Reads each grievance from portals, call-centre notes, letters and social posts, detects language and topic, sets a priority level and routes it to the responsible office.
Agents that prepare the routine work of government — reading grievances, checking beneficiary lists, extracting tender requirements, reconciling fund flows, tracking cases and actions — on the governed data fabric. In public-sector squads agents observe, suggest or prepare; a named official decides every case that affects a citizen's entitlement, a penalty, a tender award or a payment.
These are agent designs from our AI & Agentic Engineering practice, mapped to the public-sector value chain for central and state departments, PSUs and cities. The live agent demo runs on banking sample data under the same controls.
Grievances are read, masked, classified and routed the day they arrive; grievance officers approve every reply.
Reads each grievance from portals, call-centre notes, letters and social posts, detects language and topic, sets a priority level and routes it to the responsible office.
Detects and masks names, phone numbers, ID numbers and addresses before any grievance or file text reaches a language model.
Drafts a reply in the citizen's language, citing the rule or circular it relies on, for the grievance officer to edit and send.
Tracks every service request and grievance against its citizen-charter timeline and warns the owning office before it lapses.
Duplicates, ineligible entries and failed payments are flagged with evidence; officers verify and decide every inclusion or exclusion.
Finds likely duplicate and ineligible beneficiaries across registers using consented identifiers and fuzzy matching, and assembles the evidence for each match.
Groups failed and returned DBT payments by cause and drafts the corrective step for each group.
Answers field staff's questions on scheme guidelines, quoting the clause and circular each answer comes from.
Revenue gaps and arrears are surfaced and ranked; officers choose every inspection and sign every notice.
Compares who should be paying with who is paying — returns, property records and third-party information — and quantifies the gap by area and category.
Scores returns and properties for inspection, with the reasons behind each score, so officers can choose where to go.
Prioritises arrears by amount and age and drafts the recovery notice for the officer to review and sign.
Matches collections and challans with treasury credits and explains each break.
See Revenue Assurance →Agents read tenders and bids and prepare a page-linked evaluation; the tender committee decides every award.
Extracts eligibility, technical and commercial requirements from tender documents into a checklist, citing the clause and page.
Turns vendor bid PDFs, including scans, into line items with unit prices, quantities and totals, each linked to the page it came from.
Checks each bid against the tender conditions and lists deviations and missing documents with references.
Tracks delivery and payment milestones on awarded contracts and flags slippage and risk.
Fund flows are reconciled, legal exposure is tracked and audit replies are assembled; finance and law officers approve.
Follows each release from sanction through agencies to final payment and flags unspent balances and pending utilisation certificates.
Matches treasury, bank and scheme-account entries and explains each break for the accounts officer.
See CLARION →Tracks cases and arbitrations against the department by court, status, category and age, totals the financial exposure and flags critical aged and high-value cases.
Turns review decisions into owned actions, emails the owning office with the data extract, SLA and escalation path, and chases overdue actions.
Gathers the documents and data behind each audit para and drafts the reply for the officer to approve.
Civic complaints and assets are routed and prioritised; engineers and field staff act on prepared work lists.
Classifies civic complaints, detects repeats at the same location and routes them to the ward and service team.
Ranks roads, pipes, lights and other assets for maintenance from complaints, inspections and age.
Captures inspections offline with GPS, readings and photos and checks completeness before sync.
Targets, portfolio, projects and cash in one command centre; management decides, agents prepare the pack.
Keeps the strategic command centre current — targets, portfolio, disbursements, projects and cash — and explains what changed since the last review.
Flags projects and capex lines drifting from schedule or budget, with the drivers and the money at stake.
Follows budget to purchase order, receipt and consumption, and explains mismatches.
Drafts the board pack and ministry report from governed data, with commentary for management to edit.
See NARRATOR →Data is catalogued, consented and masked; every agent is inventoried, evaluated and auditable.
Proposes glossary terms, owners and lineage for new datasets and routes them through the stewardship workflow.
Checks that each use of personal data has a recorded purpose and consent under the DPDP Act and flags gaps.
Runs quality rules on critical registers and reports failures to the owning department.
Reviews agent logs, overrides and evaluation results and reports drift or policy breaches to the AI governance board.
See agent governance →Policy decides what goes straight through; people approve everything else; a kill switch halts all agents. The live demo runs on banking sample data — the same controls apply to every Public Sector squad.