EssayApril 20, 2026
The Office That Was Never Built
Third in the Yukti series. Same argument: across five very different domains, the surface is observable but the reasoning behind it has never been written down. This one is about the office of an Indian MP. Someone gets sent to Parliament from a constituency. What the office is supposed to do once there has never been formally specified.
Previously in this series: The Patch That Was Never Named
The Constitution (One Hundred and Thirty-First Amendment) Bill was introduced on 16 April 2026 and defeated the following day, 298 votes to 2301. It will return in some form. The debate it triggered has already exposed, to anyone willing to notice, how little we actually know about the work the current 543 Members of Parliament do.
The previous essay in this series argued that the 1976 freeze on Lok Sabha seats was an unnamed architectural safeguard and that the real debate is what explicit mechanism replaces it. That debate is about the chamber. About who gets to vote on what, with what weighting, under what constraints.
This essay is about the office. Not the staffed rooms in Delhi and the constituency, but the operational system that connects constituents to the state through the MP.
Even if the chamber debate is settled well, even if the replacement patch for the 1976 freeze is designed honestly, we will still have an institutional problem that the delimitation conversation does not touch. The 543 MPs in the current chamber are working inside an office that nobody has ever fully described, whose outputs are largely unmeasured, and whose core processes would be recognisable to a clerk in 1952. Adding more seats to that chamber without redesigning that office is the wrong intervention on the wrong layer. A constituency with no casework system, no MPLADS outcome tracking, and an MP who asks zero questions across five years does not become better represented by cloning that office three hundred times. It becomes three hundred more constituencies with the same gap.
The arithmetic on the chamber side is equally blunt. The seventeenth Lok Sabha sat for 274 days across five years, roughly six hours per sitting. Spread that floor time evenly across 543 MPs, and each MP’s share is about forty seconds per sitting. Expand the chamber to 850 seats and, under the same proportional assumption, the per-MP share drops to around twenty-five seconds. That is the ceiling on what seat expansion does for legislative voice. The improvement in the other five functions, absent an office redesign, is zero.
What an MP actually does
An MP does at least six different things.
The first is legislative voice. Debates, votes, questions to ministers, private members’ bills, committee work. This is the only part of the job that the parliamentary calendar bounds. It is also the part we measure most often, because the Lok Sabha Secretariat publishes it.
The second is constituency representation in Delhi. Not casework. The work of carrying a region-specific issue, a problem grounded in a particular place, into the national legislative frame. This is harder to measure, but it is visible when it happens.
The third is casework. An individual constituent with a specific problem: a pension that has not arrived, a passport stuck somewhere, a hospital bed that needs to be found, a school admission that needs pushing. By most MP accounts, this is the largest single draw on their attention. It is also the one almost nobody outside an MP’s office tracks.
The fourth is MPLADS disbursement. Each MP has five crore rupees (about six hundred thousand US dollars) a year to sanction for local development work in their constituency2. Roads, wells, community halls, school facilities. The quality of this function determines whether a constituency sees tangible change within a single term.
The fifth is political brokerage. Connecting constituents to the bureaucracy, to the state government, to central ministries. Not casework in the individual sense. Systemic connection work.
The sixth is visible presence. Weddings, funerals, festivals, openings. There is a temptation to dismiss this as theatre. In practice it is how an MP maintains the legitimacy to do the other five.
Adding seats to the chamber does different things to these six functions, and the difference is worth pausing on. The parliamentary calendar is a fixed resource. Sitting days, committee hours, debate time, and question slots do not expand when the chamber expands. Add more MPs and the same floor time gets divided into smaller shares. Functions one and two, legislative voice and constituency representation in Delhi, get diluted per MP as the chamber grows. Functions three through six are bounded per constituency rather than per chamber. Adding MPs does not improve them. It replicates them unchanged. The delimitation debate treats all six as if they scale cleanly with seats. The first two get smaller per MP. The last four get copied across more offices.
The numbers we have are already damning
Before we talk about what we do not measure, we should be honest about what we do.
The seventeenth Lok Sabha, which sat from June 2019 to February 2024, recorded 274 sittings across five years. The first Lok Sabha, between 1952 and 1957, recorded 6773. On a per-year basis, MPs now meet less than half as often as the first parliament under the Constitution. This is not a COVID anomaly. It is the continuation of a thirty-year decline that the Lok Sabha Secretariat’s own statistical handbook documents.
Thirty-five percent of bills in the seventeenth Lok Sabha were passed with less than an hour of debate. Fifty-eight percent were passed within two weeks of introduction4. The Jammu and Kashmir Reorganisation Bill in 2019 and the Women’s Reservation Bill in 2023 were both passed within two days of being tabled.
The referral rate to parliamentary committees has collapsed. In the fourteenth Lok Sabha it stood at sixty percent. In the fifteenth it was seventy-one percent. In the sixteenth it dropped to twenty-five percent. In the seventeenth it fell to sixteen percent4. Committees are the part of the legislative process where expertise is applied, where drafting is scrutinised, where line-by-line review happens. We have quietly stopped using them.
Across five years of the seventeenth Lok Sabha, 4,663 starred questions were listed. Of those, 1,116 were answered orally5. Roughly a quarter. The rest were converted to written answers or held over. Twenty-four sitting MPs asked zero questions across the entire five-year term, a figure verifiable against the Lok Sabha questions database one member at a time.
These are not inferences. They are counts from the Lok Sabha Secretariat’s own publications, the Ministry of Parliamentary Affairs statistical handbook, and the Speaker’s valedictory address at the close of the seventeenth Lok Sabha.
The numbers we do not have are worse
The figures above concern one function, legislative voice. The other five functions, which together take up most of an MP’s actual working time, are almost entirely unmeasured.
There is no published dataset on MP casework. Volume, resolution rates, response time, pattern of complaints, none of it. The Lok Sabha Secretariat does not collect it. The Ministry of Parliamentary Affairs does not collect it. No compiled study from any credible institution surfaces it. Five hundred and forty-three offices handle an unknown number of citizen grievances with no aggregate visibility at all.
MPLADS is better only in principle. Each MP receives five crore rupees a year, at that rate since financial year 2011-12. The Ministry of Statistics and Programme Implementation maintains a dashboard. But the Ministry’s own dashboard carries a disclosure that should be read slowly: MPLADS digital data for the first four years of the seventeenth Lok Sabha, financial years 2019-20 through 2022-23, is not available through the eSAKSHI portal6. Four of the five years of an MP’s principal development fund are not digitally retrievable from the primary government system.
The most recent comprehensive Comptroller and Auditor General review of MPLADS, Report Number 31 of 2010-11, found that district authorities spent between 37 and 52 percent of the funds available to them in each year from 2004-05 to 2008-09, leaving 1,788 crore rupees (about 210 million US dollars) unspent at 31 March 2009. It found twelve district authorities in six states reporting the entire advance given to implementing agencies as spent, which inflated the expenditure figures; 82.54 crore rupees left by former Rajya Sabha members in ten states that had not been passed to their successors; and diversion of funds7.
There has been no comparable performance audit in the fifteen years since.
What we have, put together, is a picture in which the parts of the MP’s job we measure are showing a thirty-year productivity decline, and the parts we do not measure are structurally unaudited.
Why more seats makes all of this worse
Set aside, for a moment, the federal questions the previous essay was about. Assume the replacement patch for the 1976 freeze is designed well and the chamber is eventually expanded. What happens to the office.
The office does not change. Each new seat is a new MP working with the same staff allocation, the same paper-based casework, the same five crore rupees (about six hundred thousand US dollars) a year, the same absence of any outcome tracking on what that five crore does, the same absence of any public metric on how many constituents were served in a year. India has gone from 489 seats serving about 738,000 people each in 1952 to 543 seats serving about 2.58 million each today8. A three-and-a-half-fold expansion in constituency load has not been matched by any expansion in office capability. Not in staffing. Not in process. Not in technology. What we have done instead is hold the office constant and hope the MP absorbs the load through personal effort.
Adding three hundred more MPs on the same office template does not improve representation. It replicates an unmeasured office, unaudited since 2011, with pre-industrial process, across three hundred more constituencies.
This is the work that delimitation, as currently debated, leaves entirely untouched.
The office we have not built
The representation problem is solvable, but not by adjusting the seat count. It is solvable by treating the MP’s office the way we already treat a district consumer grievance portal, a GST intake centre, or a court registry: as an operational function that processes structured work under public accountability and publishes its numbers.
Here is a scaffolding at the architectural level. Six modules. None exotic. Each already runs in an adjacent part of Indian public life.
A structured citizen intake layer. Multi-channel. A walk-in interface in the constituency office, the existing Common Service Centre network for rural access, WhatsApp, a web form, a phone line. Intake is classified at the point of entry: grievance, policy input, development request, emergency. Language support across the twenty-two scheduled languages. Aadhaar-linked where the constituent consents.
A ticketing and workflow layer. Every intake becomes a case with a unique identifier. Auto-routed to the correct department, ministry, or state authority based on classification. A service-level agreement by category, visible to the constituent. Status updates at defined checkpoints. This is what a district consumer grievance portal already does, scaled to a constituency.
A processing and correspondence layer. Staff draft responses with structured templates. Where a reply draws on prior correspondence, the system surfaces it. Translation across languages happens inline. The MP or a named staffer signs every outbound communication. No autonomous response.
A policy signal layer. The interesting one. When three hundred constituents in a district report the same malfunction of the public distribution system, that is not three hundred cases. It is one policy problem appearing three hundred times. An aggregation layer on top of the case database surfaces patterns weekly. What the MP sees is a digest: these are the top issues in my constituency this week, here is how they cluster, here is what is trending up. The MP then has material for a parliamentary question, a private member’s bill, or a letter to the relevant ministry that carries actual constituent weight.
An MPLADS outcome layer. Every sanctioned project is tagged with expected outcome, geotagged, photo-verified at completion, tied back to constituent feedback. A public dashboard shows each project, its status, the money released, and the outcome recorded. Not the existing disclosure of disbursement. Disclosure of effect.
A public transparency layer. Per-MP metrics visible to constituents and to national aggregators. Cases received. Cases resolved within the published SLA. MPLADS projects completed. Outcomes achieved. Parliamentary participation. The same information now buried in Lok Sabha bulletins, brought alongside the five functions the bulletins do not cover.
None of this requires a constitutional amendment. It requires a design, a budget, and implementation authority vested somewhere between the Lok Sabha Secretariat and the Ministry of Parliamentary Affairs.
Where AI fits, and where it does not
A design like the one above is where narrow machine learning does most of the useful work. Classification of intake. Routing of cases to the right department. Translation across languages. Pattern recognition across the case database to surface policy signals. Anomaly detection when a specific category of grievance spikes. These are bounded, well-scoped tasks with clear correctness criteria. They are exactly what machine learning is good at, and exactly what large language models are poor at on a cost and reproducibility basis.
Where language models earn their keep is narrower: first-draft generation of correspondence for staff review, summarisation of casework into a weekly digest for the MP, natural-language queries into the aggregated case database. In every one of these uses, a human reviews, edits, and signs the output.
This is the same architectural principle that applies to every other decision system I have worked on in industry.
It is not a system where a model decides which complainant gets escalated or whose emergency is genuine. It is not a chatbot answering grievances on an MP’s behalf, because that is a shadow-governance layer with no audit trail and no accountable signatory. It is a structured, auditable workflow that happens to have machine learning in the right places, and a named human on every material output.
What this would cost, relative to the alternative
The 131st Amendment, had it passed, would have required a delimitation exercise spanning years, redistribution of constituencies across states with the attendant political cost, and ongoing salary, staff, and chamber-side infrastructure expenditure on more than three hundred additional MP offices. The current Lok Sabha chamber, designed for 888 seats, inaugurated in May 2023 and in use since September 2023, already absorbs some of the physical cost. The operational cost has never been published in any form I could find, but the order of magnitude is clearly in the low thousands of crore (low billions of US dollars), annualised.
The office I have described would cost, at an informed estimate, between fifty and one hundred crore rupees (roughly six to twelve million US dollars) to build, and thirty to fifty crore (three-and-a-half to six million US dollars) a year to operate, rolled out across the 543 offices we already have. That figure assumes a central platform build at the higher end and per-constituency operating cost in the range of five to ten lakh rupees a year for staff time, hosting, and support, scaled by the 543 offices. This is a conservative order-of-magnitude estimate, not a line-item budget; the scoping exercise that would produce a line-item budget is itself a month’s work.
This is one order of magnitude less expensive than the delimitation alternative. It is deliverable in three years rather than ten. It does not require a constitutional amendment. And it does something that delimitation does not do at all, which is improve the quality of representation that each MP provides.
Where this argument is hardest to land
The honest pushback on all of the above is that the design is not the hard part. Roughly equivalent systems already run elsewhere in Indian public life. The hard part is that 543 MPs currently operate as independent principals whose incentive to adopt standardised central infrastructure is weak, and in some cases negative. A national casework system would make resolution rates public. It would make MPLADS outcomes comparable across constituencies. It would make zero-question MPs visible in a way that annual valedictory speeches do not. The constituencies that would benefit most from adoption are the ones where the MP has the least interest in being measured. That is a political problem and not a technical one, and I do not have a clean answer to it.
What I will say is that the same political problem exists for delimitation, and in a more acute form. Expanding the chamber requires a constitutional amendment. Building the office does not. It requires a budget line, a procurement, and a Secretariat with the authority to set a standard. If there is going to be a fight over whether MPs should be measured on the work they actually do, that fight is worth having before we triple the number of MPs doing that unmeasured work.
The most realistic path is not a national mandate. It is a central voluntary scheme, tied to performance grants or party-level adoption in two or three states, that produces comparable numbers across a few dozen constituencies. Once those numbers exist publicly, continued non-adoption becomes politically expensive for the constituencies that opt out. The scheme does not have to be mandatory to become the floor.
What this is actually a proposal for
The previous essay ended with the observation that the patch was never named, and that its replacement has to be. This one ends with the observation that the office was never designed, and that its design is overdue.
Neither the framers of the Constitution nor the drafters of the 1976 freeze described what an MP’s working life should look like in operational terms. They described the chamber, the vote, the bill, the committee. They did not describe the five other functions that together consume most of an MP’s time and shape most of what their constituents actually experience. That omission made sense in 1949 and even in 1976. It does not make sense in 2026.
The delimitation debate, for all the attention it has drawn, is operating on a twentieth-century picture of the MP’s office. The seat-count argument makes sense inside that picture. Outside it, in the actual operational reality of a Member of Parliament representing 2.58 million people with paper-based casework and no outcome tracking on roughly six hundred thousand US dollars a year, the argument looks like the wrong instrument on the wrong layer.
Designing the office is not a substitute for the replacement patch. It is the other thing that has to be done. The chamber is where the vote happens. The office is where the representation happens.
If the first essay asked who sits in the chamber and with what weight, this one asks what they actually do between votes.
The patch needs naming. The office needs building.
I design decision systems for regulated enterprises. This is the same problem at parliamentary scale: structured work, named humans, measured outputs.
Sources
5. Lok Sabha Secretariat: Highlights Pertaining to Questions, 15th Session of the 17th Lok Sabha
Share feedback
No public comments. If you want to share a reaction or correction, send it here.