Skip to content
Peepalytics AI
FIELD NOTES · DELIVERY & GOVERNMENT

The model flags. The officer decides.

How to give AI a bounded job in government, with structured output, evaluation gates and a named decision-maker.

Give the model a job you can describe

“Use AI” is not an operating requirement. Classify this document, extract these fields, identify a missing item or flag a mismatch against a stated rule: those are jobs a team can scope and evaluate. The distinction matters in government because a model’s useful contribution is often much narrower than the decision the service ultimately makes. That narrowness is a strength.

We build AI when it removes work. Much of what makes a service painful is process, not intelligence. A clear form, a configurable requirement and a routed queue can remove more friction than a model added to an ambiguous workflow. The first design task is to identify the repeated work. Only then is it sensible to ask whether that work benefits from model assistance.

Structure the output before it enters the workflow

A paragraph of plausible prose is difficult to use as a controlled system input. Structured output gives each field a defined place and type. Schema validation checks that the returned object has the expected shape before another part of the application relies on it. A missing field, an invalid type or an unexpected value becomes a condition the software can handle explicitly.

Schema validation is not a truth test. A date can have the correct format and still be the wrong date. That is why structure, rule checking, evaluation and officer review are different controls. They should not be presented as interchangeable guarantees. Each addresses a specific way the model or the surrounding process can fail.

The document prompt should also have a narrow boundary. Our specified pattern uses the document image and its expected type. The model does not need an applicant’s entire history to check one document. Restricting the input makes the task easier to inspect and reduces the amount of information sent into a processing step. Data handling terms remain part of the deployment agreement.

Evaluate the task, not the demonstration

A model that performs well on a few clean examples has passed a demonstration. A production gate requires a defined evaluation set and a measure suited to the actual task. A golden set provides examples with known expected outcomes. Precision gates in continuous integration can prevent a change from proceeding when it no longer meets the agreed threshold.

The evaluation set should be designed around the documents the service expects to encounter. Different document types can fail differently. A system that handles a clean typed certificate may struggle with a scan containing stamps, folds or an unusual layout. The point is not to promise universal interpretation. It is to establish what the bounded task is allowed to do and where it must defer.

A rules-only profile is a useful part of that design. Not every document type needs a model at every step. Some checks can be expressed directly and consistently. Keeping a rules-only option makes it possible to use model assistance deliberately rather than treating it as an unavoidable dependency across the whole service.

Keep the decision with a named officer

The interface should distinguish what the model observed from what the officer decided. A flagged mismatch is a request for attention. It is not an automatic refusal. A completeness suggestion is an aid to review. It is not authority to issue a licence. When those distinctions disappear into a single green status, the service loses the explanation of how its decision was made.

The officer needs to see the relevant document, the extracted information and the reason for a flag. The workflow should support a recorded decision and, where appropriate, a recorded reason for departing from a suggestion. The model’s output and the officer’s action belong in the same history, but they remain different events with different authors.

This is also a training question. Officers should understand which checks are deterministic, which are model-assisted and which require their judgement. Training should use the production workflow and realistic examples of uncertainty. A system does not become accountable merely because a button is labelled “human approval”. The person pressing it needs enough context to exercise responsibility.

Provenance is part of the output

Every model call should be traceable to a model identifier, a prompt version and a timestamp. Without those details, an unexpected result becomes difficult to investigate after the system has changed. Provenance links the output to the conditions under which it was produced. It supports review, incident analysis and controlled improvement.

Processing terms matter alongside application design. Zero model-provider retention and no training on client data are specified controls, not decorative phrases for the marketing page. The delivery team must ensure the chosen service and its configuration support the agreed terms. The ministry should be able to examine the processing arrangement as part of its due diligence.

The acceptance conversation should include what happens when the model cannot return a useful answer. The service might need to send the document for manual review, record a processing failure or permit a retry under controlled conditions. Those paths should be visible to the officer. An empty field should not silently become a finding that a requirement is absent. A timeout should not look like a completed check. Designing those distinctions keeps operational failures from being mistaken for administrative decisions. It also provides the support team with enough context to investigate a problem without asking an officer to reconstruct the whole application from memory.

The resulting design is deliberately restrained. A model reads a bounded input and returns structured information. Validation and evaluation constrain its use. An officer sees what matters and makes the decision. The audit trail records both contributions. AI earns its place by removing repeated work from that sequence, while leaving responsibility visible.

All field notes
OFFLINE EDITION · SEPTEMBER 2026Pages, search and PDF downloads work offline. Email requires your mail app.

Search Peepalytics.

Search the work, platform and journal. Use ↑ ↓ to choose and Enter to open.

25 results

WORKNational Business Licence PlatformAround 300 licensable activities, six referral agencies and four approval stages brought into one citizen-facing process. Licensing
WORKTaxi & Tour Operator Licensing RegistryOperator licensing reduced from months to days, with a searchable registry and automated renewals. Live since 2025. Licensing
WORKWAVE Workforce & Payroll PlatformPayroll preparation reduced from a day per cycle to hours. Five reports consolidated into one, on Caribbean-resident infrastructure. Workforce
WORKTrade & CSME Paperwork AutomationTrade paperwork reduced from weeks to minutes, with reporting prepared for Cabinet. Workflow
WORKStatutory Reporting for a National Social Security FundStatutory reporting reduced from more than two months to minutes. Analytics
WORKRenewable Energy Interconnection ApprovalsOfficer document-review time reduced by 90 percent in renewable-energy interconnection approvals. Document AI
WORKNational Festival AccreditationAccreditation workflow and credential issuance automated, with on-site devices. Live in 2026. Workflow
PLATFORMRegistry & LicensingApply, route, review, request information, decide, issue and renew. A public requirements portal explains what is needed before an applicant signs in. Online and assisted submission feed the same officer queue. One application. Every requirement. A decision you can trace.
PLATFORMWorkflow & Referral RoutingA category-to-agency routing matrix sends each application to the right reviewers. Reviews can run in parallel, with the statutory clock counted in working days and every extension recorded. Across agencies. In parallel. With a clock on every step.
PLATFORMDocument IntelligenceClassify, extract, check completeness and apply rules. Every output is structured and schema-validated before it reaches an officer. Only the document image and expected document type enter the prompt. The model flags. A named officer decides.
PLATFORMWorkforce (WAVE)Process and reconcile payroll on change-only cycles. Consolidate reports, generate letters and forms, support recruitment with AI-assisted screening and manage evaluation cycles. Payroll, people and reporting, on a stable rules base.
PLATFORMAnalytics & Plain-English QueryBring entities into one executive view. An exception engine applies thresholds and raises alerts. Ask a question in plain English and see the source and the working behind the answer. One executive view. The source behind every figure.
JOURNALThe form is online. The process is not.A digital front door is useful. The work between the desks is where a public service succeeds or fails.
JOURNALDigitising a multi-agency process when not every agency is ready.Tiered participation lets a shared service move forward without pretending every office has the same infrastructure.
JOURNALThe model flags. The officer decides.How to give AI a bounded job in government, with structured output, evaluation gates and a named decision-maker.
JOURNALWhat a rehearsed migration actually involves.A cut-over plan is useful only after the team has exercised it against the records, checks and responsibilities that matter.
JOURNALOne project, inside one budget cycle.The case for a bounded service, clear ownership and a result that can earn the next engagement.
COMPANYOne project. Delivered.HOW WE WORKOne project. Delivered.Scope first. Costing in your tender format. A service that runs, and a handover you can use.A product engineering team.A product engineering team, not a body shop. Around twenty specialists across engineering, data, machine learning and design work under one roof. Build, integration and advisory run together.We start with a working session, not a price. We need to understand the queue, the rules, the integrations and the people responsible for the service before we can define the commitment.From first meetingto live service.01Working sessionStart with the officers and the work they are responsible for.02ScopingDefine the queue, the rules, the integrations and the acceptance criteria.03Tender documentsPrepare scope and costing in your procurement format.04Build with previewsReview working software as the system takes shape.05User acceptanceRun scripted cases, record defects and verify the fixes.06Rehearsed migrationRehearse the cut-over and validate the transferred records.07Go-live & supportLaunch with 30 days of heightened support.08Training & handoverTrain by role. Hand over source, data, documentation and runbooks.Procurement, done properly.Multi-bidder tenders, evaluation panels and due-diligence responses are part of public delivery. We prepare the technical, security, hosting and costing documentation in your format.Your Procurement Board and Ministry of Finance need a scope they can evaluate. We document the dependencies, delivery steps and acceptance criteria alongside the cost.Handover is the deliverable.100%of source code, data, documentation and runbooksYou own the system and the data. The ministry is the data controller. Export your data in a machine-readable format at any time, at no charge.Versioned migrations, rehearsed cut-overs and written incident reports make the operating record available to the next person responsible.Training, on the system.Role-based training takes place in go-live week, on the production system, and is recorded. The officer needs to perform the real task: receive a record, review it, request information and make a decision.A deliberate boundary.We do not replace systems that already work.We do not make every agency adopt the same tools on the same date.We do not introduce AI where a clear form or rule removes the work.We do not make the first project conditional on a multi-year programme.THE NEXT CONVERSATIONStart with a conversation.A working session with your digital transformation team. No commitment, no proposal, no numbers.Let’s talk ↗
COMPANYDigital government, delivered inside one budget cycle.FOR GOVERNMENTSDigital government, delivered inside one budget cycle.A service your officers can operate. A delivery record your evaluators can examine. A system you own.Three places to start.Option 01 · FoundationCitizen authenticationA dependable front door for public services, with online and assisted access.Discuss this starting point↗Option 02 · ProvenBusiness licensing & permitsA requirements catalogue, routed reviews and a register you can maintain.Discuss this starting point↗Option 03 · High volumeCivil registry & certificatesA focused service where a repeatable process removes a large queue.Discuss this starting point↗One project, delivered, inside one budget cycle. Let the result decide whether there is a second.OR, START WITH YOUR PROCESSBring us your queue.A licensing backlog, a report that takes too long, a process that crosses agencies. Start with the work you are responsible for.Discuss your service↗Built for the waygovernment works.01 / MULTI-AGENCYAgencies that are not readyIntegrated agencies, PDF despatches and clerk-updated status can participate in the same process. Nobody has to wait for every office to be ready.02 / SELF-SERVICEChange without a developerYour officers configure categories, requirements, questions and fees. Routine policy changes should not become a software project.03 / TENDER-READYProcurement, done properlyScope, technical requirements, security, hosting and costing documentation, prepared in your procurement format.04 / ACCESSNobody is excludedOnline and assisted submission enter the same workflow. A service still has to work for the person who walks up to the counter.Alongside the national programme.Interoperability frameworks, electronic identity, payments and single windows establish the national context. We build a useful service within that context and integrate with what already works.A focused first project can establish a clear operating result while the wider programme progresses.THE AI, DISCIPLINEDThe model flags.A named officer decides.01 Structured and schema-validated02 Logged: model, prompt version, timestamp03 Zero model-provider data retention04 No training on your dataREAD THE CONTROLS ↗A record you can examine.Six government systems delivered or in build for the Government of St. Kitts & Nevis, over more than two years. Operator licensing live since 2025. A national workforce platform and event accreditation among the delivery record.References are available on request. Technical controls, data residency, procurement documentation and ownership terms are available for evaluation.Read the case studies↗Review the controls↗THE NEXT CONVERSATIONStart with a conversation.A working session with your digital transformation team. No commitment, no proposal, no numbers.Let’s talk ↗
COMPANYThe same engines. Your operation.FOR REGULATED ENTERPRISEThe same engines. Your operation.Document automation, workflow, analytics and workforce systems for organisations that need to show their working.01 / CAPABILITYDocument automation & compliance90% less document-review time in renewable-energy interconnection approvals.02 / CAPABILITYWorkflows that route themselvesTrade paperwork reduced from weeks to minutes.03 / CAPABILITYForecasting, exceptions & queryAI analytics platforms for regulated companies in the United States, including financial services and healthcare.04 / CAPABILITYWorkforce & payrollA live government workforce platform. Payroll preparation from a day per cycle to hours.Where the engines apply.HOSPITALITYDemand forecasting · dynamic pricing · workforce reportingUTILITIESDocument checks · approvals · exception monitoringFINANCIAL INSTITUTIONSCompliance workflows · board packs · workforce processingMANUFACTURING & DISTRIBUTIONBatch traceability · procurement and FX exposure modelling · farmer and distributor portalsSTATUTORY BODIESConsolidated reporting · document routing · payroll preparationApplications to scope with your team, using the same underlying engines.What we do not oversell.Our production record is government and regulated enterprise. The engine transfers; the domain knowledge comes from your team. We integrate with the systems you rely on.We start with the work, the source data and the rules. Where a new domain needs validation, that validation belongs in the scope.THE NEXT CONVERSATIONStart with a conversation.A working session with your digital transformation team. No commitment, no proposal, no numbers.Let’s talk ↗
COMPANYBuilt here. With a wider perspective.THE COMPANYBuilt here. With a wider perspective.Technology in the service of our people.Peepalytics AI, Inc.A Delaware C-corporation headquartered in Brooklyn, licensed to conduct business in the Federation of St. Kitts and Nevis. Offices in Brooklyn, Basseterre and Port of Spain.We build and run software for government and regulated enterprise: licensing, registries, workflows, payroll, document intelligence and analytics.PEEPALYTICS AI · BROOKLYN · BASSETERRE · PORT OF SPAIN · PFOUNDER & CEOKareem Williams.A citizen and resident of St. Kitts & Nevis, with Grenadian and Trinidadian family roots.Formerly Head of People Analytics at Pinterest, Data Science Research Lead at Amazon, and a Data Scientist with AWS AI/ML Professional Services. Earlier work at Intel in computer vision.Peepalytics brings the engineering discipline of those organisations home to the Caribbean, in systems our institutions can operate and own.Around twenty specialists.Engineering, data, machine learning and design under one roof. A team with experience at Pinterest, Amazon and AWS.Product engineeringData & machine learningDesign & deliveryPartners in delivery.A change-management practice of organisational psychologists supports the human side of adoption. A hardware partner supports field devices where a service extends beyond the screen.THREE OFFICES. ONE TEAM.From Brooklyn to the Caribbean.BrooklynNew York, United StatesBasseterreSt. Kitts & NevisPort of SpainTrinidad & TobagoArrange a conversation↗THE NEXT CONVERSATIONStart with a conversation.A working session with your digital transformation team. No commitment, no proposal, no numbers.Let’s talk ↗
COMPANYStart with a conversation.THE FIRST STEPStart with a conversation.Tell us about the work you are responsible for. We will start there.Your nameWork emailOrganisationYour roleCountrySt. Kitts and NevisAntigua and BarbudaBahamasBarbadosBelizeDominicaDominican RepublicGrenadaGuyanaHaitiJamaicaSaint LuciaSaint Vincent and the GrenadinesSurinameTrinidad and TobagoAnguillaArubaCuraçaoMontserratSint MaartenUnited StatesUnited KingdomOtherWhat are you responsible for?How did you hear about us? OptionalThis offline copy prepares an enquiry for you. Choose email to open your mail app, then review and send it with an internet connection. You can also save the enquiry as a text file. Nothing is submitted or stored by this HTML file.Prepare email ↗Save enquiry ↓DIRECT CONTACTsales@peepalytics.comA working session with the people responsible for your service. We aim to respond within one working day.BrooklynNew York, United StatesBasseterreSt. Kitts & NevisPort of SpainTrinidad & TobagoDocuments for your evaluators↗
COMPANYThe controls behind the work.SECURITY · DATA · DELIVERYThe controls behind the work.A reference for technical evaluators and procurement teams. Engagement-specific configurations and supporting documents are available on request.LAST REVIEWED · 09 SEPTEMBER 202601 / Company and incorporation.Peepalytics AI, Inc. is a Delaware C-corporation headquartered in Brooklyn and licensed to conduct business in the Federation of St. Kitts and Nevis. We work across Brooklyn, Basseterre and Port of Spain.02 / Security architecture.Public access uses email OTP. Officer accounts are invite-only. Administrators use MFA. Role-based access is mirrored by row-level security. Documents are scanned for malware and accessed through signed, 15-minute URLs. Append-only audit trails are exportable.03 / Data protection and residency.Systems can run in the government’s own cloud account or on Caribbean-resident infrastructure. The ministry is the data controller. Data can be exported in a machine-readable format at any time at no charge. Deployment location and processing terms are agreed for each engagement.04 / AI processing controls.Outputs are structured and schema-validated. Every call records the model, prompt version and timestamp. The model flags; a named officer decides. Processing is specified with zero model-provider retention and no training on client data. Per-document-type rules-only profiles and a golden-set evaluation harness constrain model use.05 / Hosting and infrastructure.TLS 1.2 or later with HSTS, encryption at rest, SHA-256 integrity hashes and managed secrets form the infrastructure baseline. Daily backups are paired with rehearsed restoration. Versioned migrations, infrastructure as code and per-change previews support controlled release.06 / Testing and quality.Scripted acceptance tests, evaluation gates and preview deployments give officers and technical reviewers a repeatable way to verify behaviour. Migration is rehearsed before cut-over. Incidents are documented in writing.07 / Handover and ownership.The deliverable includes 100 percent of source code, data, documentation and runbooks. Role-based training takes place during go-live week, on the production system, and is recorded. Ownership is not a subscription feature.08 / Documents available on request.An NDA, a DPA on your form, technical prerequisites, reference letters and procurement documentation can be provided through a working conversation. References are available on request.THE NEXT CONVERSATIONStart with a conversation.A working session with your digital transformation team. No commitment, no proposal, no numbers.Let’s talk ↗
COMPANYBuild the systems our region needs.CAREERSBuild the systems our region needs.We hire from the Caribbean diaspora deliberately. The mission is public. The engineering has to hold up.How we work.Distributed across Brooklyn, the Federation and Trinidad. Trunk-based development, continuous integration gates and preview deployments. Rehearsed migrations and written incident reports.We value clear written thinking, care for the officer using the system and the ability to turn an operating requirement into dependable software.Open roles.No open roles are currently listed.If your work belongs here, introduce yourself and tell us what you have built.Introduce yourself↗
COMPANYA record you can forward.FOR YOUR NEXT INTERNAL CONVERSATIONA record you can forward.Concise capability and trust briefs, drawn from the same content as this website.PDF · SEPTEMBER 2026Capability overviewOur delivery record, five engines and approach to government software.Download PDF ↓PDF · SEPTEMBER 2026Trust summarySecurity, data processing, residency, testing and handover controls.Download PDF ↓THE NEXT CONVERSATIONStart with a conversation.A working session with your digital transformation team. No commitment, no proposal, no numbers.Let’s talk ↗

Brooklyn · Basseterre · Port of Spain