Process Flow-MIS

Someone AI
TCIG_GenAI_Vehicle_Licence_Renewal.docx

Using Generative AI to Improve Vehicle Licence Renewal

A proposal for the leadership of the Turks and Caicos Islands Government

Management Information Systems | Group assignment

Prepared by: Vanessa Rolle and group members Submission date: [Insert date]

Executive summary

TCIG’s ePay system already produces an official receipt. The improvement opportunity examined in this report is therefore the work surrounding a vehicle licence renewal: explaining requirements, preparing the application, reviewing supporting information and communicating about outstanding issues. These activities can still require considerable staff attention even when payment and receipting are electronic. Their actual frequency and duration should be measured before implementation.

This report proposes a GenAI-assisted application preparation and review process for vehicle licence renewal. The assistant would use approved DMV guidance to answer questions, prepare draft information from submitted documents and produce a clear summary of omissions or inconsistencies for officers. Staff would verify the evidence and approve the application. Customers would continue to pay through the authorised route, with ePay retaining its existing official receipting function. Where needed and technically feasible, controlled integration would link the approved application with the payment and receipt record.

The proposal preserves officer approval, physical inspection requirements and financial accountability. Its value should be assessed through processing time, rework, receipt accuracy and reconciliation performance. A limited pilot is recommended before wider implementation. The aim is to improve staff capacity and the quality of service, rather than remove staff from the process.

1. Business domain and selected transaction

1.1 Purpose and management problem

The purpose of this report is to recommend a practical use of generative artificial intelligence within TCIG’s vehicle licence renewal transaction. The recommendation is addressed to government leadership and concerns both service delivery and financial control. A customer should be able to understand what is required, submit the necessary information, make the correct payment and receive an official receipt without avoidable repetition. Management, meanwhile, must be able to establish that the renewal was authorised, the money was received and the transaction was recorded correctly.

The management problem selected for examination is the staff effort needed to turn a customer’s request and supporting documents into a complete, reviewable renewal application. ePay already provides an official receipt, so receipt generation is not presented as a missing capability. Instead, the proposal examines whether customers need repeated guidance, officers repeatedly prepare the same information or incomplete applications create avoidable exchanges. These are specific improvement hypotheses to test with staff, not established findings about current DMV performance.

1.2 What is generative artificial intelligence?

Artificial intelligence (AI) is a broad term for computer systems that perform tasks associated with abilities such as interpreting information, recognising patterns and supporting decisions. Generative artificial intelligence, abbreviated as GenAI, is a type of AI that produces content in response to instructions and information supplied to it. That content may be text, images, audio or other digital material. This description is consistent with the National Institute of Standards and Technology’s definition (NIST, n.d.). In this report, the focus is on generating written guidance, draft application summaries and explanations of transaction exceptions.

A large language model (LLM) is the language-processing model that could power the assistant. A prompt is the instruction given to it. For example, a staff member could instruct the assistant to compare an application with an approved checklist and draft a short message explaining what is missing. The output would be a proposed response for review, rather than an official decision. The model generates a likely response based on learned patterns and the context provided; a convincing answer is therefore not proof that the answer is correct.

Consider an applicant who writes, “I paid for my vehicle renewal yesterday but cannot find my receipt.” A conventional payment system can search a reference if one is supplied. A GenAI assistant can help interpret the message, prepare a summary of the issue and draft a request for the information needed to investigate. It cannot establish that the payment was received simply because the applicant says so. That fact must come from the authorised payment record.

1.3 GenAI, digitisation and automation are different

Term

Meaning in this proposal

Example

Digitisation

Capturing information in a digital form.

Customer uploads a document instead of presenting a paper copy.

Rules-based automation

Executing a predefined instruction using approved conditions.

A fee table determines the charge, or a confirmed payment triggers a numbered receipt.

GenAI assistance

Producing a contextual draft from supplied information and approved guidance.

The assistant explains why information is incomplete and drafts a clarification request.

System integration

Allowing authorised systems to exchange structured data.

The receipting system reuses the approved payer details and transaction reference.

The proposed solution needs all four elements, but they have different responsibilities. It would be misleading to present automatic receipt generation as a GenAI achievement. GenAI helps prepare and explain the information surrounding the transaction. Integration and controlled automation allow that verified information to be used without repeated entry. This distinction makes the proposal more credible and helps management decide where the investment is justified.

1.4 Scope, stakeholders and transaction boundary

TCIG provides public services while collecting the fees associated with those services. This report focuses on the renewal of an existing vehicle licence, using the DMV and Treasury interface as the area for improvement. Initial vehicle registration, ownership transfers and driver’s licence applications are outside the scope because they involve different decisions and supporting evidence. The DMV’s published service information identifies vehicle licence registration, including new applications and renewals, as a service and lists supporting documentation such as vehicle insurance and a completed inspection checklist (TCIG DMV, n.d.).

For this assignment, vehicle licence renewal represents the service and revenue collection side of the transaction cycle. Although a government department does not sell services in the same way as a private business, the transaction still involves a request, validation, a charge, collection, receipting and completion of the service. Treasury’s role is therefore closely connected to the DMV’s ability to complete the transaction and maintain reliable records.

2. How the existing process works

2.1 Evidence and process boundary

Based on the operational information supplied for this assignment, Treasury accepts walk-in cash and card payments, bank transfers and direct online payments through ePay. ePay produces an official receipt. Staff also access the relevant system to prepare receipts in staff-handled transactions, although the precise channel-specific responsibilities must be confirmed. The current-state model therefore distinguishes ePay’s existing receipt from transactions requiring staff preparation. It does not assume that every customer’s receipt is manually created.

The process begins when a customer requests renewal and provides the required application information and supporting documents. A DMV officer checks the submission and any relevant vehicle requirements. An incomplete submission must be corrected before the transaction proceeds. Once the application is acceptable, the applicable fee and payment instructions are established.

The customer pays through an authorised channel. For an ePay transaction, the official receipt is generated by ePay and should be retrieved or verified rather than recreated. For staff-handled payments, the authorised officer performs the applicable verification and receipting steps. The confirmed payment and receipt are linked to the renewal before the DMV completes the service. Finance staff reconcile collections and investigate differences between payment, receipt and application records.

The existence of an electronic receipt does not establish whether the application itself is complete, whether the vehicle details are consistent or whether the customer understands what remains outstanding. These are information-processing tasks around the transaction. GenAI could help staff prepare and explain that information, while payment confirmation and formal approval remain under established controls. Any claim about repeated entry or delays must be supported by a walkthrough or measured baseline.

Figure 1 is a working model for officer validation. ePay’s official receipt is confirmed. The sequence of approval and payment, channel responsibilities and any existing connection between DMV and ePay records still require confirmation.

2.2 Responsibilities and information flow

The customer supplies information and pays the charge. DMV officers assess the renewal. Treasury or an authorised collecting officer handles the applicable payment controls. Finance reconciles transactions and ICT supports the systems. Exact internal hand-offs require confirmation because the published service list does not describe them.

Proposed linking fields include the application and vehicle identifiers, payer, service, amount due, amount received, payment reference and receipt number. This is not a claim that TCIG currently has an integrated database. Reliable links allow officers to find information already held without asking the customer to supply it again.

2.3 Effect on the revenue collection cycle

Incomplete information can delay assessment, while document discrepancies create clarification work. At payment, the collection must remain connected to the application. ePay provides an official receipt, but officers must still establish that renewal requirements are met. Reconciliation checks agreement between the application, collection and receipt. GenAI would primarily support preparation, review and follow-up.

The opportunity lies in the quality and movement of information. Faster responses help only when records remain accurate and reconcilable. The proposed MIS should give officers a reliable transaction status and management a clear list of unresolved cases.

3. Current process flow

Figure 1A. Working model of the current vehicle licence renewal and receipting process.

Rounded boxes show activities or start/end points. Diamonds show decisions. Arrows show the sequence and the return route when an application or payment cannot proceed. These are preliminary diagrams; final symbols can be adjusted to the lecturer’s convention.

3.1 Payment-channel detail

Figure 1B. Detail supporting the payment stage in Figure 1A. This is a model for officer validation, not a claim that every channel is available for vehicle renewals.

Symbol key: oval = start, end or continuation; rectangle = processing activity; diamond = decision; parallelogram = information output. The overview figures use rounded activity boxes for readability. Final notation can be aligned with the class convention.

4. Four threats to successful completion

4.1 Incorrect or incomplete transaction information

A customer may supply an incorrect vehicle identifier, omit a required document or provide inconsistent information across files. An officer may also make a transcription error when recording the application. These issues can lead to rework or an incorrect record even where ePay generates an accurate receipt for the payment made. The threat concerns the quality of the application and its link to the transaction, rather than an assumption that electronic receipting is defective.

4.2 Fraudulent payment evidence or unauthorised processing

A submitted payment screenshot or transfer instruction does not establish that the Government has received the funds. If staff accept such evidence without checking an authorised source, a renewal could proceed without valid payment. Unauthorised changes to receipt details or vehicle information could similarly undermine the transaction. The impact would include possible revenue loss, inaccurate records and difficulty establishing responsibility.

4.3 Payments that are omitted, duplicated or left unmatched

A valid receipt must be connected to the correct renewal application. If references or vehicle details are unclear, a genuine payment could remain unmatched to the service requested. A customer who misunderstands the status may also make another payment. The existence of ePay’s official receipt does not remove the need to identify the correct application and reconcile the records. These exposures should be tested against actual channel procedures; duplicate receipts are not alleged to be a known ePay problem.

4.4 System unavailability and interrupted processing

An interruption affecting connectivity, the payment service or the receipting system can prevent staff from verifying or completing transactions. If temporary records are not controlled, information can be lost or entered twice when service resumes. Customers may also receive conflicting advice about whether they should retry payment. The threat therefore concerns both continuity of service and the accuracy of records after recovery.

These four threats require different responses. GenAI can help explain missing or inconsistent application information, but it cannot establish receipt of funds or resolve an accounting difference by itself. Approved transaction references, independent payment verification and reconciliation remain necessary. The proposal uses GenAI where language and document preparation create work, while conventional controls protect the financial transaction.

5. Proposed use of GenAI

The recommended solution is a secure renewal support assistant embedded in the transaction workflow. It would use an approved language model and controlled access to relevant information. It is a proposed capability rather than a claim that TCIG already operates such a system. The assistant would prepare information for officers and help explain exceptions; the systems of record would retain authority over applications, payments and receipts.

5.1 Application preparation and customer guidance

A secure assistant would use an approved, version-controlled collection of DMV instructions to explain the renewal requirements in plain language. Customers could ask questions before submitting their information. Where access to technology is limited, an officer could use the same assistant during a counter visit. This preserves an assisted service rather than making online access a condition of renewal.

Document-reading tools would extract relevant details from submitted files. GenAI would use those details to prepare a draft application summary, identify apparent inconsistencies and draft requests for missing information. Each suggested field should link to the relevant source document so the reviewing officer can check it. An unclear registration number should be referred for review rather than guessed. Only information necessary for the transaction should be processed.

5.2 Officer review, payment and the existing official receipt

The DMV officer would review the draft against the original evidence and the official record. GenAI would not determine legal eligibility, authenticate a document or replace a required vehicle inspection. Once the officer approves the application, a rules-based component would apply the approved fee schedule and create a unique transaction reference.

The customer would then pay online or through an authorised counter channel. A verified payment status would come from the relevant payment source, not from GenAI’s interpretation of a customer message. For cash, the authorised cashier would record the collection under existing cash controls. For card payments or transfers, the applicable confirmation and reconciliation rules would apply.

For an ePay payment, the existing official receipt would remain the receipt for the transaction. The proposal would not generate a second receipt through GenAI. Where authorised and technically feasible, system integration would link that receipt and payment reference to the approved application. Staff-handled channels would retain their applicable receipting procedures, with approved data reused where practical. The GenAI benefit is the preparation of information and explanations around the transaction; receipt linking is a supporting integration task.

5.3 Reconciliation support and the limits of GenAI

Finance officers would receive an exception list of payments that do not match the approved transaction rules. GenAI could summarise the associated correspondence, explain the information available and draft a follow-up query. Staff would verify the explanation and authorise any correction. The assistant would have no independent authority to issue refunds, amend bank details or post financial adjustments.

The proposal combines GenAI with ordinary automation. Generating a contextual explanation or draft query is a GenAI task; calculating a fee, matching an exact reference and generating a numbered receipt are controlled system functions. This distinction matters because TCIG would otherwise risk paying for GenAI where a simpler integration would solve the problem. NIST (2024) identifies confidently incorrect generated content and information security as relevant GenAI risks, reinforcing the need to verify outputs before they affect official records.

5.4 How the proposed system would work

First, DMV management would maintain an approved knowledge base containing the renewal checklist, service guidance and current procedures. Each item would have an owner and effective date. When a question is asked, a retrieval component would select the relevant approved passages and supply them to the language model with the question. This method is commonly called retrieval-augmented generation (RAG). In plain terms, the assistant is given the relevant official material before it drafts an answer. This reduces reliance on the model’s general knowledge, although it does not eliminate mistakes.

Second, document-reading tools, including optical character recognition where necessary, would convert submitted documents into usable text. The assistant would prepare a structured draft containing the relevant renewal details, a source reference for each suggested field and a list of uncertainties. An unreadable number or conflicting name would be marked for review. It would not be silently filled in. The officer would compare the draft with the documents and official records, make corrections and approve the information.

Third, an integration layer would pass only the approved transaction fields to the charging and payment functions. An application programming interface (API) is a controlled mechanism through which systems exchange information. Access would be limited to the required fields and actions. The language model would not receive direct authority to post accounting entries, alter the fee schedule or change the customer’s payment destination.

Fourth, the authorised payment system would supply the confirmed status and existing receipt reference. For ePay this means using its official receipt, not inventing a new receipting function. Pending or uncertain payments would remain unresolved until verified. The application system should check whether the payment and receipt are already linked before recording them again. Failed links would be investigated without asking the customer to pay a second time. These are proposed integration controls; their necessity depends on what the current systems already do.

Finally, staff and management would see different views of the same controlled record. An officer would see the documents, draft suggestions and outstanding actions. Treasury would see payment and receipting status. A supervisor would see processing delays and unresolved exceptions. This is the management information systems contribution: information is captured once where possible, verified by the responsible officer and made available for authorised operational and management use.

5.5 A realistic example

Suppose a renewal applicant submits an application and supporting documents, but the vehicle identifier entered on the form differs from the identifier on one document. The assistant prepares a comparison and drafts a message explaining the inconsistency. The officer checks the original documents and decides whether the customer must correct the submission. This saves preparation time while leaving the decision with the person responsible for it.

Once the corrected application is approved, the authorised charging process determines the fee and provides the payment reference. If the customer pays through ePay, ePay produces the official receipt. The officer verifies the linked payment information and completes the approved renewal process. The improvement has occurred mainly before payment: the assistant prepared a comparison and one clear clarification message, so the officer could focus on the decision. Time saved is a pilot question rather than an assumed result.

5.6 How staff work would change

Activity

Current model

Proposed model

Customer guidance

Officers repeatedly explain requirements.

GenAI drafts guidance from approved material; officers handle uncertainty and complex questions.

Application preparation

Officers read and enter information.

Tools prepare draft fields and a GenAI summary; officers verify and approve them.

Payment and receipt follow-up

ePay issues an official receipt; staff support other channels and resolve queries.

Retain ePay receipting. GenAI drafts query summaries; controlled integration links the existing receipt where required.

Unmatched payments

Officers gather details and prepare correspondence.

GenAI summarises authorised evidence and drafts queries; finance verifies and resolves them.

Management review

Supervisors depend on transaction reports and explanations.

Structured reports identify outstanding cases; GenAI may draft a narrative for review.

Improving staff does not require every transaction to involve the same number of manual actions as before. Routine entry can be reduced while staff retain responsibility for decisions and exceptions. The proposal would therefore change the balance of work rather than justify redundancies. Its success should be judged by whether officers have better information and more time for work requiring judgement, not by a reduction in headcount.

5.7 Why GenAI rather than a simpler alternative?

Leadership should compare three options. Keeping the existing process avoids new technology costs but leaves the selected manual work in place. Integrating application, payment and receipt records could remove repeated entry without GenAI. Adding a controlled GenAI assistant could provide further value where the remaining workload involves varied questions, inconsistent documents and lengthy correspondence. The third option is justified only if these language-based tasks are sufficiently frequent and the pilot demonstrates reliable assistance.

The recommended sequence is therefore to establish sound data and transaction controls first, then add GenAI for clearly defined preparation and communication tasks. If an ordinary checklist or integration solves a problem adequately, it should be used. This avoids making an expensive language model responsible for predictable calculations or exact matching that a conventional system can perform more reliably.

6. Recommended controls

Threat

Control and responsibility

Evidence of operation

1. Incorrect or incomplete information

DMV requires mandatory fields and approved validation rules. The officer compares draft fields with source documents before approval. Finance controls revenue codes and fee data.

Validation results, linked source documents and a dated officer approval.

2. Fraud or unauthorised processing

Treasury confirms payment from an authorised source. Role-based access separates application approval, receipt adjustments and reconciliation. Cash is subject to cashier balancing and supervisory review.

Payment confirmation, access logs, cash balancing records and authorised adjustment history.

3. Omitted, duplicate or unmatched payments

Use one transaction reference throughout. Prevent repeat processing of the same payment event. Finance reconciles payment, receipt and application records daily, with supervisor follow-up of aged exceptions.

Reconciliation report, duplicate checks and an exception register showing owner and resolution.

4. System interruption

ICT maintains recovery procedures and tested backups. Staff use an approved continuity process, identify pending transactions and reconcile temporary records before resuming normal processing.

Recovery test results, outage log and a signed reconciliation of transactions processed during interruption.

6.1 Controls specifically required for GenAI

The assistant should operate in an approved environment with restricted access, encryption and contractual limits on the use and retention of government information. Customer documents should not be entered into an unrestricted public chatbot. Officers should see the source supporting each suggestion and be able to reject or correct it. Where the source is missing or contradictory, the assistant should refer the matter to a person instead of producing a definite answer.

ICT and the business owners should test the assistant against incomplete documents, misleading instructions embedded in files and plausible but incorrect answers. Uploaded content must be treated as evidence, never as instructions that override the system’s rules. Logs should record the model version, source material, generated suggestion and officer action without unnecessarily reproducing personal information. Policy changes require controlled updates and testing. These measures address the risk that staff may accept a fluent answer without checking whether it is correct.

DMV management should own eligibility decisions, Treasury should own payment and receipting controls and ICT should own technical access and recovery. No single officer should be able to alter a charge, confirm payment and approve a refund without independent review.

6.2 GenAI safeguards and accountability

Exposure introduced by GenAI

Specific safeguard

Owner and evidence

An incorrect answer or invented requirement

Retrieve current approved material; show sources; block automatic approval; refer unclear cases to officers.

DMV process owner reviews test cases and sampled outputs; correction log records errors.

Disclosure of personal or financial information

Use a restricted environment, minimise input fields, control retention and prohibit supplier training on TCIG information unless expressly authorised.

ICT and information governance owners review access logs, contracts and retention settings.

Instructions hidden in uploaded documents

Treat file contents as untrusted data; restrict available tools and prevent the model from executing document instructions or making financial changes.

ICT tests hostile documents and retains results before release or model changes.

Staff overreliance on fluent output

Train officers to verify sources; retain the right to reject suggestions; sample approved cases independently.

Supervisors review acceptance patterns, overrides and errors during the pilot.

These safeguards respond to the possibility that an assistant can produce an answer that sounds reasonable but is unsupported. NIST (2024) discusses this problem as confabulation and also identifies privacy and information security risks. The controls above are tailored recommendations for TCIG; they should be assessed against the Government’s own policies before implementation. No claim is made that they are currently in place.

6.3 Preventive, detective and corrective controls

Mandatory fields, restricted permissions and duplicate-payment checks are preventive controls because they aim to stop an error before it affects the official record. Reconciliation, exception reports and supervisory sampling are detective controls because they identify problems that have occurred. Authorised corrections, recovery procedures and investigated refunds are corrective controls. All three are needed. A system with extensive preventive checks still requires reconciliation because external payment delays, integration faults and human mistakes cannot be ruled out.

Each exception should have a named owner, a date raised, a reason and a target resolution date. Management should approve escalation periods for the pilot, including when a supervisor must intervene. A customer’s repeated query should be linked to the existing case instead of treated as a new transaction. Any refund or cancellation must follow a separate authorised workflow with traceable approval. It must not be generated automatically because GenAI describes a payment as duplicated.

6.4 Continuity, access and public accountability

If only the assistant is unavailable, staff should continue using the approved checklist and normal systems. If payment verification or receipting is unavailable, staff must follow an authorised continuity procedure and must not issue an unverified receipt. Temporary records need unique references, secure custody and reconciliation after recovery. The online route should operate alongside counter assistance so that customers with limited digital access are not unfairly excluded. Customers should also have a clear route to a human officer when the assistant cannot resolve an issue.

7. Revised process flow

Figure 2A. Proposed process with GenAI assistance, officer approval and controlled receipting.

The payment decision represents confirmation under the rules for the selected channel. An acknowledgement that a transfer was initiated is insufficient. Integration failures enter the controlled exception process; they must not trigger a second payment or receipt automatically.

7.1 Receipt linking and exception controls

Figure 2B. Supporting controls for linking the official receipt to the application in Figure 2A. GenAI may summarise an exception for staff, but the financial checks shown here are rules-based controls.

All posting attempts and corrections are logged. If a system is unavailable, the continuity procedure in Section 6.4 applies. Return routes preserve the original reference and require a check for an existing receipt before any retry.

8. Implementation and evaluation

8.1 Phased implementation and release decisions

TCIG should begin by mapping the actual renewal workflow with DMV officers, cashiers, finance staff and ICT. This would establish where information is entered more than once, which documents are necessary and whether the existing systems can exchange approved transaction data. The review should also establish the baseline for processing time and error rates. Without this evidence, a claim that the proposal will reduce costs or waiting time would be premature.

The first pilot should involve a limited set of routine renewals and a small trained team. Initially, GenAI could prepare draft information while officers continue the existing verification process. Once the quality of those drafts is acceptable, the team could test payment and receipt integration in a controlled environment. Failed, repeated and delayed payment confirmations should be tested alongside successful transactions. Expansion should depend on demonstrated accuracy and reliable reconciliation, rather than customer volume alone.

Staff participation is essential. Officers should help design the prompts, identify common customer misunderstandings and review exceptions. Training should cover how to challenge an AI suggestion, protect personal information and follow the continuity procedure. Time released from routine entry could be used for reconciliation, complex applications and direct support to customers who need assistance.

Measure

How leadership should assess it

Processing time

Compare median staff handling time and total completion time with the baseline, including delayed cases.

Accuracy and rework

Track incorrect application fields, officer corrections and repeat submissions per completed renewal.

Payment integrity

Track unmatched payments, duplicate receipts and the time taken to resolve exceptions.

GenAI quality

Measure how often officers accept, correct or reject suggestions and investigate serious errors.

Service access

Monitor customer feedback and outcomes for both online and staff-assisted applications.

A business case should include software and integration costs, security review, staff training, maintenance and ongoing monitoring. No percentage reduction in time or cost is claimed in this report because these figures require baseline and pilot evidence. If most delay is caused by inspection capacity or system outages, GenAI alone will have limited effect. Leadership should therefore assess the complete process rather than judge success by the number of AI-generated responses.

8.2 Pilot acceptance criteria and test cases

The following criteria are proposed management conditions for a pilot, not measured results or existing TCIG targets. They should be agreed before testing so that leadership does not assess success only after seeing the results. The pilot must test unusual and failed cases as well as straightforward renewals.

Test case

Expected result

Complete and eligible renewal with verified payment

Officer-approved information produces one receipt linked to one application and payment.

Missing document or conflicting vehicle identifier

The assistant flags the issue; the officer decides the required correction; the case cannot bypass approval.

Payment screenshot without authorised confirmation

The transaction remains pending; no official receipt or completed renewal is generated on that evidence alone.

Repeated payment confirmation or retry

The existing transaction is located; duplicate processing is blocked; any true second payment is referred for review.

Receipt posting fails after payment succeeds

Payment remains recorded; a controlled retry is made; the customer is not instructed to pay again.

Outdated guidance or unsupported GenAI answer

Source and effective date are checked; the answer is corrected or escalated before use.

Assistant outage or an instruction hidden in a document

The approved manual route remains available; hidden instructions do not change permissions or transaction rules.

Release conditions should include no unverified payments triggering receipts in the test set, no duplicate official receipts from replayed confirmations and full traceability of every approval and posting. These are control requirements, not a promise of perfect performance after release. For efficiency, management could set an initial pilot objective of reducing median staff handling time by 20% without increasing the correction rate. This is an illustrative proposed target; it must be revised if the baseline shows that it is inappropriate. All unresolved financial differences should be reviewed before the pilot expands.

8.3 Measuring value and recognising limitations

Median handling time should be calculated from staff start and finish records for comparable renewals, with exception cases reported separately. The correction rate is the number of completed cases requiring correction divided by total completed cases. The unmatched-payment rate is the number of confirmed collections not linked to a valid application and receipt at the reporting cut-off divided by confirmed collections. Reporting the denominator prevents a larger transaction volume from being mistaken for poorer performance.

A financial assessment could estimate annual staff hours released as transaction volume multiplied by average minutes saved, divided by 60. The value of that capacity should be considered alongside software, integration, security, training and maintenance costs. Released time is not automatically a cash saving, particularly where staff remain employed and use the time for other duties. Benefits such as fewer return visits and better reconciliation should therefore be reported separately from any verified reduction in expenditure.

The main constraints are access to reliable source information, the ability of existing systems to exchange data and the quality of staff review. A well-written AI explanation cannot correct an inaccurate official record or compensate for insufficient inspection capacity. Leadership should approve wider use only when the evidence shows that the complete transaction has improved.

9. Conclusion

Vehicle licence renewal offers TCIG an opportunity to strengthen a public service transaction by improving how information is prepared, reviewed and explained. ePay already produces an official receipt, and that capability should be retained. The strongest case for GenAI is therefore an assistant that helps customers understand requirements and helps officers prepare application summaries and resolve incomplete submissions. Supporting integration should connect approved records where a verified gap exists.

A successful approach would retain officer judgement while reducing repetitive entry and unnecessary follow-up. Approved fee rules, verified payment information, clear responsibilities and timely reconciliation must remain central to the process. With a limited pilot and measurable outcomes, TCIG can determine whether the proposal improves service and staff capacity without weakening financial accountability.

References

National Institute of Standards and Technology. (n.d.). Generative artificial intelligence. Computer Security Resource Center glossary. https://csrc.nist.gov/glossary/term/generative_artificial_intelligence

National Institute of Standards and Technology. (2024). Artificial intelligence risk management framework: Generative artificial intelligence profile (NIST AI 600-1). https://doi.org/10.6028/NIST.AI.600-1

Turks and Caicos Islands Government, Department of Motor Vehicles. (n.d.). Services. https://www.gov.tc/dmv/services

Turks and Caicos Islands Government, Department of Motor Vehicles. (n.d.). Department of Motor Vehicles. https://www.gov.tc/dmv/

Key abbreviations

AI: artificial intelligence. GenAI: generative artificial intelligence. DMV: Department of Motor Vehicles. TCIG: Turks and Caicos Islands Government. MIS: management information systems. LLM: large language model. RAG: retrieval-augmented generation. API: application programming interface.

Appendix A. Rubric coverage and evidence to confirm

Assessment requirement

Marks

Where addressed

General introduction of the domain

5

Section 1: purpose, GenAI definition, public service domain and transaction scope.

Current process and cycle impact

5

Section 2: roles, information flow and revenue-cycle consequences.

Current process flow

30

Section 3 and its payment-channel detail: decisions, exceptions and reconciliation.

Four process threats

12

Section 4: four distinct threats, their causes and consequences.

GenAI improvement and explanation

20

Section 5: staff use cases, system design, worked example and alternative options.

Controls for threats and GenAI

15

Section 6: matched controls, owners, evidence and GenAI safeguards.

Revised process flow

40

Section 7 and its receipt-control detail: AI assistance, approvals and financial controls.

Conclusion

5

Section 9: leadership recommendation and limits.

Creativity and presentation

8

Consistent figures, comparison tables and a practical integrated solution.

Total

140

Coverage identifies where marks are addressed; it does not guarantee a grade.

Operational evidence needed before final submission

ePay’s production of an official receipt is confirmed by the operational information supplied. Confirm the remaining sequence of DMV review, fee assessment, payment and renewal issuance. Identify whether the current systems already link the official receipt to the application, which payment channels are available specifically for vehicle renewals and who performs reconciliation. Establish the actual volume of incomplete applications, repeated queries and staff handling time. Check the applicable renewal checklist and physical inspection requirements with the responsible officers.

Where permitted, support the final report with a short anonymised process walkthrough, an approved checklist and aggregate measures of handling time, correction volumes and unmatched payments. Do not use identifiable customer records in the assignment. Revise the current-state diagram where officer confirmation differs from the working model. These checks strengthen the evidence supporting the proposal without inventing weaknesses or performance figures.

image1.png

image2.png

image3.png

image4.png