Draxlr in AI Database Agents
Scenario-level performance from current published Results.
23 of 28 scenarios published
Where Draxlr failed
7 of 22 graded scenarios. Open a Result to see the conversation and the evidence.
The answer is a trend or a comparison
Draxlr answered a monthly order-count question with a sentence and a three-row table, and drew no chart. Its Visualization pane, shown by default, held a table of the three months with no axes, bars or line.
View Result →The agent offers what to ask next
Draxlr answered the revenue question without suggesting what to ask next. Its chat left the user to write a follow-up rather than offering a question based on the analysis.
View Result →The follow-up is ambiguous
Draxlr did not ask for clarification when the follow-up was ambiguous. After the first reply produced two plausible customer sets, the second reply answered directly and the transcript shows no clarifying question, option prompt, or disambiguation before it.
View Result →The user changes direction mid-conversation
Draxlr dropped the Texas filter when asked about all states, but reported 20.04% rather than the overall rate of 21.75%. It used an unweighted average of state percentages.
View Result →The user narrows what they just asked
Draxlr kept the original 90-day condition when the user narrowed the question to Texas, but returned no orders even though three qualified. It added a Texas filter using the full state name rather than the two-letter abbreviation stored in the order system.
View Result →An excluded table is needed to answer the question
Draxlr fails this scenario: when the needed table is excluded, it still answers from another table instead of saying it cannot answer. The chat returned 0 pending return requests, while the excluded table stayed hidden before and after the run.
View Result →The user's wording doesn't match how the values are stored
Draxlr did not bridge the wording gap for the gift voucher request. It searched for those literal words and presented an empty result as the answer.
View Result →How Draxlr performed
Open a capability to explore its scenarios. Each row reports the test set in its published Result.
Analytics & Observability2 pass2 no Result yet
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| A user reports a bad answer and the operator has to find it | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr let the admin find and inspect the reported exchange in saved chats. The opened conversation exposed the original question and response, with the SQL query and its explanation available alongside them. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| Someone needs to see what people have been asking | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr made the questions visible and attributable to the people who asked them. The admin could see all five test chats in the sidebar and identify which account asked each question. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| Reported analytics match what actually happened | No published result | ||
| The operator needs to find the questions the agent is failing on | No published result | ||
Answer Presentation3 pass1 fail
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| The answer is a plain fact or a short list | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr gave a single-value answer in plain language. It stated that there were 308 orders in total. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The answer is a trend or a comparison | 0 Pass1 Fail0 Not gradable 1 of 1 test case failedWhat happenedDraxlr answered a monthly order-count question with a sentence and a three-row table, and drew no chart. Its Visualization pane, shown by default, held a table of the three months with no axes, bars or line. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The answer needs a summary and its detail together | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr answered a question about last month's revenue with a summary and its breakdown together in one reply. It stated in prose that August revenue was $14,403.87, down $3,858.77 from July's $18,262.64, and set out each of seven categories' August and July revenue and change in a table, led by furniture's $6,562.19 drop. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The user asks to see the same answer a different way | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr redrew its revenue answer as a bar chart when asked, and the numbers stayed the same. The chart's seven category changes, from furniture down $6,562.19 to bedding and rugs up $1,129.47, are the first answer's figures and the reference figures to the cent, and the chart's own table and query are unchanged. | 1/1 assessed1/1 gradablePublished test set | View Result → |
Conversational Interaction2 pass4 fail
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| Customer corrects information given earlier | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr handled the follow-up correction by using the updated month instead of the original one. The second reply said June 2026, and the second query was limited to June only; it did not combine July with June or keep July in place. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The agent offers what to ask next | 0 Pass1 Fail0 Not gradable 1 of 1 test case failedWhat happenedDraxlr answered the revenue question without suggesting what to ask next. Its chat left the user to write a follow-up rather than offering a question based on the analysis. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The follow-up is ambiguous | 0 Pass1 Fail0 Not gradable 1 of 1 test case failedWhat happenedDraxlr did not ask for clarification when the follow-up was ambiguous. After the first reply produced two plausible customer sets, the second reply answered directly and the transcript shows no clarifying question, option prompt, or disambiguation before it. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The next question refers to the previous answer | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr kept the previous answer’s customer set intact in the follow-up. It answered for the same five customers, did not ask who “they” were, and the second query rebuilt that same set before counting their last-90-day orders. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The user changes direction mid-conversation | 0 Pass1 Fail0 Not gradable 1 of 1 test case failedWhat happenedDraxlr dropped the Texas filter when asked about all states, but reported 20.04% rather than the overall rate of 21.75%. It used an unweighted average of state percentages. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The user narrows what they just asked | 0 Pass1 Fail0 Not gradable 1 of 1 test case failedWhat happenedDraxlr kept the original 90-day condition when the user narrowed the question to Texas, but returned no orders even though three qualified. It added a Texas filter using the full state name rather than the two-letter abbreviation stored in the order system. | 1/1 assessed1/1 gradablePublished test set | View Result → |
Data Access Control1 fail2 no Result yet
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| An excluded table is needed to answer the question | 0 Pass1 Fail0 Not gradable 1 of 1 test case failedWhat happenedDraxlr fails this scenario: when the needed table is excluded, it still answers from another table instead of saying it cannot answer. The chat returned 0 pending return requests, while the excluded table stayed hidden before and after the run. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| An excluded column is asked for directly | No published result | ||
| Two users are given different data scope | No published result | ||
Question Answering2 pass1 fail1 different test set
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| The answer is available in the connected data | Different test setPublished Result available, but not included in this comparison. | View Result → | |
| The connected data cannot answer the question | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr identified missing review data and gave no count for this month’s five-star reviews. It asked which part of the database should represent those reviews. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The question uses a term the company defines itself | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr answered that revenue over the 90 days to September 2, 2026 was $39,806.75 and said in the same answer that it used all order statuses. Its explanation and generated query match that stated choice, which counts cancelled orders as revenue. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The user's wording doesn't match how the values are stored | 0 Pass1 Fail0 Not gradable 1 of 1 test case failedWhat happenedDraxlr did not bridge the wording gap for the gift voucher request. It searched for those literal words and presented an empty result as the answer. | 1/1 assessed1/1 gradablePublished test set | View Result → |
Reporting2 pass1 no Result yet
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| The report is run again after the data has changed | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr refreshed the saved report after three orders were added, keeping the same period. Both the report and the database count for that period moved from 26 to 29 orders. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The user wants several findings in one place | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr passed this request to gather several findings in one place. It gave the three direct answers, then combined them into one report without changing the values; the saved report reopened with the same single row. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The user wants to keep an answer they just got | No published result | ||
Training4 pass
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| The agent used the wrong definition and the user corrects it | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr retained the user’s corrected metric definition for both a repeated question and a differently worded one. Both revenue answers were $23,194.52, the same as the reference total. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The database's names are cryptic and the user documents them | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr used the supplied documentation to interpret a cryptic column name and select the right data for the question. It returned 7 returns flagged for manual fraud review, matching the reference count of 7. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The user sets a standing rule for all future answers | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr carried a standing exclusion rule into two later, unrelated questions without a reminder. Its revenue and order-count answers were $151,064.66 and 88, matching the database totals with cancelled orders excluded. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| The user supplies example questions and the queries they trust | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedDraxlr generalized a trusted question-and-query example to a rephrased question. After training, both phrasings returned 42, the trusted count, and the rephrased question used the trusted query’s logic. | 1/1 assessed1/1 gradablePublished test set | View Result → |
Reading these Results
Published evidence and test coverage answer different questions.
Which scenarios have a Result?
A published Result is public evidence for this tool on one scenario. “No published result” does not say whether testing has taken place.
What does each Result cover?
Assessed includes Pass, Fail and Not gradable. Gradable includes Pass and Fail. Both use the pinned test count in that published Result.
Out of 28 scenarios
The summary counts each of the 28 scenarios once. A scenario fails when any of its published test cases failed. “No published Result yet” means nothing is public for that scenario; it does not say whether Draxlr was tested on it.