06 — Public Sector
Casework, eligibility and citizen service modelled as outcomes per pound — with the highest bar for explainability, appeal rights and bias testing.
North Star
Cost per resolved citizen outcome
The value engine, end to end
The same six moves we run everywhere, expressed in the physics of this industry.
Where do citizens wait and cases stall?
Service value map
What drives cycle time and error?
Casework causal model
What if we change triage or staffing?
Service delivery twin
Who assists the caseworker?
Assistive agent layer
Is it explainable and appealable?
Algorithmic transparency record
Does backlog fall and stay down?
Outcome contract
Where the value is
Before a twin is built, we agree where the money sits and what moves it.
Value pool
Casework throughput
Driver
Cycle time, rework, backlog
Measured as
Cases resolved per FTE
Value pool
Error & fraud
Driver
Overpayment, eligibility error
Measured as
Value recovered or prevented
Value pool
Contact cost
Driver
Avoidable contact, channel shift
Measured as
Cost per interaction
Value pool
Outcome quality
Driver
Appeal rate, satisfaction, equity
Measured as
Rework and trust cost
Use cases
Each one is a number somebody already owns, a twin that models it, and an agent that moves it inside a defined authority. Filter by function.
5 use cases
Reads the application and supporting evidence, checks eligibility against the published rule set, and shows the citation for every determination.
Structures submissions against the published criteria so evaluators compare like with like, and records why each score was given.
Detects organised and opportunistic misuse while keeping the false-positive burden off legitimate claimants.
Answers the routine question correctly the first time, in the citizen's language, with a named route to a human when it is not routine.
Ranks maintenance and renewal against condition, consequence and public value instead of the order the requests arrived in.
Outcome twins
Each twin owns a set of levers. Each feature moves one lever. Each agent executes inside a defined authority.
Model the queue, not just the case.
Levers
Features
Multi-agent layer
Casework Assist Agent
Prepares the case pack; the officer decides.
Find error and fraud without punishing legitimate claimants.
Levers
Features
Multi-agent layer
Integrity Agent
Flags anomalies with reasons and an appeal path.
Answer first time, in the citizen's channel and language.
Levers
Features
Multi-agent layer
Service Agent
Resolves status and guidance queries with citations.
Unit economics
Set your volume, switch features on or off, and move each impact to your own evidence. Everything is expressed per case.
Cases per year
600,000
Contribution per case
$210
Public value: benefit is cost avoided and error prevented, not revenue.
Annual program investment
$3.5m
Casework Twin
Unit cost — Cost per case
6.0% — Right case to the right officer first time.
$7.6m
annual value
Casework Twin
Unit cost — Handling time
5.0% — Officers read a pack, not a folder.
$6.3m
annual value
Integrity Twin
Leakage — Error and overpayment
20.0% — Prevention before payment, with an appeal path.
$1.5m
annual value
Citizen Service Twin
Unit cost — Contact cost per case
3.0% — Status answered without a phone call.
$3.8m
annual value
Value by feature
Annual value contributed by each active feature at current settings.
Value and ROI vs volume
How engineered value and ROI move as cases per year change. The marker is your current setting.
Cumulative value vs investment
Where the curve crosses the investment line is payback.
Value mix by lever
Which lever the engineered value is actually coming from.
Baseline contribution
$126.0m
Engineered annual value
$19.2m
ROI on program
447%
Payback
2.2 months
Directional model. In an engagement every number here is replaced by your measured baseline, signed off by finance, and written into the outcome contract before any agent goes live.
Scale
Prove cycle time and equity metrics side by side.
Reuse the pattern across adjacent services.
Shared standards, shared assurance evidence.
Integrations
Vendor-agnostic by design. Connectors are added per engagement — no platform lock-in.
Case & CRM
Identity
Data & cloud
AI
Regulations and standards
Compliance is a design input, not a review gate. Every agent action is logged, attributable and reversible.
EU AI Act
Essential-services eligibility AI is high-risk: FRIA, oversight, registration.
GDPR / national DP law
Lawful basis, Art. 22 rights, DPIA per service.
Algorithmic transparency
Published record for any citizen-affecting model.
Equality duty
Bias testing across protected characteristics.
Accessibility law
WCAG-conformant AI interfaces.
Standards supported
Next industry
Insurance
