Democracy Is Infrastructure

There is a constituency office in a mid-sized Kenyan town that I have visited four times over the past two years. The MP is conscientious. Her staff are dedicated. The constituency has real needs — bursary requests, CDF project petitions, constituent casework, public participation meetings for county projects. And her office manages all of it on two WhatsApp groups, a shared Google Sheet that nobody updates consistently, and the personal phone numbers of the three staff members who have been there longest.

On my first visit, I sat with the office manager as she worked through a backlog of constituent requests. A woman from one of the wards had come to ask about the status of a bursary application her son had submitted eight months earlier. The office manager opened WhatsApp, scrolled through hundreds of messages looking for the student's name, found a message from the ward administrator confirming receipt, found a later message suggesting the bursary committee had considered it, found nothing further. She called the ward administrator. The administrator called someone else. Forty minutes later, the student's mother left without a definitive answer.

I want to be precise about what I observed. The people in that office were not failing at democracy. They were attempting to do genuine constituency service work on infrastructure that was not designed for it. And the gap between what they were attempting and what their tools could support was producing exactly what you would expect: case backlogs, constituent frustration, lost institutional memory, staff burnout, and an MP who had no way to measure whether her office was actually effective.

The content doesn't shrink. The container it's poured into does.

1. A tooling argument, not a verdict on governance

This paper argues that a specific civic institution — the Kenyan constituency office — operates on infrastructure unequal to its mandate, and that this is a tooling problem with a tooling-shaped portion of a solution. Here is precisely what kind of claim that is.

What it argues from. Repeated, first-person observation of one constituency office over four visits across two years — illustrative, not sampled. Kenya's constitutional structure: 290 constituencies under Article 89(1) of the Constitution of Kenya, 2010.¹ The Auditor-General's own published, nationwide audit of the National Government Constituencies Development Fund (NG-CDF) for the 2023/2024 financial year, covering all 290 constituency funds — real, systemic, and independently produced.² A political-science framework, from US think-tank literature rather than Kenyan or African scholarship, for treating democratic institutions as infrastructure rather than only as values.³ One peer-reviewed public-sector study, from South Africa rather than Kenya, on why losing staff means losing institutional knowledge absent deliberate retention or documentation practice.⁴

What it does not argue from. No systematic survey of how Kenya's 290 constituency offices actually manage casework exists in this paper — the WhatsApp-and-spreadsheet pattern is this paper's own repeated observation of one office, generalised by argument, not by sampling. No study isolates how much of the NG-CDF Auditor-General findings in §4 trace specifically to case-tracking tooling, as distinct from procurement practice, staffing capacity, or compliance culture — the Auditor-General's own report is explicit that many of its gravest findings are governance and legal-compliance failures, not tooling failures, and this paper does not conflate the two. No independent audit of BungeConnect's own claimed pilot results exists; those remain Flux's internal, unaudited data.

What it does not claim. It does not claim software alone would resolve the irregularities the Auditor-General documents — procurement without competitive bidding, Project Management Committees bypassed, late Board disbursement are compliance and governance failures a case-management system does not, by itself, fix. It does not claim every one of Kenya's 290 constituency offices operates identically to the one described in the opening. It does not claim the "democratic infrastructure" framework in §2 was developed for, or tested against, the Kenyan context; §7 states plainly that it is borrowed political-science vocabulary, not an empirical finding about Kenya.

What would refute it. Given in full in §9. The sharpest version: if constituency offices already running structured case-management software are found to show audit and compliance outcomes no better than offices relying on WhatsApp and spreadsheets, the causal link this paper proposes — that the tooling gap meaningfully contributes to the outcomes documented in §4 — would be substantially undermined, even though the underlying governance failures the Auditor-General documents would remain real regardless.

2. The Infrastructure Frame

We talk about democracy primarily as values — free and fair elections, civil liberties, the rule of law, accountability. These are real and important. But democracy is also, and just as fundamentally, infrastructure. It is the set of systems that allow a citizen to communicate a need to a representative, allow that representative to act on it, and allow both parties to know what happened and why.

This is not only this paper's own framing. A 2019 analytical framework from New America's Political Reform program, developed with Demos, argues for treating "democratic infrastructure" as the underlying institutional, political, and social supports — not only formal rights — that let a polity function, building on the sociologist Charles Tilly's account of democracy as requiring "mutually binding consultation" between the state and the public, where state agents have clear, enforceable obligations to deliver benefits responsive to constituents' demands.³ That framework was built from US political institutions and a US policy-reform context; it was not developed for, or tested against, a Kenyan constituency office, and I am borrowing its vocabulary rather than its evidence. But the conceptual move — treating accountability and responsiveness as something that runs on concrete institutional and technical supports, not only on the values a constitution declares — transfers cleanly to the case this paper is about.

When the infrastructure fails, the values don't disappear. People still vote. MPs still win constituencies. Offices still open. Staff still answer phones. But the gap between the democratic promise — that a citizen's voice will be heard and acted upon — and the democratic reality widens. And the widening happens not from malice but from missing operational layer.

The infrastructure frame is useful because it shifts the question from "who is to blame?" to "what is broken and what does the fix look like?" It moves the conversation from accountability rhetoric to engineering.

Kenya's 290 constituencies each have a constituency office.¹ Each office manages a National Government Constituency Development Fund (NG-CDF) allocation. Nationally, in the 2023/2024 financial year — the most recent year for which the Auditor-General has published a complete summary report — a total of Kshs 57,452,681,385 was actually disbursed from the NG-CDF Board to the 290 constituency funds, against an approved national budget of Kshs 74,752,868,447: a 23% shortfall against the approved figure.² Divided evenly across 290 constituencies, that works out to roughly Kshs 198 million actually disbursed per constituency on average for the year — a national average, not a per-constituency figure; press reporting on the following financial year's allocations shows individual constituencies receiving anywhere from roughly Kshs 70 million to over Kshs 200 million, so no single office's experience should be read off the average.⁵ Each office handles constituent casework: bursaries, referrals, petitions, complaints. Each office is supposed to run public participation processes for projects, maintain a constituent database, and report to the MP and the national CDF Board.

Almost none of them have the software infrastructure to do this at the scale their mandate requires.

3. What a Constituency Office Actually Does

A constituency office is, functionally, a combination of a casework CRM, a project management system, a grants administration platform, a public communications channel, and a small government treasury — all operating simultaneously, often with three to five staff.

The casework load alone is staggering. An urban constituency office might receive a hundred new requests in a week — bursaries, medical assistance, emergency relief, referrals to county services, project petitions. Each case has a constituent, a request, a status, a handler, and an outcome. To manage a hundred cases a week with any reliability, you need a case management system. What most offices have is a notebook and a WhatsApp group.

The CDF administration is particularly consequential. An MP's office manages hundreds of projects across dozens of wards — boreholes, classrooms, dispensary equipment, road repairs, community halls. Each project has a procurement process, a contractor, progress milestones, payment disbursements, and a completion report. The CDF Board requires that offices maintain records of all of this. In practice, the records are whatever the person who attended the site visit wrote in their notebook.

When an audit comes — and audits do come, nationally, every year — the reconstruction of project records from WhatsApp histories and paper files is, on the evidence in §4, a genuinely weeks-long exercise for a large share of Kenya's 290 constituency funds, not an occasional hardship. When a contractor disputes payment, the resolution depends on whether someone remembers the right conversation. When a ward demands to know why their project hasn't started, the office makes phone calls until it can produce an answer that might or might not be accurate.

The hardest part of my job is not the work itself. It is that I have no way to know, at any given moment, what the status of any given thing is. Everything is in someone's head or someone's phone. When that person is sick or leaves, we start over.

— Constituency office manager, Central Kenya

4. What the Auditor-General's own numbers show

The previous section's "weeks-long exercise" claim and the office manager's "we start over" quote are not, on their own, more than one person's testimony and this paper's own inference. The Auditor-General's 2023/2024 summary report on the NG-CDF — covering all 290 constituency funds, nationally, for one financial year — is independent of Flux and of this paper, and it is worth stating plainly what it actually found, rather than only gesturing at "audits do come."

Two hundred and fifty-eight of 290 NG-CDFs (89%) carried unresolved prior-year audit issues into the 2023/2024 review, and the report states that "Management had not resolved the issues or given any explanation for the delay in resolving the issues."² Ninety-nine NG-CDFs (34%) had expenditure totalling Kshs 1,266,628,897 that was "not supported by the relevant documentation and supporting schedules."² Thirty-nine had inaccuracies in their financial statements totalling Kshs 610,198,588; sixty-four had unreconciled variances totalling Kshs 1,328,327,065.² Every one of the 290 — all of them — was flagged for systemic issues in both project management and financial reporting, the two largest categories in the report's own systemic-issues summary.² One hundred and thirty-six were flagged specifically for bursary-management issues, the category closest to the opening anecdote of this paper.² A disclaimer of audit opinion was issued for one fund and adverse opinions for three others.²

This evidence establishes something specific and limited. The Auditor-General's report audits the NG-CDF's financial and project records — budgets, expenditure, procurement, bank reconciliations — not the kind of constituent-casework record (the status of one student's bursary application) the opening anecdote is about. Those are related but distinct record-keeping problems, sitting inside the same office, often handled by the same overstretched staff, but not the same dataset. What the report does establish, independent of this paper's own observation, is that systemic record-keeping and compliance failure in Kenyan constituency-level institutions is not an edge case affecting one office I happened to visit four times. It is close to universal across the 290. And the report is equally clear that a meaningful share of its findings — non-compliance with procurement law in 53 funds,² bypassing Project Management Committees, late disbursement by the Board itself, irregular fund reallocation without Board approval — are compliance and governance failures that a better case-management tool does not, by itself, resolve.² §7 returns to this distinction directly.

5. Why This Matters Beyond Efficiency

One response to everything I've described is: "So what? Constituencies still function. MPs still get re-elected. The system still works, more or less." And there is something to this. Democracy does not require software to exist.

But consider what the absence of institutional infrastructure actually produces over time.

First, it concentrates power in staff with institutional memory. When the case management system is one person's phone, that person has power — they know things that aren't written anywhere. When they leave, that power leaves too. This is not a Kenya-specific or constituency-office-specific pattern; a 2024 peer-reviewed study of South African public enterprises found that tacit organisational knowledge loss is "largely caused by voluntary and involuntary employee turnover and a lack of retention strategies" — a different country and a different kind of public institution, but the same underlying mechanism this paper is describing: knowledge that lives only in a person, not a system, leaves when the person does.⁴ Institutions that operate this way become dependent on individuals in ways that make them fragile, and that create informal hierarchies that are resistant to accountability.

Second, it makes it impossible to measure outcomes. An MP who cannot tell you how many bursaries her office distributed in the last three years, what they were worth, and what happened to the recipients, cannot claim with any evidence that her constituency service work is effective. She may be doing genuinely good work. But without records, it is undocumented work — and undocumented work cannot be replicated, improved, or held accountable.

Third, and most importantly, it degrades the constituent experience in ways that erode trust in democratic institutions broadly. A constituent who cannot get a status update on a bursary application after eight months is not thinking "the infrastructure is underfunded." She is thinking "these people don't care about me." That erosion of trust is real and cumulative, and it happens across every constituency, ward, and county office that operates without adequate infrastructure.

6. What BungeConnect Is Doing About This

BungeConnect is a constituency operations platform — built specifically for the Kenyan parliamentary context, shaped by conversations with constituency staff across multiple election cycles. It is not enterprise software that has been adapted for government. It is software designed from scratch for what a Kenyan constituency office actually does.

Case management with constituent profiles, status tracking, and outcome recording. Bursary workflow — from application through committee review to disbursement, with audit trail. CDF project tracking with milestone management, contractor records, and payment history. Public participation tooling — for recording inputs, publishing summaries, and linking community feedback to specific decisions. Constituent database with ward-level organisation. Reporting dashboards for MPs and for CDF Board compliance.

It is intended for the 2027 election cycle, and Flux's own account is that it is already being piloted — a claim this paper states as Flux's own, not as independently verified. The premise behind it is simple: that democracy works better when its infrastructure works.

The values are already there. The mandate is already there. The people are already there. What has been missing is the operational layer — and that is precisely what can be built. What software cannot do is resolve the governance and compliance failures §4 documents on its own; a case-management system that makes bursary status trackable does not, by itself, stop a Project Management Committee from being bypassed or a procurement process from skipping competitive bidding. Those require enforcement and political will a product does not supply.

Infrastructure is never glamorous. Nobody campaigns on "better case management software for constituency offices." But the woman who waited forty minutes for an answer she never got — she felt the absence of that infrastructure in her body, in the time she spent waiting, in the defeat she carried home.

Democracy is what runs on the infrastructure. When the infrastructure is a WhatsApp group, the democracy is whatever fits in one.

7. Objections I take seriously

"Flux has a commercial interest in this conclusion." Flux builds BungeConnect, and this paper's central argument — that constituency offices need purpose-built case-management software — is close to BungeConnect's own sales pitch. The opening observation is Flux's own fieldwork, and BungeConnect's claimed pilot engagement is Flux's own internal, unaudited data. I am not going to pretend otherwise. The honest mitigation is narrow: the NG-CDF audit evidence in §4 is independent of Flux, produced by Kenya's own Auditor-General, and it would show the same systemic pattern whether or not Flux or BungeConnect existed. It establishes that the problem is real and national in scale. It does not establish that BungeConnect, specifically, is the right or only fix, and this paper does not have independent evidence that it is.

"The Auditor-General's findings are governance and compliance failures, not tooling failures — better software would not have stopped them." This is largely correct, and §4 and §6 say so directly rather than only here. Procurement without competitive bidding, Project Management Committees bypassed, funds reallocated without Board approval, late disbursement by the Board itself — these are failures of process adherence and institutional discipline that a case-management tool does not enforce on its own. What this paper claims is narrower: that the specific sub-problems of unsupported expenditure, unreconciled variances, and unresolved prior-year findings — the kind of gap that opens when nobody has a reliable, searchable record of what happened — are plausibly, though not provenly, worsened by the absence of structured record-keeping tools. Whether better tooling would measurably reduce those specific findings, as distinct from the governance failures, is exactly the kind of question §9's falsification conditions are built to test, and this paper does not have the evidence to answer it yet.

"One constituency office, visited four times, is an anecdote, not a sample." Correct, and §1 says so plainly. I do not know what share of Kenya's 290 constituency offices manage casework the way the office in this paper's opening does, versus offices that may already have more developed internal systems. The Auditor-General's report, in §4, establishes that the financial and project-management record-keeping problem is close to universal across all 290 — but it does not establish that casework and bursary-status tracking specifically follows the same WhatsApp-and-spreadsheet pattern everywhere, only that it does in the office this paper observed.

"The 'democratic infrastructure' framework is American political theory, not Kenyan or African scholarship." Also correct. Rahman and Gilman's framework was developed from US democratic institutions for a US policy-reform audience.³ This paper uses it for conceptual vocabulary — treating accountability as something infrastructure either supports or forecloses — not as evidence about Kenya. The claim that this vocabulary usefully describes Kenya's constituency-office problem is this paper's own extension, not something the framework's authors argued.

8. Tooling, governance, and where the line sits

Software alone would not resolve the NG-CDF's documented governance and compliance failures. Procurement irregularities, bypassed Project Management Committees, and late Board disbursement require enforcement and political will, not only better tools.

It does not claim a representative sample of Kenya's 290 constituency offices; the WhatsApp-and-spreadsheet pattern is this paper's own repeated observation of one office, not a survey finding.

It does not claim the Auditor-General's financial and project-management audit findings in §4 are the same record-keeping problem as the constituent-casework tracking described in the opening anecdote; they are related but distinct, sitting in the same offices but covering different records.

It does not claim BungeConnect's pilot results are independently verified; they remain Flux's internal, unaudited data, and this paper's underlying evidence for the scale of the problem does not depend on them.

It does not claim the "democratic infrastructure" framework in §2 was developed for or tested against Kenya; it is borrowed conceptual vocabulary from US political-reform literature, not an empirical finding about this context.

And it does not claim the tacit-knowledge-loss mechanism in §5 has been measured specifically in Kenyan constituency offices; the supporting study is South African and public-sector generally, cited for the underlying mechanism rather than as direct Kenyan evidence.

9. What would falsify this

Stated as commitments, in decreasing order of how much damage each would do to this paper's claims.

Constituency offices already running structured case-management software are found to show audit and compliance outcomes — unresolved prior-year findings, unsupported expenditure, unreconciled variances — no better than offices relying on WhatsApp and spreadsheets. That would directly undermine the causal link this paper proposes between tooling and the specific record-keeping failures documented in §4, even though the underlying governance failures the Auditor-General reports would remain real either way.

A systematic survey of Kenya's 290 constituency offices finds that informal, WhatsApp-and-spreadsheet-based casework management is uncommon rather than the norm. That would undermine this paper's generalisation from one repeatedly observed office to a national pattern, and would require the claim to be narrowed to the specific office described rather than offered as representative.

Independent examination of BungeConnect's own pilot data finds it does not show the engagement or outcome improvements Flux has claimed internally. That would not undermine the evidence for the scale of the problem (§4 is independent of BungeConnect), but it would remove this paper's implicit suggestion that BungeConnect specifically is a working answer to it.

The specific NG-CDF figures cited in §2 and §4 are revised, restated, or found to contain reporting errors by the Auditor-General in a subsequent report. The narrowest and least damaging: it would require correcting specific numbers, not withdrawing the argument.

Appendix A — Evidence table

Appendix B — Citation register

CITATION-NEEDED — share of constituency offices using informal casework tracking. Appendix A row 14. The WhatsApp-and-spreadsheet pattern is this paper's own repeated observation of one office; no survey of all 290 was found.

CITATION-NEEDED — causal contribution of tooling to the documented audit findings. Appendix A row 15. The Auditor-General's report establishes the findings; it does not isolate how much, if any, traces to case-tracking tooling specifically as opposed to procurement practice, staffing, or compliance culture — the central uncertainty §9's sharpest falsification condition is built to test.

CITATION-NEEDED — independent verification of BungeConnect's results. Appendix A row 16. Pilot engagement claims are Flux's own internal, unaudited data.