Chief Information Officer India

The system's answer needed a better business question

Technology responsibility through three leads: 25 colleagues; ₹4.8 crore FY 27 expense; separate ₹1.9 crore change programme
Business applications and language-service information: 28 years; Career from July 1998; CIO from March 2022
Provisional allowance for a first external board role: 23 days; Preparation, consent and actual calendar still to be examined

Decisions & consequences

Replacing my latest-file rule with the script's approved picture relationship

In 2022 I supported a workflow that selected the most recently uploaded script for the next recording session. The rule was easy to explain and reduced the time coordinators spent comparing filenames. It did not establish which approved picture cut the script accompanied. Anjali challenged the arrangement after a session team found that the selected material belonged to another version. The application had followed its instruction; my approval had made that instruction too broad.

We examined 120 jobs. One hundred and four already had an agreed script-to-cut relationship. Sixteen required further review. The programme sponsor wanted to preserve automatic selection and rely on the recording team to notice mismatches. Production colleagues argued that this would move an avoidable ambiguity into booked time with performers and reviewers. I supported a ₹79 lakh programme to examine the relationships, revise the workflow and make the appropriate approval visible before a session was treated as ready.

Ten of the sixteen cases were reconciled by the responsible production and client owners. Six remained held. The resulting account showed 114 supported relationships and six unfinished decisions in the reviewed cohort. A mismatch did not necessarily establish that the translation was poor, and a held job was not automatically lost work. The question concerned which material was authorised for the intended cut. Language professionals and client owners retained that decision; technology recorded and applied it.

The revised process required more coordination before a session and removed a simple default that colleagues had found convenient. Some new uploads could no longer move directly into the next activity. I accepted that friction because the earlier convenience depended on an assumption the system could not establish. We kept the six unresolved jobs visible instead of assigning them the newest file simply to close the review. The programme cost was not presented as a calculated saving from hypothetical re-recording.

I would ask the version-relationship question during workflow design next time. A timestamp remains useful for understanding when a file arrived. It should not become a proxy for the professional and customer acceptance that determines how the file may be used. The correction made that distinction operational, with my responsibility for the earlier rule acknowledged rather than transferred to the people who had followed it.

Separating a dialogue track from the caption service our system claimed

In 2023 our service register marked eighty films as having captions available. Qualified review found that sixteen met the agreed full-caption service criteria, while sixty-four had dialogue-only tracks. Those tracks could serve a useful purpose, but the label implied a broader account of relevant sound than they supplied. The issue was not solved by confirming that a text file could be downloaded. We needed the service description to match the work actually present.

The commercial sponsor worried that splitting the status would make the catalogue appear less complete and require awkward customer explanations. Anjali and accessibility practitioners wanted the distinction visible before a user relied on the label. I backed a ₹44 lakh programme covering workflow changes, professional review and the additional work needed for the examined group. Technology represented the intended service and acceptance evidence. Qualified practitioners determined what the captions needed to convey and whether the agreed work was suitable.

Forty-eight of the sixty-four dialogue-only cases acquired captions accepted through that review. Sixteen remained pending. Together with the original sixteen, sixty-four films in the eighty-film cohort now had the supported service status. We did not describe the result as universal compliance with every accessibility requirement or as proof that every user would encounter the track correctly on any device. Translation, timing, relevant non-speech information and actual delivery conditions needed their appropriate assessment.

The correction reduced the apparent breadth of the initial offer and required funded work that a single completed field had concealed. Commercial colleagues had to explain which service was available and which remained in preparation. I wanted that explanation to be precise without dismissing the value of dialogue subtitles. Different audiences and commissions may need different outputs. The error was allowing one convenient system label to make those outputs appear interchangeable.

My change for future programmes is to establish the service vocabulary with practitioners before designing the status list. A database can store a confident label cheaply. The company still has to supply the professional work that makes the label meaningful. The case gave us a more defensible account of the reviewed films and sixteen explicit unfinished items, rather than a broad accessibility success story unsupported by the scope of the examination.

Testing the text people saw after the source had arrived intact

A 2024 delivery examination considered seventy-two predefined text-and-environment cases: six representative clips in each of three language-script sets, viewed under four agreed playback conditions. Source text had transferred without alteration, but language reviewers found that fourteen cases did not meet the agreed display criteria. Fifty-eight did. Missing or inadequately presented characters in particular combinations meant that faithful storage was not enough to establish a usable delivery route.

The technical sponsor wanted to accept the transfer because the source content was intact and the ordinary demonstration worked. Production colleagues asked for the actual displayed result to be assessed by people able to read it. I supported a ₹71 lakh programme to investigate the environment, font and shaping conditions with qualified technology and language owners. The exercise examined Odia, Assamese and Hindi work within the specified conditions. It did not assume that every language or receiving device behaved alike.

Ten of the fourteen failing cases met the agreed criteria after correction. Four remained outside the approved route, with supported alternatives retained for the corresponding work. The final account was sixty-eight accepted cases and four exceptions in the seventy-two-case examination. These were test combinations, not seventy-two customers, films or distinct languages. Passing a display case did not certify the translation, the complete accessibility of the film or all future playback conditions. Those limits stayed with the result.

The choice prolonged support for more than one delivery arrangement. It also required language-review time that a technical file comparison would have avoided. I regarded that as part of establishing the service, not a reason to ask the reviewers to accept a display they could not defend. Technology investigated the rendering conditions while the relevant professionals retained judgement about the text people encountered. The programme was not described as correcting the writing system itself or as a universal font solution.

I would include representative scripts and receiving conditions earlier, before a successful demonstration becomes the expected answer for the whole service. A character sequence and its visible presentation are connected but not identical. The CIO's role is to make that distinction part of enterprise acceptance, with a bounded claim and a supported exception where the current route still falls short.

Correcting a report that counted each contract once for every language

In 2025 an internal management report combined eighty commissions with their language-version records. Each of the eighty had exactly three versions in this examined cohort, making 240 version records. The contracted fees totalled ₹11.4 crore at commission level. The report's join repeated each commission amount three times and displayed ₹34.2 crore. The arithmetic followed the joined rows, but the result did not describe additional contracts or money the company was entitled to receive.

The operating sponsor valued the version count because it represented work across languages and did not want the correction to erase that activity. Finance objected to the repeated monetary amount. I supported a ₹52 lakh programme to repair the model, review the affected report and establish ownership of the measures. We preserved both levels: eighty commissions with their actual combined contracted value, and 240 language versions whose production status could be examined separately. A version would carry a separately agreed fee only where the relevant agreement actually established one.

The corrected report showed ₹11.4 crore for the examined commission cohort. The ₹22.8 crore difference from the earlier display was duplicated reporting, not a saving, cash recovery or revenue reduction caused by the technology programme. The underlying agreements and finance ledger had not been changed by this internal report error. Finance retained accounting treatment and any recognition decisions. We did not convert the contracted fee value into a claim about recognised revenue, profit or collections.

The programme required report owners to explain their measures before a dashboard combined them. That slowed some requests for immediate new views and made a formerly attractive total unavailable. It also left the useful version count intact, addressing the operating sponsor's concern that the language work should remain visible. A data correction should preserve valid distinctions rather than choose one team's preferred total and discard the other team's legitimate question.

I would now require an explicit account of the relationship between a commercial commitment and its deliverables before approving a combined report. The case was not a failure of multiplication. It was a failure to preserve what could properly be multiplied. Technology needed finance and operating owners to define that boundary, and the final explanation needed to say clearly which business facts had remained unchanged while the presentation was corrected.

Discuss a leadership challenge