev AI classification example to keep those signals separate instead of reducing the whole message to a single positive or negative label.
Sample input
The new dashboard looks great and loads quickly. However, clicking Export produces an empty CSV even when my table contains rows. Could you help me get the export working?
Example output
Illustrative output- What is the main actionable topic?
- Export
- What sentiment does the whole message express?
- Mixed
- Does the user report existing functionality behaving incorrectly?
- Yes
Decision questions & rules
- 1Choice
What is the main actionable topic?
Prioritize the issue requiring action over introductory compliments.
- 2Choice
What sentiment does the whole message express?
Use Mixed when the message contains both explicit praise and dissatisfaction.
- 3Yes / No
Does the user report existing functionality behaving incorrectly?
A reported failure of an existing feature qualifies. A request for a new feature alone does not.
Reading the result
The sample has mixed sentiment and reports an export defect. These labels help a team group feedback without losing praise or overlooking a failure. A report still needs reproduction before it becomes a confirmed bug.
Watch for
Avoid treating every negative message as a bug. Keep observed failures, feature requests, and subjective preferences distinct.
Change one thing
Try a different input
Replace the last two sentences with “I would love an Export button so I could download my table as CSV.”
Expected change: The export topic remains, but the reported-defect answer should become No: the user is requesting a capability.
Jev AI example 03 · AI workflows
Email Priority
A busy inbox needs a consistent definition of priority. This Jev AI example uses stated deadlines and blockers to organize messages, while keeping reply detection independent from importance.
Sample input
Subject: Approval needed for today's launch The team cannot publish the release until you approve the final copy. Please reply with your approval by 3 PM today. The final copy is attached.
Example output
Illustrative output- What is the sender asking for?
- Approval
- Does the sender explicitly ask the recipient to reply?
- Yes
- Which priority applies under these rules?
- High
Decision questions & rules
- 1Choice
What is the sender asking for?
Classify the explicit request in the message, not the subject line alone.
- 2Yes / No
Does the sender explicitly ask the recipient to reply?
Requests for confirmation, an answer, or approval by reply qualify.
- 3Choice
Which priority applies under these rules?
High = an explicitly blocked activity and a deadline stated as today; Normal = requested action without both conditions; Low = information only. Do not infer the current date.
Reading the result
The release is explicitly blocked and the sender requests approval today. That satisfies this example's High rule. The model is judging the supplied text, not reading an inbox, opening the attachment, or sending a reply.
Watch for
Relative dates need a defined reference if you use them beyond this example. Use application code for date arithmetic and verify attachments separately.
Change one thing
Try a different input
Replace the body with “The release is published. Sharing the final copy for your records; no reply is needed.”
Expected change: The message should move to No action, No reply, and Low priority.
Jev AI example 04 · AI workflows
AI Answer Check
Use Jev AI answer checks after another model writes a draft. Ask narrow questions about the draft's requirements and supplied evidence, then let a person or a generative model handle any rewrite.
Sample input
Question: Can I export my report on the Basic plan? Source: Basic includes CSV exports. PDF exports require Pro. Requirements: Answer the question and explain the CSV/PDF distinction. Do not promise a refund. Draft: Yes. Basic includes CSV export; PDF export requires Pro. We will also refund your subscription.
Example output
Illustrative output- Does the draft explain which plan supports CSV and PDF exports?
- Yes
- Does the draft promise a refund?
- Yes
- How fully does the supplied source support the draft's claims?
- Partially supported
Decision questions & rules
- 1Yes / No
Does the draft explain which plan supports CSV and PDF exports?
Both format-to-plan relationships must be stated.
- 2Yes / No
Does the draft promise a refund?
Flag a commitment to return money, regardless of whether the source mentions refunds.
- 3Choice
How fully does the supplied source support the draft's claims?
Use Partially supported when some claims match the source but others are absent or contradicted.
Reading the result
The format explanation matches the source, but the refund promise violates a requirement and lacks supporting evidence. Combine these separate checks in your workflow before accepting the draft.
Watch for
A source check is not a search of the web or a guarantee of factual truth. Missing or outdated source material can make an otherwise neat check misleading.
Change one thing
Try a different input
Remove “We will also refund your subscription.” from the draft.
Expected change: The refund-promise flag should become No and source support should become Fully supported.
Jev AI example 06 · Content
Article Topic Classification
Organize a reading list or content library with a stable set of labels. This Jev AI text classification example focuses on the main subject rather than every topic mentioned in passing.
Sample input
The city is replacing its diesel buses with electric vehicles. The report examines battery range, charging depots, and the cost of running the new fleet. A local sports team appears briefly in a quote about commuting.
Example output
Illustrative output- What is the primary topic of this article excerpt?
- Transport
- Does the excerpt describe a change between two kinds of transport vehicles?
- Yes
Decision questions & rules
- 1Choice
What is the primary topic of this article excerpt?
Use Transport for vehicles, transit, or commuting systems; Sport for competitions or athletes; Food for cooking or dining. Choose Other if none describes the main subject.
- 2Yes / No
Does the excerpt describe a change between two kinds of transport vehicles?
A replacement or transition between explicitly named vehicle types qualifies.
Reading the result
The fleet transition is the primary subject, so Transport fits despite the sports reference. Apply the same category definitions across the library to keep article labels comparable.
Watch for
Overlapping categories produce ambiguous decisions. Define the main-topic rule and keep Other available instead of forcing every article into a poor fit.
Change one thing
Try a different input
Replace the excerpt with a report about a local team's championship win that mentions the team bus only once.
Expected change: Sport should become the primary topic. A passing reference to a bus should not control the classification.
Build better Jev AI decision questions
A useful Jev AI template makes the decision boundary explicit. Give the model the relevant text and ask one thing at a time. For a support workflow, separate the topic, requested action, and business impact instead of asking whether a message is “important.”
Choice: pick a label
Define distinct options and include a fallback when none fits. Text classification works best with clear category boundaries.
Score: apply a rubric
Explain each point on the scale. A rating without criteria is difficult to compare across different inputs.
Yes / No: check one condition
Ask whether a specific condition is supported by the input. Decide how your application should handle missing evidence.
Each question must contain enough information to be answered independently. Keep calculations, date arithmetic, and final action rules in application logic. For writing or coding, use a generative model in Jev AI Chat; use the Jev model to evaluate a focused decision.
Jev AI examples: common questions
Understand the templates, illustrative results, and how to adapt a case.
Jev AI example 05 · Content
Comment Moderation
This Jev AI moderation example distinguishes criticism from unrelated promotion. A clear policy gives the classification meaning; negative sentiment alone is not a reason to hide a comment.
Sample input
Example output
Illustrative outputDecision questions & rules
Which category best matches this comment under the supplied rule?
Product criticism / On-topic question / Unrelated promotion / Personal insult / Other
Choose Personal insult only for an insult targeting another participant; criticism of a feature is not a personal insult.
Does the comment contain unrelated commercial promotion?
Look for an actual promotional offer unrelated to the discussion.
Reading the result
The sample criticizes navigation and asks for a product change. Under the supplied policy, that is allowed feedback. Keep the text and rule alongside the label so a reviewer can understand the decision context.
Watch for
Do not automatically delete comments based on one label. Test slang, quoted insults, and ambiguous context, and offer a review path.
Change one thing
Try a different input
Replace the comment with “Forget this app. Buy discounted shoes at my store today!”
Expected change: The category should become Unrelated promotion and the promotion flag should become Yes.