Leadership review
Mira Zheqerni
Chief Information Officer · India
Highlights
82
Technology employees
Six heads; sixteen contractors counted separately
₹22.6 crore
Annual operating expense
₹8.3 crore approved change investment is a separate authority
28 years
Systems career
Analysis entry July 1998; CIO since January 2021
A letter from Mira Zheqerni
A component can pass its test while a colleague still cannot finish the work. I ask what the evidence actually demonstrates, which dependency is outside the test and how the remaining limitation will be handled. A narrower honest result is more useful than a successful label that asks operations to discover the missing condition later.
The database had been restored, and I told management that the recovery test was successful. In 2019 that seemed a reasonable description of the work we had just demonstrated. It was not a sufficient description of the insurance service. When colleagues attempted 240 invented representative tasks, 192 completed and 48 could not retrieve attachments stored outside the database. We had recovered an important component while leaving a dependency outside the test, and my summary had turned the component result into a stronger claim.
Mira Zheqerni
Chief Information Officer
Strategic priorities
01
02
03
04
Performance
Priority outcome 01
In 2019 I described a database-restoration exercise as a successful recovery test. The database had returned, but the service also depended on attachments in a separate document store. When colleagues attempted 240 invented representative tasks, 192 completed and forty-eight could not retrieve the required material. The narrower technical result was real; the broader statement I made from it was not justified. We had tested an important component without testing the work that management understood the result to represent. I withdrew the service-readiness claim and worked with operating colleagues to define a broader exercise. Qualified technology owners included the document store and the identity, permission and network dependencies required by the tasks. The programme cost ₹1.4 crore, with ₹18 lakh annual exercise and support expense retained separately. We did not describe the annual amount as a saving or assume that a one-time repair removed the need to rehearse recovery as the service changed. Retesting the same 240 tasks produced 231 passes and nine still needing an assisted route while repair continued. The nine remained in the denominator and in the operating account. They were not relabelled successful because someone could work around the problem with help. This was a rehearsal using invented tasks; it did not establish that real customer records were lost, that a claim decision failed or that the insurer had achieved a universal recovery-time promise. The result applied to the scope actually exercised. The correction changed how I approve a readiness statement. I now ask which complete task was demonstrated, what supporting material it used and which dependency remained outside the exercise. A component result is useful when it is accurately named. It becomes misleading when it inherits the meaning of a larger service without the corresponding evidence. My responsibility was not merely to expand the technical checklist, but to correct the management account I had supplied and fund the continuing work required to keep that account credible.
Priority outcome 02
In 2022 a review examined seventy-four offline lookups of policy-status information at branches. Eleven displayed a snapshot older than the latest authorised source; sixty-three were current under the review conditions. The interface labelled the view as current even when the connection was unavailable. A cached record could be useful, but the presentation did not explain the condition under which its freshness could no longer be confirmed. That was a problem in the system's account of what it knew. We funded ₹58 lakh of change and ₹11 lakh annual support across the eighteen-branch scope. The revised view displayed snapshot time and connection state, with a route to authorised staff when a current verification was needed. Operations helped define the explanation and the next practical step. Technology did not infer whether a policy's terms had changed or permit a colleague to alter a contractual consequence because the screen was disconnected. The relevant insurance professionals retained their decisions. The review did not establish eleven incorrect policies, eleven rejected claims or a loss suffered by eleven customers. It compared the displayed snapshot with the latest authorised source in a defined population. Nor did the change make all future cached information current; a disconnected service still has a limit. The response made that limit visible and supported an appropriate verification route. We retained the support expense because a label alone would not help a colleague who needed a current answer to continue the work. The difficult choice was to give up the reassuring simplicity of one status. A visitor could see that a record was available while also seeing that its currentness had not been confirmed. Some colleagues initially regarded the extra explanation as unnecessary complexity. I supported it because the complexity already existed in the connection state; hiding it transferred the interpretation problem to the person using the screen. The case taught me to ask what a service may truthfully say while disconnected, rather than treat offline display as an inferior copy of the connected interface with the same claims attached.
Priority outcome 03
In 2023 certificate expiry interrupted two of twelve partner interfaces for ninety-four minutes. The incident concerned those connections; the case does not establish a breach or that every customer was affected. A subsequent review examined thirty-six service certificates. Twenty-seven had an accountable owner and nine did not. Technical records contained dates, but a recorded date was not the same as a person accepting the continuing work needed to renew and test the connection appropriately. We allocated ₹48 lakh to the inventory, ownership and renewal-exercise response. A reporting release moved by two weeks to make capacity available. Technology and service colleagues assigned accountable owners across all thirty-six certificates and established renewal testing through the authorised workflow. The public account does not expose credential details or prescribe production commands. The work required both technical responsibility for the connection and an operating understanding of what interruption meant for the service using it. The delay was a real tradeoff. Reporting colleagues had expected their release and needed to understand why another responsibility had taken priority. I did not describe certificate administration as a trivial background task that could be absorbed without consequence. Equally, assigning ownership did not prove that future interruption was impossible. The company still needed the work to be performed, tested and reviewed as services and partners changed. A named owner is the beginning of accountability, not evidence that every later renewal has already succeeded. The incident changed how I examine a completed integration. Implementation can leave behind a recurring obligation that does not fit neatly inside the original project team. If that obligation is not accepted by a continuing owner, the service depends on somebody noticing the date and finding time when it arrives. My contribution was to connect the technical inventory with the people and resources needed to act on it. The ninety-four-minute interruption remained part of the account, rather than being overshadowed by a larger claim that the subsequent review had solved all partner-service risk.
Priority outcome 04
A 2024 testing proposal relied on 650 production-derived fixture cases. The programme team valued their resemblance to real work and wanted to preserve the count as evidence of coverage. Data owners and role specialists needed to examine two different questions: what material the tests were permitted to use, and whether the cases represented the requirements being tested. Neither question could be answered simply by saying that the set was large or that it had come from actual operations. We replaced the proposed set with 420 synthetic illustrative cases and 230 restricted metadata-only cases reviewed by the data owners. The latter were not labelled anonymous or unrestricted. Business structures and relationships can remain sensitive even when direct personal details are absent, so access conditions stayed attached to that part of the set. The case contains no actual policyholder information or production fixture content. Qualified owners retained decisions about permitted use and the professional requirements relevant to the real system. The work cost ₹86 lakh and extended the release plan by three weeks. Role owners examined requirement coverage rather than accept a restored total of 650 as proof that the replacement set was adequate. Synthetic cases needed to exercise meaningful situations, and restricted metadata cases needed a clear reason for inclusion. We did not claim a legal-compliance certificate, perfect anonymity or complete coverage of every possible service condition. The result was a differently constituted test set with explicit limits, not a universal assertion about privacy or testing quality. The tradeoff was time and design effort in place of convenient reuse. It was possible to preserve a reassuring test count while failing to improve either representativeness or information handling, so I asked the team to explain those purposes separately. A CIO should be able to defend why material belongs in an exercise and what the exercise demonstrates. The decision taught me to examine the composition of test evidence before celebrating its volume, while avoiding the opposite error of assuming that invented data automatically supplies every scenario the business needs to examine.
Priority outcome 05
A 2025 pre-release pilot examined sixty illustrative policy-summary documents. Seventeen Hindi-script documents had clipped text in PDF export even though the browser preview was legible. Forty-three did not show that reported defect under the test conditions. The preview had become the convenient demonstration of readiness, but it was not the document the recipient would actually use. We needed to examine the output path rather than infer its quality from a different presentation of the content. We approved ₹72 lakh for document-export and layout review, with ₹14 lakh annual regression and support expense retained separately. Technology worked with language and authorised content reviewers on the actual exported pages. The exercise used illustrative material, not real policyholder records. Content and legal owners retained approval responsibilities; correcting clipping did not establish that a translation was equivalent, that every policy statement was valid or that the whole service met all applicable requirements. Retesting the same sixty documents produced fifty-seven passes and three withheld from release. An assisted verified-document route remained for the unfinished scope. We kept those three in the account instead of removing them from the test set to claim complete success. The result did not show that every future document would export correctly, and the continuing regression expense recognised that templates, content and output behaviour could change. No claim consequence or customer loss was inferred from this pre-release finding. The decision widened what we meant by demonstrating the service. A reviewer should see the artefact the intended reader receives, through the path in which it will be used. An attractive browser preview is evidence about that preview. It cannot establish the layout of an exported page merely because both began with the same source. I wanted the release discussion to retain the actual output, the remaining failures and the authorised review needed before use. The case is a reminder that a technical demonstration can be persuasive while stopping one step before the place where the service must work.
Stakeholder voices
Mira retained the nine unfinished recovery tasks in the account. She did not let an improved result erase the support and repair still required.
She asked what the disconnected screen could honestly tell a colleague and funded the verification route, rather than treating the extra explanation as a substitute for operating support.
Outlook
A recovery rehearsal passed at the component level while forty-eight tasks could not retrieve their attachments. The correction changed both the test scope and the management claim.
Read the essayA partner-interface interruption exposed recurring work that had outlived its project. The response connected the certificate inventory with people accountable for renewal and service consequences.
Read the essayReplacing production-derived material required separate judgements about permitted use and meaningful coverage. Restoring the number 650 did not answer either question by itself.
Read the essayAn invitation
A first outside non-executive contribution after employer agreement and qualified company-specific assessment. I have no statutory directorship and hold 0.4% inside the employee vehicle. Databank registration and proficiency assessment were completed in March and September 2026 in this fictional account; twenty-two days remain provisional.
office@mirazheqerni.example