Chief Information Officer / India
The recommendation changed when the missing obligation became clear
- Technology employees
- 68
- Five heads; twelve contractors counted separately
- Annual technology operating expense
- ₹17.8 crore
- Separate from ₹5.9 crore approved change investment
- Systems and technology experience
- 34 years
- Operations, integration, applications and CIO leadership
Decisions & consequences
012019: nine historical report types outlived the migration milestone
As head of IT, I endorsed retirement of an older application after its current transactions moved to a replacement. Business review then identified thirty-six required historical report types. Twenty-seven were supported by the new environment, while nine still depended on the old one. I had treated the current transaction migration as evidence for a broader conclusion it did not establish. The records had continuing business uses that a successful new-workflow demonstration had not tested.
I paused the retirement recommendation and worked with business, records and legal owners to define the remaining retrieval needs and access conditions. The CEO approved ₹86 lakh over fourteen weeks for the retrieval programme, with ₹18 lakh of transition support expense recorded separately. The work concerned controlled access and usable evidence, not permission to retain or delete every record indefinitely. Applicable retention and disposal decisions remained with the authorised owners.
Six of the nine outstanding report types passed the defined retrieval review. Three remained unfinished at that point, so the old environment retained a controlled purpose and the full shutdown was not approved. We did not claim that copying files alone made the reports interpretable, or that a majority pass established readiness for all historical work. The result also did not produce an asserted saving from retirement, because the service had not yet been fully withdrawn.
The correction changed the order of the decision. Instead of treating the retirement date as proof that the obligation had moved, we required evidence that the defined historical work could still be performed. My responsibility was to acknowledge that the earlier recommendation had left this question out. I learned to examine the business questions that survive a transaction migration and to bring their continuing cost into the board account. A newer platform can be useful while an older obligation still needs a funded and authorised route.
2023: four connectors stayed outside a release after the logging review
A pre-release review examined 180 synthetic diagnostic scenarios across a set of insurer-service connectors. Forty-three emitted identity fields that responsible information owners considered unnecessary for the diagnostic purpose. The other 137 did not show that reported issue under the selected conditions. The material was synthetic; the finding did not establish that real customer data had been exposed or that a security incident had occurred. It did show that the proposed design could collect more business detail than the technical question required.
I brought security, privacy and operating owners into a ₹72 lakh, twelve-week correction programme. They defined the permitted event content and the evidence needed for useful investigation. The technical team changed the relevant logging paths rather than simply hide fields in the display after they had already been collected. Required operational and security monitoring remained part of the design; the response was not to disable all logging to make the test quiet.
Thirty-nine of the forty-three affected scenarios passed the defined recheck. Combined with the 137 without the original finding,176 passed that content check, while four remained unfinished. The four associated connectors were withheld from the release until the relevant work could be completed. This account does not invent their later outcome or describe the selected set as proof of universal privacy, anonymity or legal compliance. Access controls, retention, transmission and ongoing monitoring still required their own review.
The decision cost time because the programme could not claim that its largest completed portion represented the whole release. I considered that distinction necessary. A diagnostic record should support an authorised question without inheriting every field that happens to pass through the application. The case taught me to ask why a field is collected at the source, not only who can see it later, while preserving the technical evidence the organisation legitimately needs to operate and investigate its systems.
2025: the quiet IT window overlapped active hospital-service work
The proposed maintenance plan used a two-hour Saturday window from 02: 00 to 04: 00 because it was outside branch-office hours. Operating review of eight selected Saturdays showed 360 of 2,000 hospital-service messages arriving during that window. The count did not establish the clinical urgency or outcome of any individual message, but it did challenge the assumption that the supported service was effectively quiet. IT had chosen a convenient time without a sufficient account of the operating work continuing through it.
I worked with Rakesh Qezrlavi and authorised service owners on staged maintenance, an additional service instance and a controlled fallback route. The CEO approved ₹96 lakh over fourteen weeks, with ₹21 lakh annual running expense shown separately. The programme did not give technology staff authority to determine coverage or clinical decisions. Operations needed to define how legitimate service work would be handled while the technical change proceeded, including the responsibilities and limits of the fallback.
Four representative maintenance exercises were reviewed. Three completed the agreed service sequence, while one failed the fallback step. We deferred unattended rollout of that route and kept the unresolved scenario in the readiness account. The result was not presented as a seventy-five-percent assurance that the whole arrangement was safe, and the extra service instance was not described as a promise of uninterrupted availability. The failed exercise required further work before the broader commitment could be supported.
My decision was to change the maintenance proposition rather than ask operations to accept the office timetable as a complete answer. The additional cost and delayed rollout were part of the recommendation. I learned that a technical team’s off-hours label can conceal someone else’s active service period. A credible plan needs evidence about that period and a tested operating response, with the outstanding limitation visible to the people who must authorise the change.