AiSewak
Government Helpline Playbooks

CM Helpline Performance Dashboard: KPIs and Voice AI

What a CM helpline performance dashboard should show: answer rate, SLA breaches and citizen-verified resolution, not just disposal. How voice AI feeds it.

22 min readUpdated 10 Nov 20264,319 words

A CM helpline performance dashboard should track four things: the share of calls answered, grievances breaching their deadline, the share of closed cases the citizen confirms were fixed, and time to resolution. Disposal alone counts closed files. Voice AI supplies the first three by answering every call and phoning citizens back after closure.

Executive Summary

The Chief Minister's helpline is one of the most visible pieces of citizen-service infrastructure a state runs. It carries the CM's name, it promises one number for every department, and its performance is read as a proxy for the government's competence. The question for a state is what its dashboard actually measures.

The public record shows how different the available numbers are. Madhya Pradesh's CM Helpline portal displays a running counter of 4,07,64,614 complaints registered and 4,01,99,211 disposed, which works out to 98.6% disposed (cmhelpline.mp.gov.in, read on 10 November 2026; the percentage is our arithmetic). That counts closed files. It does not say how many citizens were satisfied. The central government's CPGRAMS asks citizens directly, through a feedback call centre: in November 2024 it found 44% of 55,206 citizens contacted were satisfied with the resolution (DARPG monthly report, November 2024), and DARPG reports 76% for January to June 2026 (PIB backgrounder, 9 August 2026).

Executive Callout — Closure is not cure. A disposal rate and a satisfaction rate answer different questions, and a CM helpline dashboard needs both. Voice AI helps because it can instrument the full lifecycle: intake, classification, routing, deadline-tracked escalation and a call back to ask the citizen whether the problem is fixed. National proof of the intake half exists: DARPG launched the CPGRAMS voice chatbot "Samadhan Didi" on 30 May 2026, which lets citizens register grievances by speaking and identifies the ministry, department and category automatically (PIB backgrounder, 9 August 2026).

This article is the CM-helpline overview: what published data says, how a CM helpline works and where it breaks, the dashboard metrics worth tracking, how a state grievance line with verified closure changes the workflow, and a 30-day pilot design. AiSewak has not run a CM-helpline deployment, so every outcome described below as a result belongs to the cited source, and everything else is design. For the UP deep dive, see AI for the UP CM Helpline 1076; for the broader grievance workflow, see AI for Public Grievance Redressal.

Introduction: One Number, Every Department

The premise of a CM helpline is simple. A citizen does not need to know which department owns the problem or which form to fill: they call one number and the state routes the complaint. When it works, it is an instrument of responsive governance. When it does not, it becomes a call that rings out and a complaint logged into a queue that goes nowhere.

Large states run one: 1076 in Uttar Pradesh, and 181 in Rajasthan and Madhya Pradesh. This article makes one argument: a CM helpline is a high-value place to deploy Voice AI, because it is cross-cutting (it touches many departments), measurable (every call and complaint is logged), politically owned (it carries the CM's name) and exposed to the gap between closing a file and solving a problem.

What the Published Data Says

The table lists what AiSewak could verify from the primary page or a named news report. Gaps are stated, not filled.

HelplineWhat is publishedSource and scope
Madhya Pradesh CM Helpline, 1814,07,64,614 complaints registered; 4,01,99,211 disposed (98.6%, our arithmetic)State portal counter, read 10 November 2026. A disposal count, not a satisfaction measure
Rajasthan Sampark, 181Chief Secretary: satisfaction above 80%, pending cases down to about one lakh, 25,000 cases disposed a day against an expected 5,000 to 6,000, and AI bots used to collect satisfaction feedbackStatements by the Chief Secretary in a First India feature; not independently audited
Rajasthan Sampark, police complaints1,60,292 registered in the past year, 1,52,444 resolved (about 95%)IANS report, 28 April 2026, official data quoted at an inspection
Uttar Pradesh CM Helpline, 1076Launched 5 July 2019 as a toll-free, 24x7 line with a 500-seat centre designed for 80,000 inbound and 55,000 outbound calls a dayKRC Times, 5 July 2019; launch-day design capacity, not current volume. AiSewak found no current outcome data
CPGRAMS (central)Satisfaction 44% (November 2024) and 76% (January to June 2026) from feedback callsDARPG and PIB

Two patterns follow. First, the headline number is usually disposal, which counts closure; satisfaction appears only where someone calls the citizen back, as CPGRAMS does and as Rajasthan's Chief Secretary describes. Second, the figures are not comparable across states: each uses its own definition of resolved, its own period and its own source. A dashboard that compares them needs common definitions first.

How a CM Helpline Works, and Where It Can Break

A complaint travels through six stages. The failure points below are the places a performance dashboard should be able to see; they describe the typical design, not a finding about any one state.

Stage 1: Intake. The citizen dials. If lines saturate, calls drop before anyone answers; if intake handles only some languages, some citizens are excluded. The measures: answer rate, abandonment, languages served.

Stage 2: Classification. An operator maps the complaint to a department. A wrong mapping sends it into the wrong queue, where it ages. The measure: misrouted-and-transferred rate.

Stage 3: Cross-department routing. The complaint reaches the department's officer, often without live confirmation that it was received or acted on. The measure: time to first departmental action.

Stage 4: Deadline tracking. Each grievance carries a resolution deadline. If breaches are only visible in a monthly report, the deadline is advisory. The measure: SLA-breach rate and age of open cases.

Stage 5: Resolution. The department marks the file closed and the disposal counter rises. This is the step the disposal rate counts.

Stage 6: Citizen call-back. The step that verifies the fix is the one a manual process samples at best. DARPG's feedback call centre and Rajasthan's AI feedback bots are two Indian examples of doing it at volume. The measure: citizen-verified resolution rate and reopen rate.

A helpline measured only on stage 5 will produce stage 5. The structural fix is to measure stages 1, 4 and 6 as well.

Five Weak Points Worth Testing For

None of these is fixed by adding seats alone.

  1. Closure and cure look the same on a disposal dashboard. Only a citizen call-back separates them.
  2. Human capacity limits quality as well as volume. A fixed headcount can answer every call or follow every grievance to cure, and the contract usually measures the first.
  3. Language coverage is a design choice. A helpline that serves a few languages filters out some of the citizens it was built for. Test per-language acceptance before launch; see per-language acceptance testing.
  4. Demand is uneven. Complaint volume varies by season and event, and a centre sized for the average either idles or drowns. Ask for hourly and weekly volume data before sizing anything.
  5. Leadership visibility depends on reporting cadence. If performance reaches the CM's office as a monthly document, the office steers by a lagging signal.

How Voice AI Changes the Lifecycle

Voice AI does more than answer extra calls. It makes each stage measurable, escalatable and verifiable. The mapping below is a design, not a measured result.

Intake: every call answered, in the caller's language. A voice agent can answer around the clock. The government language stack it would run on is Bhashini, which the Ministry of Electronics and IT describes as covering 36 text and 23 voice languages, powering more than 500 government websites and processing over 15 million inferences a day (PIB, 12 March 2026). Listing a language is not the same as passing it on phone audio, so each language needs its own acceptance test.

Classification: automatic categorisation at intake. A model trained on the state's own department taxonomy can propose the category while the citizen is still speaking. Samadhan Didi does this at national level for CPGRAMS, identifying ministry, department, category and sub-category from speech, with follow-up questions where needed (PIB, CPGRAMS backgrounder). PIB reports the launch; AiSewak has found no published outcome data for it yet.

Routing: a handoff into the grievance portal. Where the state's portal exposes a write interface, the agent can create the record with department attached. Where it does not, a read-only design with a human entering the case is safer; see read-only versus write integration.

Deadline tracking: automatic escalation. A grievance that passes its deadline can move up the state's own escalation ladder, with an SMS to the citizen at each step. The ladder (for example officer, district head, divisional head, CM office) must be the state's actual rules, not a template.

Resolution: routine queries settled on the call. Status checks and procedural questions are candidates for end-to-end handling. The share a given helpline can contain is unknown until its own call data is classified; AiSewak does not quote a containment figure. Human agents take everything sensitive or ambiguous; see Human-in-the-Loop: Augmenting Government Call-Centre Agents.

Call-back: asking the citizen whether it is fixed. After a file is closed, an automated call in the citizen's language asks whether the problem is solved. A "no" reopens the grievance and restarts the clock. This turns a disposal metric into a satisfaction metric on every case rather than a sample.

The CM-Helpline AI Ladder (AiSewak's own framework)

This is a planning model, not a measured benchmark of any state.

LevelStageWhat it looks like
0ManualHuman IVR and operators; disposal-only metrics
1Assisted intakeWhatsApp or chatbot channel; text only; humans still route
2Voice intakeRound-the-clock multilingual voice answering; the agent logs the grievance
3Routed and trackedThe agent classifies, routes and escalates on deadline breach
4Verified resolutionAutomated citizen call-back; satisfaction becomes the headline KPI
5Command-centre nativeCalls, deadlines and sentiment stream live to the CM's office dashboard

The 30-day pilot below is designed to take one district cluster to Level 3, with Levels 4 and 5 following only if the pilot's own data supports them.

Indian Precedents

  • Samadhan Didi (national, CPGRAMS). Launched on 30 May 2026 by Dr Jitendra Singh, Minister of State for Personnel and Public Grievances, built by DARPG with Bhashini. The 30 May release says the CPGRAMS platform already supports the 22 languages of the Eighth Schedule, with regional and indigenous languages being added in phases, and that the Minister urged States to adopt similar voice tools in their own grievance portals (PIB, 30 May 2026). The later PIB backgrounder describes the chatbot as using Bhashini's speech and translation services (PIB, 9 August 2026).
  • Rajasthan Sampark 181. The Chief Secretary describes a "Next-Gen Jan Sampark" reform in which AI bots collect satisfaction feedback and unsatisfied citizens' grievances are raised a level (First India). These are the state's own statements; no independent evaluation was found.
  • Uttar Pradesh CM Helpline 1076. A 24x7 line launched in 2019 that sends the citizen an SMS on registration (KRC Times). The turnaround case is examined in AI for the UP CM Helpline 1076.
  • Madhya Pradesh CM Helpline 181. Publishes a live disposal counter, which is the metric this article argues is necessary but not enough (portal).

The Performance Dashboard: What to Track

The dashboard is where a CM helpline's data becomes steering. These metrics are AiSewak's proposed definitions; a state should fix its own before a pilot, and each needs a named data source.

MetricDefinitionWhat it exposesWhere the data comes from
Answer rateCalls answered within a set time as a share of calls receivedCapacity failure at intakeTelephony logs
AbandonmentCalls dropped before an answerPeak-load failureTelephony logs
Misroute rateGrievances transferred to another department after first routingClassification qualityPortal audit trail
SLA-breach rateOpen grievances past their deadline, by department and ageWhere the queue is stuckPortal records
Disposal rateGrievances marked closed as a share of registeredClosure (not cure)Portal records
Citizen-verified resolutionShare of closed cases where the citizen confirms the problem is fixedWhether closure was realCall-back results
Reopen rateClosed cases reopened after a "not fixed" answerQuality of closureCall-back results
Time to resolutionMedian and 90th-percentile days from registration to verified fixSpeed that the citizen feelsPortal and call-back records
Language mixShare of calls by language, with per-language containment and errorWhether some groups are poorly servedCall transcripts

The pairing to watch is disposal rate against citizen-verified resolution. If the first is high and the second is low, the dashboard has found closure without cure. This gap is the reason CPGRAMS built a feedback call centre.

Every AI-handled call produces structured data: department, category, location, deadline status, outcome, citizen-verified result and transcript sentiment. Streamed to the CM's office, that data lets it see which departments are breaching deadlines this week, which districts are surging and where disposal and satisfaction diverge. The helpline becomes the sensing layer and the command centre the steering layer. The architecture of that steering layer is treated in AI for the Chief Minister's Command Centre.

Before and After: The CM-Helpline Operating Model

DimensionTraditional CM helplineVoice-AI-modernised CM helpline (design)
AvailabilityLimited by shifts and seat countRound the clock; capacity scales with demand
LanguagesWhatever the staff speakBhashini voice languages, each acceptance-tested
ClassificationManualModel proposes the category at intake
RoutingManual forwarding, no confirmationPortal handoff where an interface exists
EscalationDeadline visible in a periodic reportAutomatic escalation on breach, SMS to the citizen
Resolution checkSampled survey, if anyCall-back on every closed case
Headline metricDisposal rateCitizen-verified resolution next to disposal
CMO visibilityPeriodic reportLive dashboard

On cost, AiSewak does not publish per-call or pilot prices in this article. Any state should ask each vendor for a priced, capped proposal against its own call volumes, and compare it with the fully loaded cost of its current seats.

The 30-Day Pilot

Start with a scoped, measurable pilot, not a statewide replacement.

Scope. Choose two or three districts that include an urban baseline and at least one district where non-Hindi or regional-language speech is common. Cover three call types end to end: status checks, new grievance registration and escalation alerts.

Integration. Connect to the state grievance portal (write access only if the state grants it, otherwise read-only with human entry), a Bhashini-based speech layer, an SMS or WhatsApp gateway for confirmations, and a read-only feed to the CMO dashboard. Handle personal data as described in DPDP Act, Data Privacy and Security for Government Voice AI.

Baseline first. Before launch, record the helpline's current answer rate, misroute rate, SLA-breach rate and disposal rate from its own logs. A pilot with no baseline cannot show improvement.

KPIs. Measure answer rate, containment of routine queries, escalation accuracy and, above all, citizen-verified resolution on call-back. Set the targets against your own baseline; AiSewak does not offer a benchmark because it has no CM-helpline deployment data.

Pathway. The sequence from a 30-day pilot to statewide scale is covered in The 30-Day Pilot to Statewide Scale Roadmap.

Expected Impact

If the pilot works, four things change. Access widens, because a citizen who cannot use a web form can register a grievance by voice in their own language. The gap between disposal and satisfaction becomes visible, because the call-back reopens files that were closed but not fixed. Departments see their own deadline breaches sooner. And the CM's office gains a live view in place of a periodic report. The size of each effect is an empirical question for the pilot, and this article does not forecast it.

Risks and Mitigation

Vendor concentration. Treat the AI layer as an overlay with clear exit and data-portability terms, not a second single point of failure.

Empathy and edge cases. Grievances can involve distress, corruption allegations or safety. Build human handoff in from the start: the AI takes routine traffic, and anything ambiguous or sensitive goes straight to a person with the context attached.

Data privacy. CM-helpline calls hold personal data and citizen identifiers. Plan consent capture, access controls and data residency under the DPDP Act, which is only partly in force: its consent and penalty provisions commence on 13 May 2027.

Metric gaming. Any new KPI can be gamed. Make citizen-verified resolution the headline metric, because it is the hardest to fake, and audit a sample of call-backs by listening to them.

Workforce concerns. Frame and staff the deployment as augmentation, with the AI taking repetitive traffic so human agents work on cases that need judgment.

Future Outlook

Bhashini now runs as operating infrastructure, and Samadhan Didi shows voice-first grievance intake at national level, with the Minister publicly urging States to follow. What is still missing is published outcome data for AI in grievance lines, which is the reason the first step is a pilot with its own baseline and its own call-back measurement.

Key Takeaways

  • A CM helpline dashboard needs disposal rate and citizen-verified resolution side by side; closure alone does not show whether the problem was fixed.
  • The published numbers differ in definition, period and source, so compare states only after fixing common definitions.
  • CPGRAMS found 44% satisfaction in November 2024 and 76% for January to June 2026 by calling citizens back, which shows the measurement exists and works.
  • Voice AI can instrument all six lifecycle stages; the automated call-back is the keystone.
  • Start with a 30-day, multi-district pilot with a recorded baseline, and measure verified resolution rather than disposal.

Conclusion

The CM helpline was designed to collapse the whole of government into one number a citizen could trust. A dashboard that counts only closed files cannot tell whether it does. Voice AI redesigns the workflow so "resolved" is checked with the citizen, and so the Chief Minister's office can see the result in time to act.

Government leaders exploring AI-powered citizen engagement can begin with a focused pilot in one district to validate impact before scaling statewide. AiSewak helps public institutions deploy multilingual Voice AI built for Indian governance, including grievance intake and routing and live multilingual voice demos you can try in the browser. To scope a CM-helpline pilot, book a demo.

FAQ

What is a CM helpline? A Chief Minister's helpline is a state-run, single-number citizen-service line, such as 1076 in Uttar Pradesh or 181 in Rajasthan and Madhya Pradesh, where a citizen can lodge a grievance without knowing which department owns it. The helpline classifies the complaint and routes it to the responsible department.

What should a CM helpline performance dashboard show? At minimum: answer rate, abandonment, misroute rate, SLA-breach rate, disposal rate, citizen-verified resolution, reopen rate and time to resolution, with a language breakdown. The key pairing is disposal rate against citizen-verified resolution.

How does AI help a CM helpline specifically? Voice AI can answer every call in the caller's language, propose the department and category at intake, create the record in the grievance portal where an interface exists, escalate on deadline breach, and call the citizen back to ask whether the problem is fixed. The share of calls it can complete without a human depends on the helpline's own call data.

Why do disposal and satisfaction differ? Disposal counts files marked closed. Satisfaction asks the citizen. CPGRAMS's feedback call centre found 44% of the citizens it contacted satisfied in November 2024 and 76% for January to June 2026 (DARPG; PIB). Different helplines measure different things, so compare states with caution.

Does Voice AI replace human helpline operators? No. The design is human-in-the-loop: AI takes routine traffic, and distress cases, allegations and ambiguous grievances go to a person with full context.

What languages can an AI CM helpline support? Bhashini lists 36 text and 23 voice languages (PIB, 12 March 2026). A listed language still has to be tested on real phone audio from the state's own callers before it is promised.

How is a CM helpline linked to the CM's command centre? Every AI-handled call produces structured data, including department, category, location, deadline status, verified outcome and sentiment, which can stream to a CMO dashboard.

How much does an AI CM-helpline pilot cost? AiSewak does not publish a price in this article. Ask each vendor for a capped proposal against your own call volumes and compare it with the fully loaded cost of your current seats.

Is voice-based grievance lodging proven in India? It is live at national level: DARPG launched the CPGRAMS voice chatbot Samadhan Didi on 30 May 2026 (PIB). AiSewak has not found published outcome data for it yet.

How is data privacy handled? CM-helpline calls contain personal data, so a deployment needs consent capture, access controls and data residency. The DPDP Act is only partly in force; its consent and penalty provisions commence on 13 May 2027. See DPDP Act, Data Privacy and Security for Government Voice AI.

How quickly can a state see results? A 30-day pilot can produce a first measured result against a recorded baseline. Whether that result justifies scaling is for the state's own data to show.

Schema Markup Suggestions

  • Article (or TechArticle): headline, description, datePublished (2026-07-04), dateModified (2026-11-10), publisher, keywords, articleSection ("Government Helpline Playbooks").
  • FAQPage: mark up the FAQ section, each Question with its acceptedAnswer Answer.
  • GovernmentService: model the CM helpline as a service with serviceType ("Citizen Grievance Redressal"), provider (State Government), areaServed (State) and availableChannel (telephone number, for example 1076 or 181).
  • BreadcrumbList: Home, Blog, Government Helpline Playbooks, AI for CM Helpline.
  • Organization: publisher (AiSewak Editorial) with logo and sameAs.

Suggested External References

  • Madhya Pradesh CM Helpline portal (cmhelpline.mp.gov.in): cumulative registered and disposed counters.
  • DARPG monthly report, November 2024, published via PIB: Feedback Call Centre satisfaction rates.
  • PIB backgrounder on CPGRAMS, 9 August 2026: satisfaction January to June 2026, Samadhan Didi, NextGen CPGRAMS.
  • PIB, 30 May 2026: launch of Samadhan Didi.
  • PIB, 12 March 2026: Digital India BHASHINI Division, languages and scale.
  • First India feature on Rajasthan Sampark 181, quoting the Chief Secretary; IANS report of 28 April 2026.
  • KRC Times, 5 July 2019: launch of UP CM Helpline 1076.

Social Media Summary

A CM helpline can report that most files are disposed, and that still does not tell you whether citizens' problems were fixed. A good dashboard puts disposal next to citizen-verified resolution, answer rate and deadline breaches. CPGRAMS found 44% satisfaction in Nov 2024 and 76% in Jan to Jun 2026 by calling citizens back. Voice AI can make that call-back routine. #GovTech #VoiceAI #Governance

LinkedIn Executive Summary

Every state's CM helpline carries the government's name, and its dashboard decides what the Chief Minister sees. Madhya Pradesh's portal shows 98.6% of registered complaints disposed. CPGRAMS, which phones citizens after closure, found 44% satisfied in November 2024 and 76% in January to June 2026. Those are different measurements, and both are needed.

Voice AI can instrument the whole lifecycle: round-the-clock intake in the caller's language, automatic classification, portal handoff, deadline escalation and an automated call-back that asks whether the problem is fixed. It is a design, not a proven result: DARPG's Samadhan Didi shows voice-first grievance intake at national level, but outcome data for AI in grievance lines is still thin.

The fastest, safest way in is a 30-day, multi-district pilot with a recorded baseline that measures verified resolution, not disposal.

AI Search Optimization Summary

  • Primary entities: CM Helpline, UP CM Helpline 1076, Rajasthan Sampark 181, MP CM Helpline 181, Bhashini, DARPG, CPGRAMS, Samadhan Didi, MeitY, CMO Command Centre.
  • Core topics: CM helpline performance dashboard, disposal versus citizen-verified resolution, SLA escalation, multilingual voice AI, citizen call-back, government call-centre modernisation, 30-day pilot.
  • Semantic keywords: CM helpline dashboard KPIs, grievance auto-classification, citizen call-back verification, answer rate, SLA-breach rate, per-language acceptance testing, command-centre dashboard, human-in-the-loop government.
  • Question intents to capture: "what is a CM helpline", "what should a CM helpline dashboard show", "how does AI improve a CM helpline", "disposal rate versus satisfaction", "which languages can an AI helpline support", "does AI replace helpline operators".
  • Answer-ready facts (each from a fetched source): MP CM Helpline portal counter read 10 November 2026: 4,07,64,614 registered and 4,01,99,211 disposed; CPGRAMS feedback satisfaction 44% (November 2024) and 76% (January to June 2026); Samadhan Didi launched 30 May 2026; Bhashini lists 36 text and 23 voice languages (PIB, 12 March 2026); UP 1076 launched 5 July 2019 with a 500-seat centre.

AiSewak (AI Sewak) is a Voxdonna company, made in India.

© 2026 Donna AI Labs Private Limited · CIN U62013DL2026PTC464877. All rights reserved.