Ask IIB in all four verticals
IIB Board AI Pack, page 19. First phase. Built here on synthetic data.
Today. When a Vertical Head or senior leader has a new question, it goes to the analytics team as a request and waits its turn alongside other work.
What would change. Leaders type the question in their own words and get an answer from IIB's data straight away, with the source and the definition used shown underneath. Personal data never appears in an answer.
The live test, by vertical
Last live run
| Vertical | Asked as | Asked | Correct | Share | Against 85% |
|---|---|---|---|---|---|
| Motor | Motor Vertical Head | 20 | 19 | 95% | |
| Health | Health Vertical Head | 20 | 18 | 90% | |
| Life | Life Vertical Head | 20 | 20 | 100% | |
| Other Lines | Other Lines Vertical Head | 20 | 18 | 90% |
Run started 7 Oct 2026, 09:19 and finished 7 Oct 2026, 09:22. It asked 80 questions. Wrong 5, declined by the AI 0, failed the checks 0, no reply 0. 2 needed the one automatic correction the Ask page also makes.
Earlier live runs (2)
| Started | Kind | Asked | Correct | Note |
|---|---|---|---|---|
| 7 Oct 2026, 07:39 | Sample | 10 | 10 | A cap on Azure OpenAI calls while the page was built stopped this run after 10 questions. O02 and O03 were never sent, so they are left out. |
| 7 Oct 2026, 07:35 | Sample | 16 | 10 | Run with an earlier version of the prompt, before the rules on one-row totals and one row for each hospital or agent. A cap on Azure OpenAI calls while the page was built stopped it, so O02, O03, O04, O05 were never sent and are left out. L04 failed on a tie in its own reference query, since fixed in the bank. |
What counts as correct. Each question is asked as its vertical's Vertical Head, through the same path as the Ask page. Azure OpenAI writes the query, the checks run with one automatic correction, and the query runs.
The result is compared with a hand-written reference query. The essential figures must match within 0.5%. Column names and layout may differ. A ranking must have the same rows at the top.
The test scores the figures each answer is built from, not the wording.
First time. The person asks once and does not have to rephrase. The one automatic correction happens before any answer is shown, and the run counts how often it was needed.
Honest limits. The 80 questions were written for this demo, and the problems they find were built into the synthetic test data. The real test is a pilot with questions from IIB's own staff.
What Ask IIB finds in each vertical
These four answers come from the golden library and its answer templates, governed for your role. Nothing was sent to an AI service. Each pill checks the query's own result against the story planted in the test data.
The 80 test questions
| No. | Question | Reaches | Offline | Last live run |
|---|---|---|---|---|
| M01 | How many motor policies were in force at the end of September 2026?
Reference querySELECT SUM(policies_in_force) AS policies_in_force FROM srv_policies_monthly WHERE month = DATE '2026-09-01' LIMIT 1 |
Declined | Correct | |
| M02 | Which five states had the most motor claims in the last 12 months?
Reference querySELECT state_name, SUM(claims) AS claims FROM srv_claim_frequency WHERE month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01' GROUP BY state_name ORDER BY claims DESC LIMIT 5 |
Declined | Correct | |
| M03 | How did two-wheeler claims in Telangana change this quarter, compared with last year?
Reference querySELECT CASE WHEN month >= DATE '2026-07-01' THEN 'Jul to Sep 2026' ELSE 'Jul to Sep 2025' END AS period,
SUM(claims) AS claims
FROM srv_claim_frequency
WHERE state_code = 'TS' AND vehicle_class = '2W'
AND (month BETWEEN DATE '2026-07-01' AND DATE '2026-09-01' OR month BETWEEN DATE '2025-07-01' AND DATE '2025-09-01')
GROUP BY 1
ORDER BY 1
LIMIT 200 |
Telangana two-wheeler claims rise | Answered Q11 | Correct |
| M04 | Which RTOs in Hyderabad had the most theft claims in the last six weeks?
Reference querySELECT rto_code, SUM(claims) AS theft_claims
FROM srv_claims_weekly
WHERE rto_code IN ('TS09', 'TS10', 'TS11', 'TS12', 'TS13') AND claim_type = 'THEFT'
AND week_start BETWEEN DATE '2026-08-24' AND DATE '2026-09-28'
GROUP BY rto_code
ORDER BY theft_claims DESC, rto_code
LIMIT 10 |
Telangana two-wheeler claims rise | Declined | Correct |
| M05 | What was the average own-damage claim amount for private cars in each state over the last 12 months? Show the ten highest.
Reference querySELECT state_name, SUM(claims) AS claims, ROUND(SUM(claim_amount) / SUM(claims)) AS avg_claim_amount FROM srv_claim_frequency WHERE vehicle_class = 'PCAR' AND claim_type = 'OD' AND month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01' GROUP BY state_name ORDER BY avg_claim_amount DESC LIMIT 10 |
Declined | Correct | |
| M06 | How much premium was written for each vehicle class this year?
Reference querySELECT vehicle_class, ROUND(SUM(premium)) AS premium FROM srv_policies_monthly WHERE month >= DATE '2026-01-01' GROUP BY vehicle_class ORDER BY premium DESC LIMIT 200 |
Declined | Correct | |
| M07 | How many new motor policies were issued each month in Telangana this year?
Reference querySELECT month, SUM(new_policies) AS new_policies FROM srv_policies_monthly WHERE state_code = 'TS' AND month >= DATE '2026-01-01' GROUP BY month ORDER BY month LIMIT 200 |
Declined | Correct | |
| M08 | Which five states have the highest theft claim rate per 1,000 policies over the last 12 months?
Reference queryWITH m AS ( SELECT state_name, month, SUM(claims) AS claims, SUM(policies_in_force) AS pif FROM srv_claim_frequency WHERE claim_type = 'THEFT' AND month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01' GROUP BY state_name, month ) SELECT state_name, SUM(claims) AS theft_claims, ROUND(SUM(claims) * 1000.0 / AVG(pif), 2) AS theft_claims_per_1000_policies FROM m GROUP BY state_name ORDER BY theft_claims_per_1000_policies DESC LIMIT 5 |
Declined | Correct | |
| M09 | How many vehicles are on the register in Rajasthan, by vehicle class?
Reference querySELECT vehicle_class, SUM(registered_stock) AS registered_stock FROM srv_uninsured_estimate WHERE state_code = 'RJ' GROUP BY vehicle_class ORDER BY registered_stock DESC LIMIT 200 |
Declined | Correct | |
| M10 | Which five states had the most motor vehicle thefts reported to police in the latest NCRB year?
Reference querySELECT state_name, year, mv_theft_cases FROM srv_theft_ncrb WHERE year = (SELECT MAX(year) FROM srv_theft_ncrb) ORDER BY mv_theft_cases DESC LIMIT 5 |
Declined | Correct | |
| M11 | How many people were killed in road accidents in Telangana each year?
Reference querySELECT year, SUM(killed) AS killed FROM srv_accidents WHERE state_code = 'TS' GROUP BY year ORDER BY year LIMIT 200 |
Declined | Correct | |
| M12 | Third-party claims in Maharashtra, July to September 2026, against the same months of 2025?
Reference querySELECT CASE WHEN month >= DATE '2026-07-01' THEN 'Jul to Sep 2026' ELSE 'Jul to Sep 2025' END AS period,
SUM(claims) AS tp_claims
FROM srv_claim_frequency
WHERE state_code = 'MH' AND claim_type = 'TP'
AND (month BETWEEN DATE '2026-07-01' AND DATE '2026-09-01' OR month BETWEEN DATE '2025-07-01' AND DATE '2025-09-01')
GROUP BY 1
ORDER BY 1
LIMIT 200 |
Declined | Correct | |
| M13 | What share of two-wheelers on the register in Rajasthan are insured?
Reference querySELECT state_name, ROUND(insured_share * 100, 1) AS insured_share FROM srv_uninsured_estimate WHERE state_code = 'RJ' AND vehicle_class = '2W' LIMIT 1 |
Declined | Correct | |
| M14 | Which insurers reported the most motor claims in the September 2026 cycle?
Reference querySELECT insurer_code, SUM(claims) AS claims FROM srv_claims_weekly WHERE cycle = '2026-09' GROUP BY insurer_code ORDER BY claims DESC LIMIT 10 |
Declined | Correct | |
| M15 | What share of the amount claimed has been paid so far, by claim type, over the last 12 months?
Reference querySELECT claim_type, ROUND(SUM(claim_amount)) AS claim_amount, ROUND(SUM(paid_amount)) AS paid_amount,
ROUND(SUM(paid_amount) * 100.0 / SUM(claim_amount), 1) AS paid_share_pct
FROM srv_claims_weekly
WHERE cycle BETWEEN '2025-10' AND '2026-09'
GROUP BY claim_type
ORDER BY claim_type
LIMIT 200 |
Declined | Wrong | |
| M16 | How many own-damage claims did goods carrying vehicles have each month this year?
Reference querySELECT month, SUM(claims) AS od_claims FROM srv_claim_frequency WHERE vehicle_class = 'GCV' AND claim_type = 'OD' AND month >= DATE '2026-01-01' GROUP BY month ORDER BY month LIMIT 200 |
Declined | Correct | |
| M17 | Which ten RTOs in Telangana have the most two-wheeler policies in force?
Reference querySELECT rto_code, SUM(policies_in_force) AS policies_in_force FROM srv_policies_monthly WHERE state_code = 'TS' AND vehicle_class = '2W' AND month = DATE '2026-09-01' GROUP BY rto_code ORDER BY policies_in_force DESC LIMIT 10 |
Declined | Correct | |
| M18 | What is the average premium per new private car policy by cover type this year?
Reference querySELECT cover_type, SUM(new_policies) AS new_policies, ROUND(SUM(premium) / SUM(new_policies)) AS avg_premium FROM srv_policies_monthly WHERE vehicle_class = 'PCAR' AND month >= DATE '2026-01-01' GROUP BY cover_type ORDER BY avg_premium DESC LIMIT 200 |
Declined | Correct | |
| M19 | Show theft claims in Telangana week by week for the last 12 weeks.
Reference querySELECT week_start, SUM(claims) AS theft_claims FROM srv_claims_weekly WHERE state_code = 'TS' AND claim_type = 'THEFT' AND week_start BETWEEN DATE '2026-07-13' AND DATE '2026-09-28' GROUP BY week_start ORDER BY week_start LIMIT 200 |
Telangana two-wheeler claims rise | Declined | Correct |
| M20 | How many vehicles were registered across India in each of the last five years of VAHAN data?
Reference querySELECT year, SUM(registrations) AS registrations FROM srv_registrations WHERE year > (SELECT MAX(year) FROM srv_registrations) - 5 GROUP BY year ORDER BY year LIMIT 200 |
Declined | Correct |
| No. | Question | Reaches | Offline | Last live run |
|---|---|---|---|---|
| H01 | How many health claims were there in each of the last 12 months?
Reference querySELECT month, SUM(claims) AS claims FROM srv_health_claims_monthly WHERE month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01' GROUP BY month ORDER BY month LIMIT 200 |
Declined | Correct | |
| H02 | Which hospitals charge the most compared with similar hospitals over the last four months, among those with at least 20 discharged patients?
Reference querySELECT hospital_name, SUM(claims_discharged) AS discharged_claims,
ROUND(SUM(billed_amount) / SUM(peer_billed_amount), 2) AS cost_vs_peers
FROM srv_health_hospital_monthly
WHERE month BETWEEN DATE '2026-06-01' AND DATE '2026-09-01'
GROUP BY hospital_name
HAVING SUM(claims_discharged) >= 20
ORDER BY cost_vs_peers DESC
LIMIT 10 |
Hospitals billing far above peers | Answered HQ1 | Correct |
| H03 | Which hospitals had the longest average stay since June 2026, among hospitals with at least 50 discharged patients?
Reference querySELECT hospital_name, SUM(claims_discharged) AS discharged_claims,
ROUND(SUM(avg_los_days * claims_discharged) / SUM(claims_discharged), 2) AS avg_stay_days
FROM srv_health_hospital_monthly
WHERE month BETWEEN DATE '2026-06-01' AND DATE '2026-09-01'
GROUP BY hospital_name
HAVING SUM(claims_discharged) >= 50
ORDER BY avg_stay_days DESC
LIMIT 10 |
Hospitals billing far above peers | Declined | Correct |
| H04 | Which hospitals billed bypass surgery or angioplasty without declaring a cardiac unit or cath lab?
Reference querySELECT hospital_name, SUM(claims) AS claims
FROM srv_health_hospital_procedures
WHERE procedure_code IN ('CARD-CABG', 'CARD-PTCA') AND NOT has_required_capability
GROUP BY hospital_name
ORDER BY claims DESC
LIMIT 10 |
Heart procedures without the facility | Answered HQ2 | Wrong |
| H05 | For each procedure, how many claims came from hospitals that do not declare the facility it needs?
Reference querySELECT procedure_code, SUM(claims) AS claims, COUNT(*) AS hospitals FROM srv_health_hospital_procedures WHERE NOT has_required_capability GROUP BY procedure_code ORDER BY claims DESC LIMIT 200 |
Heart procedures without the facility | Declined | Correct |
| H06 | Show dengue admissions in Kolkata week by week for the last 12 weeks.
Reference querySELECT week_start, SUM(dengue_claims) AS dengue_claims FROM srv_health_weather_weekly WHERE district_code = 'WB-KOL' AND week_start BETWEEN DATE '2026-07-13' AND DATE '2026-09-28' GROUP BY week_start ORDER BY week_start LIMIT 200 |
Dengue after heavy rain | Declined | Correct |
| H07 | Which three districts had the most dengue admissions in the last six weeks?
Reference querySELECT district, SUM(dengue_claims) AS dengue_claims FROM srv_health_weather_weekly WHERE week_start BETWEEN DATE '2026-08-24' AND DATE '2026-09-28' GROUP BY district ORDER BY dengue_claims DESC LIMIT 3 |
Dengue after heavy rain | Answered HQ3 | Correct |
| H08 | Show weekly rainfall and dengue admissions in Lucknow since July 2026.
Reference querySELECT week_start, SUM(rainfall_mm) AS rainfall_mm, SUM(dengue_claims) AS dengue_claims FROM srv_health_weather_weekly WHERE district_code = 'UP-LKO' AND week_start >= DATE '2026-07-01' GROUP BY week_start ORDER BY week_start LIMIT 200 |
Dengue after heavy rain | Declined | Correct |
| H09 | Which five health products reject the highest share of settled claims, among products with at least 300 settled claims?
Reference querySELECT product_name, claims_settled, ROUND(rejected_share * 100, 1) AS rejected_share_pct FROM srv_health_products WHERE claims_settled >= 300 ORDER BY rejected_share DESC LIMIT 5 |
Policy terms that cut claims | Declined | Correct |
| H10 | What share of settled claims were cut or rejected, by the product's waiting period for pre-existing diseases?
Reference querySELECT ped_wait_months, SUM(claims_settled) AS claims_settled,
ROUND((SUM(claims_settled_part) + SUM(claims_rejected)) * 100.0 / SUM(claims_settled), 1) AS cut_or_rejected_pct
FROM srv_health_products
GROUP BY ped_wait_months
ORDER BY ped_wait_months
LIMIT 200 |
Policy terms that cut claims | Declined | Correct |
| H11 | What share of settled claims were cut or rejected for each kind of room rent limit?
Reference querySELECT room_rent_limit, SUM(claims_settled) AS claims_settled,
ROUND((SUM(claims_settled_part) + SUM(claims_rejected)) * 100.0 / SUM(claims_settled), 1) AS cut_or_rejected_pct
FROM srv_health_products
GROUP BY room_rent_limit
ORDER BY cut_or_rejected_pct DESC
LIMIT 200 |
Policy terms that cut claims | Declined | Correct |
| H12 | What share of health claims in Maharashtra were cashless over the last 12 months?
Reference querySELECT SUM(claims) AS claims, SUM(claims_cashless) AS cashless_claims,
ROUND(SUM(claims_cashless) * 100.0 / SUM(claims), 1) AS cashless_share_pct
FROM srv_health_claims_monthly
WHERE state_code = 'MH' AND month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01'
LIMIT 1 |
Declined | Correct | |
| H13 | What is the average bill per discharged patient by hospital type over the last 12 months?
Reference querySELECT hospital_type, SUM(claims_discharged) AS discharged_claims,
ROUND(SUM(billed_amount) / SUM(claims_discharged)) AS cost_per_admission
FROM srv_health_claims_monthly
WHERE month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01'
GROUP BY hospital_type
ORDER BY cost_per_admission DESC
LIMIT 200 |
Declined | Correct | |
| H14 | Which ten districts had the most health claims this year?
Reference querySELECT district, SUM(claims) AS claims FROM srv_health_claims_monthly WHERE month >= DATE '2026-01-01' GROUP BY district ORDER BY claims DESC LIMIT 10 |
Declined | Correct | |
| H15 | Which five insurers rejected the highest share of their settled health claims over the last 12 months?
Reference querySELECT insurer_code, SUM(claims - claims_open) AS settled_claims, SUM(claims_rejected) AS rejected_claims,
ROUND(SUM(claims_rejected) * 100.0 / SUM(claims - claims_open), 1) AS rejection_rate_pct
FROM srv_health_claims_monthly
WHERE month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01'
GROUP BY insurer_code
ORDER BY rejection_rate_pct DESC
LIMIT 5 |
Declined | Correct | |
| H16 | How many network hospitals in each state are NABH accredited?
Reference querySELECT state_name, COUNT(*) AS hospitals, COUNT(*) FILTER (WHERE nabh) AS nabh_hospitals FROM srv_health_hospitals GROUP BY state_name ORDER BY nabh_hospitals DESC, state_name LIMIT 200 |
Declined | Wrong | |
| H17 | How many network hospitals declare a cath lab, by city tier?
Reference querySELECT tier, COUNT(*) AS hospitals FROM srv_health_hospitals WHERE list_contains(string_split(capabilities, '|'), 'cath_lab') GROUP BY tier ORDER BY tier LIMIT 200 |
Declined | Correct | |
| H18 | Which eight procedures have the longest average stay in hospital?
Reference querySELECT procedure_name, avg_los_days FROM srv_health_procedures ORDER BY avg_los_days DESC LIMIT 8 |
Declined | Correct | |
| H19 | Which five states had the most malaria admissions this year?
Reference querySELECT state_name, SUM(malaria_claims) AS malaria_claims FROM srv_health_weather_weekly WHERE week_start >= DATE '2026-01-01' GROUP BY state_name ORDER BY malaria_claims DESC LIMIT 5 |
Declined | Correct | |
| H20 | How did health claims in Telangana change this quarter compared with the same quarter last year?
Reference querySELECT CASE WHEN month >= DATE '2026-07-01' THEN 'Jul to Sep 2026' ELSE 'Jul to Sep 2025' END AS period,
SUM(claims) AS claims
FROM srv_health_claims_monthly
WHERE state_code = 'TS'
AND (month BETWEEN DATE '2026-07-01' AND DATE '2026-09-01' OR month BETWEEN DATE '2025-07-01' AND DATE '2025-09-01')
GROUP BY 1
ORDER BY 1
LIMIT 200 |
Declined | Correct |
| No. | Question | Reaches | Offline | Last live run |
|---|---|---|---|---|
| L01 | How many life policies were in force at the end of September 2026?
Reference querySELECT SUM(policies_in_force) AS policies_in_force FROM srv_life_policies_monthly WHERE month = DATE '2026-09-01' LIMIT 1 |
Declined | Correct | |
| L02 | Which two districts had the most early death claims from May to September 2026?
Reference querySELECT district_name, SUM(early_claims) AS early_claims FROM srv_life_death_claims_monthly WHERE month BETWEEN DATE '2026-05-01' AND DATE '2026-09-01' GROUP BY district_name ORDER BY early_claims DESC LIMIT 2 |
Early death claims in two Bihar districts | Answered LQ1 | Correct |
| L03 | How many early death claims came through PoS agents in Bihar each month this year?
Reference querySELECT month, SUM(early_claims) AS early_claims FROM srv_life_death_claims_monthly WHERE state_code = 'BR' AND channel = 'pos' AND month >= DATE '2026-01-01' GROUP BY month ORDER BY month LIMIT 200 |
Early death claims in two Bihar districts | Declined | Correct |
| L04 | Which PoS agents have the most death claims within a year of the policy start?
Reference querySELECT agent_id, early_death_claims FROM srv_life_agents WHERE registry = 'pos' ORDER BY early_death_claims DESC, agent_id LIMIT 12 |
Early death claims in two Bihar districts | Declined | Correct |
| L05 | How many different insurers had early death claims in Gaya and Nalanda since May 2026?
Reference querySELECT COUNT(DISTINCT insurer_code) AS insurers
FROM srv_life_death_claims_monthly
WHERE district_code IN ('BR-GAY', 'BR-NAL') AND month BETWEEN DATE '2026-05-01' AND DATE '2026-09-01' AND early_claims > 0
LIMIT 1 |
Early death claims in two Bihar districts | Declined | Correct |
| L06 | What share of death claims in Gaya and Nalanda since May 2026 were on policies issued without a medical examination?
Reference querySELECT SUM(claims) AS claims, SUM(non_medical_claims) AS non_medical_claims,
ROUND(SUM(non_medical_claims) * 100.0 / SUM(claims), 1) AS non_medical_share_pct
FROM srv_life_death_claims_monthly
WHERE district_code IN ('BR-GAY', 'BR-NAL') AND month BETWEEN DATE '2026-05-01' AND DATE '2026-09-01'
LIMIT 1 |
Early death claims in two Bihar districts | Declined | Correct |
| L07 | What share of decided death claims did insurers repudiate, by channel, over the last 12 months?
Reference querySELECT channel, SUM(paid) AS paid, SUM(repudiated) AS repudiated,
ROUND(SUM(repudiated) * 100.0 / NULLIF(SUM(paid) + SUM(repudiated), 0), 1) AS repudiation_rate_pct
FROM srv_life_death_claims_monthly
WHERE month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01'
GROUP BY channel
ORDER BY repudiation_rate_pct DESC
LIMIT 200 |
Answered LQ4 | Correct | |
| L08 | What share of ULIPs lapsed within their first year, by sales channel?
Reference querySELECT channel, SUM(policies) AS policies,
ROUND(SUM(lapsed_within_1_year) * 100.0 / SUM(policies), 1) AS lapse_rate_pct
FROM srv_life_cohorts
WHERE product_type = 'ulip' AND full_year_observed
GROUP BY channel
ORDER BY lapse_rate_pct DESC
LIMIT 200 |
ULIPs sold through banks | Answered LQ3 | Correct |
| L09 | How many ULIPs were sold through banks to each age band from October 2024 to September 2026?
Reference querySELECT age_band, SUM(policies) AS policies FROM srv_life_cohorts WHERE product_type = 'ulip' AND channel = 'bancassurance' GROUP BY age_band ORDER BY age_band LIMIT 200 |
ULIPs sold through banks | Declined | Correct |
| L10 | What share of ULIPs sold through bancassurance went to buyers aged 55 or over?
Reference querySELECT SUM(policies) AS policies,
SUM(policies) FILTER (WHERE age_band IN ('55 to 64', '65 and over')) AS policies_55_plus,
ROUND(SUM(policies) FILTER (WHERE age_band IN ('55 to 64', '65 and over')) * 100.0 / SUM(policies), 1) AS share_55_plus_pct
FROM srv_life_cohorts
WHERE product_type = 'ulip' AND channel = 'bancassurance'
LIMIT 1 |
ULIPs sold through banks | Declined | Correct |
| L11 | How many new life policies did each channel sell this year?
Reference querySELECT channel, SUM(new_policies) AS new_policies FROM srv_life_policies_monthly WHERE month >= DATE '2026-01-01' GROUP BY channel ORDER BY new_policies DESC LIMIT 200 |
Declined | Correct | |
| L12 | Show life policy lapses and surrenders each month over the last 12 months.
Reference querySELECT month, SUM(lapses) AS lapses, SUM(surrenders) AS surrenders FROM srv_life_policies_monthly WHERE month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01' GROUP BY month ORDER BY month LIMIT 200 |
Declined | Correct | |
| L13 | What is the sum assured in force for each product type at the end of September 2026?
Reference querySELECT product_type, ROUND(SUM(sum_assured_in_force)) AS sum_assured_in_force FROM srv_life_policies_monthly WHERE month = DATE '2026-09-01' GROUP BY product_type ORDER BY sum_assured_in_force DESC LIMIT 200 |
Declined | Correct | |
| L14 | What is the average sum assured of new term policies by channel this year?
Reference querySELECT channel, SUM(new_policies) AS new_policies, ROUND(SUM(sum_assured) / SUM(new_policies)) AS avg_sum_assured FROM srv_life_policies_monthly WHERE product_type = 'term' AND month >= DATE '2026-01-01' GROUP BY channel ORDER BY avg_sum_assured DESC LIMIT 200 |
Declined | Correct | |
| L15 | What were the causes of death claims within the first year of the policy over the last 12 months?
Reference querySELECT cause_group, SUM(claims) AS claims FROM srv_life_death_claims_monthly WHERE duration_band = 'under 1 year' AND month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01' GROUP BY cause_group ORDER BY claims DESC LIMIT 200 |
Declined | Correct | |
| L16 | How many death claims were intimated more than 30 days after the death, by channel, over the last 12 months?
Reference querySELECT channel, SUM(late_intimations) AS late_intimations, SUM(claims) AS claims FROM srv_life_death_claims_monthly WHERE month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01' GROUP BY channel ORDER BY late_intimations DESC LIMIT 200 |
Declined | Correct | |
| L17 | Show death claims across all insurers week by week for the last 12 weeks.
Reference querySELECT week_start, SUM(claims) AS claims FROM srv_life_death_claims_weekly WHERE week_start BETWEEN DATE '2026-07-13' AND DATE '2026-09-28' GROUP BY week_start ORDER BY week_start LIMIT 200 |
Declined | Correct | |
| L18 | How many agents in each registry are active, suspended, lapsed or terminated?
Reference querySELECT registry, status, COUNT(*) AS agents FROM srv_life_agents GROUP BY registry, status ORDER BY registry, status LIMIT 200 |
Declined | Correct | |
| L19 | Which five insurers wrote the most new life annual premium this year?
Reference querySELECT insurer_code, ROUND(SUM(annual_premium)) AS annual_premium FROM srv_life_policies_monthly WHERE month >= DATE '2026-01-01' GROUP BY insurer_code ORDER BY annual_premium DESC LIMIT 5 |
Declined | Correct | |
| L20 | How many death claims on PoS policies were paid, repudiated or still under investigation over the last 12 months?
Reference querySELECT SUM(paid) AS paid, SUM(repudiated) AS repudiated, SUM(under_investigation) AS under_investigation FROM srv_life_death_claims_monthly WHERE channel = 'pos' AND month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01' LIMIT 1 |
Declined | Correct |
| No. | Question | Reaches | Offline | Last live run |
|---|---|---|---|---|
| O01 | What is the total sum insured in force across fire, property and marine cargo today?
Reference querySELECT SUM(policies_in_force) AS policies_in_force, ROUND(SUM(sum_insured)) AS sum_insured FROM srv_property_exposure_district LIMIT 1 |
Declined | Correct | |
| O02 | Which district and property type had the biggest rise in sum insured over the last eight weeks?
Reference querySELECT district, property_type,
COALESCE(SUM(sum_insured) FILTER (WHERE week_start = DATE '2026-09-28'), 0)
- COALESCE(SUM(sum_insured) FILTER (WHERE week_start = DATE '2026-08-03'), 0) AS rise_in_sum_insured
FROM srv_property_exposure_weekly
WHERE week_start IN (DATE '2026-08-03', DATE '2026-09-28')
GROUP BY district, property_type
ORDER BY rise_in_sum_insured DESC
LIMIT 10 |
Warehouse value building up in Thane | Answered OQ1 | Correct |
| O03 | Show warehouse sum insured in force in Thane week by week for the last 12 weeks.
Reference querySELECT week_start, ROUND(SUM(sum_insured)) AS sum_insured FROM srv_property_exposure_weekly WHERE district_code = 'MH-THN' AND property_type = 'warehouse' AND week_start BETWEEN DATE '2026-07-13' AND DATE '2026-09-28' GROUP BY week_start ORDER BY week_start LIMIT 200 |
Warehouse value building up in Thane | Declined | Correct |
| O04 | How many new warehouse policies started in Thane in the last eight weeks, and how much sum insured did they add?
Reference querySELECT SUM(new_policies) AS new_policies, ROUND(SUM(new_sum_insured)) AS new_sum_insured FROM srv_property_exposure_weekly WHERE district_code = 'MH-THN' AND property_type = 'warehouse' AND week_start BETWEEN DATE '2026-08-10' AND DATE '2026-09-28' LIMIT 1 |
Warehouse value building up in Thane | Declined | Correct |
| O05 | How many flood claims did each district in Assam report in the last two weeks?
Reference querySELECT district, SUM(claims) AS flood_claims FROM srv_property_claims_weekly WHERE state_code = 'AS' AND cause = 'flood' AND week_start BETWEEN DATE '2026-09-21' AND DATE '2026-09-28' GROUP BY district ORDER BY flood_claims DESC LIMIT 200 |
Flood in lower Assam | Answered OQ2 | Correct |
| O06 | How much has been claimed for floods in Assam since 21 September 2026?
Reference querySELECT SUM(claims) AS claims, ROUND(SUM(claim_amount)) AS claim_amount FROM srv_property_claims_weekly WHERE state_code = 'AS' AND cause = 'flood' AND week_start >= DATE '2026-09-21' LIMIT 1 |
Flood in lower Assam | Declined | Correct |
| O07 | Which five states hold the most insured value in factories?
Reference querySELECT state_name, ROUND(SUM(sum_insured)) AS sum_insured FROM srv_property_exposure_district WHERE property_type = 'factory' GROUP BY state_name ORDER BY sum_insured DESC LIMIT 5 |
Declined | Correct | |
| O08 | How many fire and property claims were reported for each cause of loss over the last 12 months?
Reference querySELECT cause, SUM(claims) AS claims FROM srv_property_claims_weekly WHERE week_start BETWEEN DATE '2025-10-06' AND DATE '2026-09-28' GROUP BY cause ORDER BY claims DESC LIMIT 200 |
Answered OQ4 | Wrong | |
| O09 | How much premium did each line of business write over the last 12 months?
Reference querySELECT line, ROUND(SUM(premium)) AS premium FROM srv_property_premium_monthly WHERE month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01' GROUP BY line ORDER BY premium DESC LIMIT 200 |
Declined | Correct | |
| O10 | Which five insurers wrote the most fire premium this year?
Reference querySELECT insurer_code, ROUND(SUM(premium)) AS premium FROM srv_property_premium_monthly WHERE line = 'fire' AND month >= DATE '2026-01-01' GROUP BY insurer_code ORDER BY premium DESC LIMIT 5 |
Declined | Correct | |
| O11 | What is the average premium per policy for each property type over the last 12 months?
Reference querySELECT property_type, SUM(policies_written) AS policies_written, ROUND(SUM(premium) / SUM(policies_written)) AS avg_premium FROM srv_property_premium_monthly WHERE month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01' GROUP BY property_type ORDER BY avg_premium DESC LIMIT 200 |
Declined | Correct | |
| O12 | What was the loss ratio for each line of business over the last 12 months?
Reference queryWITH c AS (SELECT line, SUM(claim_amount) AS claim_amount FROM srv_property_claims_weekly
WHERE week_start BETWEEN DATE '2025-10-06' AND DATE '2026-09-28' GROUP BY line),
p AS (SELECT line, SUM(premium) AS premium FROM srv_property_premium_monthly
WHERE month BETWEEN DATE '2025-10-01' AND DATE '2026-09-01' GROUP BY line)
SELECT p.line, ROUND(c.claim_amount * 100.0 / p.premium, 1) AS loss_ratio_pct
FROM p JOIN c USING (line)
ORDER BY loss_ratio_pct DESC
LIMIT 200 |
Declined | Correct | |
| O13 | How many fire and property policies are in force in Mumbai, by property type?
Reference querySELECT property_type, SUM(policies_in_force) AS policies_in_force FROM srv_property_exposure_district WHERE district_code = 'MH-MUM' GROUP BY property_type ORDER BY policies_in_force DESC LIMIT 200 |
Declined | Correct | |
| O14 | How has the total sum insured in force changed week by week over the last 12 weeks?
Reference querySELECT week_start, ROUND(SUM(sum_insured)) AS sum_insured FROM srv_property_exposure_weekly WHERE week_start BETWEEN DATE '2026-07-13' AND DATE '2026-09-28' GROUP BY week_start ORDER BY week_start LIMIT 200 |
Declined | Correct | |
| O15 | How many cyclone claims did each state report over the last 12 months?
Reference querySELECT state_name, SUM(claims) AS claims FROM srv_property_claims_weekly WHERE cause = 'cyclone' AND week_start BETWEEN DATE '2025-10-06' AND DATE '2026-09-28' GROUP BY state_name ORDER BY claims DESC LIMIT 200 |
Declined | Correct | |
| O16 | Which ten map cells hold the most insured value?
Reference querySELECT cell_lat, cell_lon, SUM(policies_in_force) AS policies_in_force, ROUND(SUM(sum_insured)) AS sum_insured FROM srv_property_exposure_grid GROUP BY cell_lat, cell_lon ORDER BY sum_insured DESC LIMIT 10 |
Declined | Wrong | |
| O17 | Show the sum insured of new business in Maharashtra week by week over the last eight weeks.
Reference querySELECT week_start, ROUND(SUM(new_sum_insured)) AS new_sum_insured FROM srv_property_exposure_weekly WHERE state_code = 'MH' AND week_start BETWEEN DATE '2026-08-10' AND DATE '2026-09-28' GROUP BY week_start ORDER BY week_start LIMIT 200 |
Declined | Correct | |
| O18 | What share of all insured value in force is in Maharashtra?
Reference querySELECT ROUND(SUM(sum_insured) FILTER (WHERE state_code = 'MH')) AS maharashtra_sum_insured,
ROUND(SUM(sum_insured)) AS sum_insured,
ROUND(SUM(sum_insured) FILTER (WHERE state_code = 'MH') * 100.0 / SUM(sum_insured), 1) AS share_pct
FROM srv_property_exposure_district
LIMIT 1 |
Declined | Correct | |
| O19 | How many marine cargo claims were there by cause over the last 12 months?
Reference querySELECT cause, SUM(claims) AS claims, ROUND(SUM(claim_amount)) AS claim_amount FROM srv_property_claims_weekly WHERE line = 'marine_cargo' AND week_start BETWEEN DATE '2025-10-06' AND DATE '2026-09-28' GROUP BY cause ORDER BY claims DESC LIMIT 200 |
Declined | Correct | |
| O20 | Which five districts in Assam hold the most insured value?
Reference querySELECT district, ROUND(SUM(sum_insured)) AS sum_insured FROM srv_property_exposure_district WHERE state_code = 'AS' GROUP BY district ORDER BY sum_insured DESC LIMIT 5 |
Flood in lower Assam | Declined | Correct |
Offline, a question is answered only when it names the same things as a golden question: the vertical, places, the measure, the breakdown and the period. Answered means the golden query gives the same figures as the reference. The code beside it, such as Q11 or HQ1, is the golden question used. Declined means Ask IIB says it cannot answer offline rather than guess. Matching took 0.35 seconds for all 80.
How Ask IIB reads four verticals
A vertical hint. Ask IIB may read 27 tables across four verticals. The prompt lists one vertical's tables first, with every column described, and the rest in short form. The hint comes from the words of the question, then the tab on the Ask page, then the asker's role.
The query checks. A query may read only the serving views, whose names start srv_, and never a file or another table. The checks also catch the mistakes that give plausible but wrong figures: adding up a count taken on one day across weeks or months (policies in force, sum insured, sum assured), averaging a ratio such as cost against peers, and the motor averaging mistakes they caught before.
Who sees what. Leadership sees state-level figures. Ask IIB does not give Leadership answers at district, hospital, agent or map-cell level, just as it holds back RTO detail in Motor. The Analyst sees counts under 10 as <10. Insurers show as code names to every role but Data Operations.
The definitions. The glossary now has terms for Health, Life and Other Lines, such as cashless claim, length of stay, early death claim, PoS agent and sum insured. Each points at the columns it describes, and an answer shows the ones it used.
Personal data is masked before any person or the AI sees it. The AI flags and drafts. A person decides. This use needs no outside data, so nothing on it is mock.