CAG’s Big Reset: AI Audits, Scrutiny and a Push for Accountability
CAG Office India (Image official website)
The 32nd AG Conference signals a bold tech-driven overhaul—but the real test is whether India sees quicker audits, fewer leakages and measurable public impact.
By P SESH KUMAR
New Delhi, November 20, 2025 — The glittering 32nd Accountants General Conference in the prestigious Bharat Mandapam in New Delhi marks an important inflection point for SAI India. Under the banner “Leading Change & Reaffirming Values: Trust, Innovation, Sustainability, Accountability,” the CAG has unveiled a package of reforms that sit squarely within the Strategic Plan 2023–30: a Centre of Excellence for Financial Audit in Hyderabad, a research awards scheme with ICSSR, a decisive tilt towards data-informed audits, and even an indigenous CAG-LLM to power AI-enabled analytics.
The rhetoric is ambitious: move from sample-based post-mortems to full-population, real-time, AI-assisted scrutiny; embed ESG, Ease of Living and Ease of Doing Business, and Foundational Literacy and Numeracy into the audit lens; and turn Accounts Offices from passive book-keepers into “goldmines” of data for auditors.
Yet, as we all know, ambition is not the same as impact. In a country with tens of thousands of audited entities and long-entrenched habits of delay, non-compliance and legislative indifference to audit, the real test is not whether CAG can build a data lake or an LLM, but whether citizens can see faster audits, fewer leakages, better programmes and a visible rupee value of CAG’s work.
The new strategic and digital initiatives are directionally sound, but that they will remain a glossy superstructure unless they explicitly confront four hard problems: timeliness of audit, chronic non-submission of Action Taken Notes (ATNs), the near-impunity with which PAC/COPU recommendations are ignored, and the opacity of CAG’s own value-for-money contribution.
Also, the ubiquitous question- has CAG formally abandoned the all-India Performance Audits which used to examine the soundness and efficacy of government policies and whether these served public interest transparently enough.
The way forward lies in coupling the Strategic Plan 2023–30 and AI Strategy Framework with concrete, public performance metrics: time-bound audit cycles including all India Performance Audits; quantified savings and revenue gains; graded recognition of well-performing entities; and digital tracking of ATNs and PAC recommendations modelled on the Centre’s Audit Para Monitoring System, but pushed down to every state and large local body.
The 32nd AG Conference press release is not just another ceremonial communiqué; it is effectively a status report on how far SAI India has travelled since the Strategic Plan of SAI India 2023–30 was adopted and how far it still has to go. The Annual Report 2023–24 describes that Plan as an effort “to transform, standardise and institutionalise our processes” over a seven-year horizon, with ten strategies under four broad goals shaping everything from risk-based audit to outreach and internal capacity.
The conference theme – “Leading Change & Reaffirming Values: Trust, Innovation, Sustainability, Accountability” – deliberately weaves that strategic vocabulary into the day-to-day marching orders for Accountants General. It is no accident that the headline announcements are all institutional and technological: a national Centre of Excellence for Financial Audit at Hyderabad; a research awards scheme with ICSSR on public finance and accountability; and a strong directive to break down the wall between Audit and Accounts so that vouchers, sanctions and challans become the raw material for “data informed” audits.
On paper, this is exactly what many critics of CAG have been demanding: more specialisation, more research, and more intelligent use of the unimaginably rich data already lying inside government treasuries and PAOs. The idea (not at all a new one) that Accounts Offices are “goldmines of invaluable financial data” and that the boundary between Audit and Accounts should be a “seamless bridge” rather than a Chinese wall is both conceptually correct and long overdue.
A Supreme Audit Institution that continues to rely on small samples and paper vouchers in 2025, despite having access to huge data and soft copies, when ministries themselves are running AI-enabled tax engines, would rightly be dismissed as an anachronism.
The digital thrust is fleshed out by two important strands. First, the conference highlights the development of an in-house Large Language Model – the “CAG-LLM” – and pilots where AI has been deployed in audit for anomaly detection, risk scoring and pattern recognition.
Second, the separate Artificial Intelligence Strategy Framework issued in April 2025 lays out, in considerable detail, how AI should be embedded “into the core of audit operations” across planning, execution, reporting and follow-up, while also setting guardrails on ethics, privacy and model governance. Together, these documents signal that CAG may not be treating AI as a toy for conferences, but is at least attempting to define a serious institutional roadmap.
Equally noteworthy is the thematic re-orientation: the press release explicitly ties future audits to ESG criteria, citizen-centric priorities like Ease of Living and Ease of Doing Business, and foundational literacy and numeracy, and announces work on a “comprehensive assessment matrix” for autonomous bodies.
This is consistent with the commitment, in the Strategic Plan 2023–30, to align audit with SDGs and broader governance outcomes rather than treat departments as silos. It also reflects global trends among SAIs, where performance and value-for-money audits increasingly look at cross-cutting themes instead of narrow compliance.
The challenge, however, is not in writing down the right buzzwords. It is in doing this at credible scale in a country where the audit universe spans Union ministries, PSUs, autonomous bodies, state governments, local bodies, schemes and special purpose vehicles numbering in the tens of thousands, many with weak IT systems and patchy data quality.
An AI-enabled, data-informed audit architecture is only as good as the underlying ledgers. In practice, that means the CAG must confront mundane but critical constraints: legacy treasury software (different versions in different states), manual postings in smaller states and districts, inconsistent chart of accounts, missing metadata, and serious human-resource gaps in IT and analytics at the field level. The AI Strategy Framework itself candidly acknowledges that adoption will require “infrastructure and resource” investments and sustained training and capacity building before AI can move beyond a small set of early use-cases.
There is also the question of audit relationships. Turning Accounts Offices into data mines for Audit is conceptually sound, but if handled clumsily it will be perceived as yet another surveillance layer and will deepen the instinctive defensiveness many departments already display. There is a successful pilot in at least one state in Rajasthan-it would appear-but many in CAG are unaware of basic details. Also why did it take so long? Voucher level computerization was introduced almost 25 years back with much fanfare.
To “take thousands of audited entities along” on the digital journey, CAG will need a carrot as well as a stick: standardised APIs and templates that reduce their reporting burden, collaborative data-cleaning exercises, and, crucially, a visible commitment to recognise and highlight good practice in audit reports rather than only parade failures.
The conference press note speaks of CAG as a “partner in achieving national priorities” and as an “enabler of good governance,” but this partnership language must be reflected in the tone and structure of reports on the ground.
One question goes to the heart of the matter: where are the demonstrable proofs of value addition? Thus far, CAG has not institutionalised a public metric that resembles, for example, the UK National Audit Office’s long-standing practice of reporting the estimated pounds saved or benefits generated for each pound spent on the NAO’s budget, or GAO’s regular claims about billions of dollars in savings and revenue enhancements attributable to its work.
Individual Indian audit reports, of course, quantify specific recoveries or avoided losses, but there is no consolidated “audit dividend” table that tells Parliament and citizens, in one glance, what this ocean of work has actually achieved for the exchequer over a decade. Nor is there a systematic practice of highlighting positive deviance – departments or states that used audit to improve their schemes, plug leakages and redesign processes.
On timeliness, the gap between rhetoric and reality is even starker. The whole promise of data-driven, AI-enabled audit is that it allows for early warning and near real-time red flags, not just autopsies years after the patient has died.
Yet many CAG reports still appear with a time lag that makes their findings more suitable for academic analysis than immediate corrective action, and the audit reports themselves routinely document cases where Action Taken Notes (ATN) promised years ago have not translated into actual change.
In one recent Union report, for example, a recommendation on enhancing a benefit for mine workers remained unimplemented even eight years after the ministry committed in an ATN to “revisit the matter.”
In another, the failure to follow up on land acquisition and deposit utilisation persisted despite ATN assurances. An AI-powered audit factory that still produces slow-motion follow-up will be rightly accused of putting new wine into old skins.
The ATN and PAC/COPU problem is precisely where the concern about “almost 10,000 ATNs (reflected in CAG’s own activity reports) where even first replies were not furnished” bites.
There is no single official national figure in the public domain that aggregates pending ATNs across all legislatures, but multiple state and sectoral reports show very large backlogs of explanatory notes and ATNs, with some paras pending for a decade or more.
The Centre’s Audit Para Monitoring System (APMS), hosted by the Controller General of Accounts and NIC, was explicitly designed as a web-based tool to track submission of ATNs on CAG paras and PAC recommendations in real time, and to reduce this backlog.
More recently, the CAG’s field offices have pushed state governments and assemblies—Delhi is a live example—to adopt APMS-style portals so that PACs can monitor which departments are responding and which are dragging their feet.
But here again, the underlying constraint is political and behavioural, not technological. The CAG can recommend, nudge and design dashboards, but it cannot force a legislature to convene PAC meetings or compel ministers to treat PAC recommendations as binding.
Unless the Strategic Plan 2023–30 is accompanied by an explicit outreach strategy to Speakers, PAC chairs and Chief Ministers, and unless the CAG is willing to publish uncomfortable league tables that name departments and states with the worst ATN compliance, the ATN backlog will simply migrate from physical files to a digital limbo.
The 32nd AG Conference press release speaks of “a sharpened focus on outcomes that matter to citizens” and of CAG as a “dynamic and future-ready audit institution,” but it is silent on any new institutional compact with legislatures to close the feedback loop.
Finally, there is the question of readability and perceptibility of audit reports. Over the last few years, CAG has experimented with more attractive layouts, special compendia, thematic reports, and public-facing initiatives like Audit Diwas.
However, the core Union and State reports are still written primarily for accountants and insiders: long paragraphs of technical text, dense tables, and little visualisation of the actual impact on citizens or the exchequer. The contrast with many NAO and GAO products, which increasingly use infographics, case narratives and prominent “Key Facts” boxes, is striking.
If CAG truly wants to demonstrate its economic value added, it must not only modernise how it audits but also how it tells the story: citizen-level vignettes, before-and-after examples, clear quantification of leakages plugged and services improved, and a standard section in each report summarising “Estimated Financial Impact” over a five-year window.
CAG’s Entertainment Sector Audit: A Bureaucratic Hygiene-closing
The way forward, therefore, has to be more concrete than a general endorsement of data and AI. The Strategic Plan 2023–30 and the AI Strategy Framework need to be translated into a handful of public, measurable promises.
One set of promises should relate to timeliness: average time from the close of a financial year to tabling of key reports; number of ongoing “concurrent” or near-real-time audits; and average lag between PAC discussion and verified implementation of recommendations, all disclosed annually.
A second set should cover value addition: total recoveries, avoided losses, additional revenues and efficiency gains credibly linked to CAG findings, published once a year, together with examples where audit led to changes in law, rules or programme design.
A third should speak to behaviour: positive recognition for audited entities that respond quickly, implement recommendations and innovate, perhaps through a formal “Audit Excellence” chapter or awards, so that audit is seen not only as a threat but also as an opportunity.
None of this is beyond the reach of the current leadership. The Centre of Excellence at Hyderabad, the ICSSR research awards, the CAG-LLM and the AI Strategy Framework show that the institution is willing to invest in new capabilities and ideas.
The real question is whether it will use those capabilities to challenge its own comfort zone: to move from pilots to scaled-up practice, from eloquent conferences to hard numbers, and from a culture that sees its job as “issuing reports” to one that judges itself by the rupees protected, policies improved and citizens’ lives quietly made better.
Only then will the Strategic Plan 2023–30 and the digital agenda read not as aspirational brochures, but as the backbone of a genuinely transformed SAI India.
(This is an opinion piece, and views expressed are those of the author only)
Follow The Raisina Hills on WhatsApp, Instagram, YouTube, Facebook, and LinkedIn
About The Author
