Procurement accelerator pilot - ai-enabled rfq/rfb automation for rand water

Closing date: 20 October 2026 (from tender record; time not stated in annexures — refer to RFQ cover page RW SCM 00016 F).

Indicative programme (assuming appointment by early November 2026):

• Phase 0 — Mobilisation: Week 1 (Deliverable D1).
• Phase 1 — Inception and discovery: Weeks 1–3 (Deliverables D2, D3).
• Phase 2 — Build and configuration: Weeks 2–6 (Deliverables D4, D5).
• Phase 3 — RFQ pilot run: Weeks 5–7, with interim results by December 2026 (Deliverables D6, D7).
• Phase 4 — RFB pilot run (parallel): Weeks 6–22 (Deliverable D8).
• Phase 5 — Evaluation, roadmap and close-out: Weeks 20–26 (Deliverables D9, D10, D11).

Total pilot period: six months.

Key milestone: RFQ interim results must be submitted by December 2026.

Note: Rand Water staff availability is reduced over the year-end period; bidder must plan around this constraint.

Enquiries

• Method: written questions to the buyer's email address on the RFQ cover page (RW SCM 00016 F).
• Process: Rand Water will share answers with all prospective bidders.

No named SCM or technical contacts, telephone numbers, or physical addresses are provided in the annexures. Refer to the RFQ cover page for these details.

Governance contacts (for information only — not for bid enquiries):

• Business owner: Chief Procurement Officer (CPO) — owns pilot outcome, approves deliverables.
• Innovation and New Technologies (I&NT) Division — project coordination, scope management.
• GCFO / Finance PMO — Finance project governance and controls.
• CTIO / DTIM — architecture, security, Azure and SAP environments.
• Supply Chain Management — process owners, pilot users, appointed buyer, user acceptance testing.

Returnable documents required:

• Schedule B (Data Use, Confidentiality and AI Model Training Undertaking): signed undertaking covering data use restrictions, AI model training prohibitions, sub-processor disclosure, POPIA compliance, and data deletion obligations. Mandatory for pre-qualification.
• Schedule C (Security and AI Disclosure Questionnaire): completed questionnaire covering solution architecture, data residency, AI models, security controls, SAP integration, and deletion process.
• Schedule D (Detailed Programme): showing how RFQ pilot results will be delivered by December 2026 and RFB use case completion within the six-month pilot period.

Submission method: not stated in the annexures provided; refer to the RFQ cover page (RW SCM 00016 F) for the buyer's email address and submission instructions.

Disqualification risks:

• Failure to sign and submit Schedule B.
• Incomplete or unsigned returnable schedules.
• Late submission after the closing time.

Schedule C — Security and AI Disclosure Questionnaire: bidder must answer all 12 questions with supporting documents where relevant. Rand Water may verify answers during demonstration.

Questions cover:

• C1: Solution architecture, components inside and outside Rand Water's Azure environment.
• C2: Data storage and processing location (Azure region/country); all sub-processors.
• C3: Every AI model used — provider, model, version, hosting location, purpose.
• C4: Whether any model or embedded AI service retains, logs or learns from Rand Water prompts, documents or outputs; how to switch this off.
• C5: Model configuration/grounding for Rand Water; who can change configuration.
• C6: AI output logging, explanation, auditing; accuracy measurement and reporting.
• C7: AI-agent actions requiring human approval versus those that do not.
• C8: Identity and access management — Entra ID single sign-on, multi-factor authentication, role-based access control.
• C9: Encryption in transit and at rest; key management.
• C10: SAP integration method (interfaces, data objects, frequency); S/4HANA compatibility approach.
• C11: Security certifications held and date of most recent independent security test.
• C12: Data return and deletion process at close-out.

Evaluation approach: the document does not specify a points-based scoring split (80/20 or 90/10) or a minimum qualifying score in the annexures provided. The RFQ form (RW SCM 00016 F) should contain the preference point system and price/functionality weighting.

Pre-qualification requirement (RFQ 3.1.5): signed Schedule B undertaking.

Proposal requirements: bidders must state measurable targets for each success measure and explain how they will be measured. Baselines and targets are confirmed at inception (Deliverable D2).

Success measures (bidder must commit to targets for each):

• Procurement cycle time: elapsed time per RFQ/RFB stage from approved request to award or purchase-order handover, versus baseline.
• Reduction in manual processing and rework: number of manual touchpoints, hand-offs and clarification loops per RFQ, versus baseline.
• Conversion rate (speed and accuracy): share of requests progressing to award within target time without return for correction.
• Compliance and audit-trail quality: complete audit trail share; control exceptions detected; ability to respond to recurring AGSA findings.
• Bid-splitting detection: potential splitting cases and trends flagged, validated against historical data review.
• Data accuracy and AI performance: accuracy of contract and document metadata extraction against verified sample; SAP reconciliation match rate.
• Real-time reporting and payment tracking: dashboard data latency; purchase-order and payment status availability; usefulness rating by SCM and Finance users.
• User adoption and satisfaction: active use by pilot user group; user satisfaction survey results.
• Security and interoperability: security test results; zero-trust compliance; SAP interfaces working; open APIs.
• Viability of scaling: architecture roadmap quality; S/4HANA compatibility; full rollout life-cycle cost.
• Pilot delivery (earned value): schedule and cost performance using earned value management.

Evaluation decisions on functionality, price and specific goals remain with Rand Water officials; AI outputs are decision support only.

Procurement: pilot implementation of an AI-enabled, agentic-ready Procurement Accelerator to streamline and automate Rand Water's sourcing value chain.

Primary use case — Request for Quotation (RFQ):

• End-to-end RFQ process support from approved requisition to award and purchase-order handover.
• Need capture and requisition validation: budget availability, account assignment, approval routing per Delegation of Authority (DoA).
• RFQ document preparation from Rand Water templates, with AI-assisted drafting and checking of specifications, evaluation criteria and pricing schedules.
• Supplier selection and compliance checking: Central Supplier Database status, tax compliance, National Treasury Register for Tender Defaulters and List of Restricted Suppliers.
• RFQ issuance, clarification and addenda management, quotation receipt.
• Evaluation support: responsiveness, functionality, price and specific goals, with scoring worksheets and full audit trail. Decisions remain with Rand Water officials.
• Recommendation, approval and award routing per DoA, with handover to SAP for purchase-order creation.
• Cycle time and status tracking at every step, with automated reminders and escalation of delays.

Secondary use case — Request for Bid / tender (RFB):

• Configured and tested on one parallel RFB/tender process.
• Same control and tracking principles extended to bid committee processes.
• Full outcome may extend into 2027; bidder must propose which RFB stages can be demonstrated within the pilot period.

Supplier management:

• Consolidated supplier view: registration, compliance, participation and award history, activity tracking.
• Flagging of suspended or restricted suppliers before invitation or award.

Finance controls and visibility (Section 6.4 and Annexure 2):

• Budget and approval validation against available budget, account assignment and DoA.
• Real-time purchase-order and supplier-payment status visibility.
• Exception dashboards: aged orders, approval delays, procurement deviations.
• SAP reconciliations supporting completeness, accuracy and reconcilability.
• Invoice-control analytics (Annexure 2 classifies as Mandatory, Desirable or Future): duplicate detection, three-way-match exceptions, missing documents, receipting validation.

Reporting and dashboards:

• Real-time and periodic dashboards for SCM, Finance and executive users.
• Payment tracking, drill-down to individual transactions, export in standard formats.

Usability and support:

• Intuitive, role-based interface for end users, buyers, approvers and managers.
• Standard browser access, minimal training required.
• User support during pilot with defined response times.

Environment and integration:

• Deployment in Rand Water's Microsoft Azure non-production environment.
• Zero-trust security: Microsoft Entra ID single sign-on, multi-factor authentication, role-based access control with least privilege, encryption in transit and at rest.
• SAP integration through standard, supported interfaces without modifying SAP core; compatible with future SAP S/4HANA.
• Open, documented APIs and open-format data export for future integration and exit.
• Data residency: South African Azure regions preferred; any processing outside South Africa must be disclosed in Schedule C and approved by Rand Water.

AI governance:

• Disclosure of all AI models: provider, model, version, hosting location, purpose, configuration, and who can change it (Schedule C).
• Prohibition on using Rand Water data to train, fine-tune or improve models outside the pilot environment without written consent.
• Traceability of AI outputs: inputs, outputs, confidence levels, user actions.
• Human in the loop: AI agent actions with financial, contractual or compliance consequences require authorised user approval.
• AI performance measurement: accuracy of contract and document metadata extraction against verified sample.
• Controls for hallucination, bias and outdated information in AI-generated content.

Boundaries and exclusions:

• No changes to production SAP environment; non-production (development/QA) environments only.
• Does not replace or duplicate planned e-procurement / procure-to-pay solution.
• Transactional invoice processing and payment remain in SAP.
• Full-scale rollout, production cut-over and organisation-wide licensing excluded; subject to separate procurement if pilot succeeds.

Rand Water inputs:

• Azure environment (subscription/resource group) provisioned by DTIM.
• Non-production SAP environment access and interface documentation.
• Approved SCM business process maps, templates, policies (DoA, SCM policy), historical procurement data subject to data-sharing approval.
• Nominated SMEs from SCM, Finance, DTIM and end-user areas; pilot user group.

Known constraints:

• Development and QA environment readiness is a material schedule risk.
• DTIM capacity constrained by recovery and normalisation work.
• Reduced Rand Water staff availability over year-end.
• Bidder must minimise dependence on Rand Water resources.

Bidder methodology must address:

SAP integration:

• Describe interfaces used, data objects exchanged, and data reconciliation approach.
• Standard, supported interfaces without SAP core modification.
• Future SAP S/4HANA compatibility required.

Architecture roadmap:

• Show how solution scales and embeds in current and future-state architecture.
• Identify gaps and configuration/design decisions for full rollout.

AI transparency and governance:

• Disclose all AI models: provider, model, version, hosting location, purpose, configuration, and who can change it (Schedule C).
• Prohibit use of Rand Water data for training, fine-tuning or improving models outside pilot environment without written consent.
• Disclose whether embedded third-party AI services retain or learn from prompts/outputs.
• Ensure AI output traceability: inputs, outputs, confidence levels, user actions.
• Human in the loop: AI agent actions with financial, contractual or compliance consequences require authorised user approval.
• Measure and report AI performance: accuracy of contract and document metadata extraction against Rand Water-verified sample.
• Describe controls for hallucination, bias and outdated information in AI-generated content.

Functional requirements summary (full detail in Annexure 2, classified Mandatory/Desirable/Future):

• Process automation and tracking: configurable RFQ/RFB workflows, end-to-end status tracking, automated reminders/escalations, time stamps for cycle-time measurement, document generation from templates, electronic audit trail.
• Compliance and control: automated checks against SCM policy thresholds, DoA, supplier compliance and restricted-supplier lists; bid-splitting detection and trend reporting; procurement deviation flagging; support for addressing recurring AGSA findings.
• Supplier management: consolidated view of registration, compliance, participation/award history, activity tracking, suspended/restricted supplier flagging.
• Finance controls and reporting: budget and approval validation; real-time purchase-order and payment status; exception dashboards for aged orders, approval delays, deviations; SAP reconciliations.
• Reporting and dashboards: real-time and periodic dashboards for SCM, Finance and executives; payment tracking; drill-down; standard-format export.
• Usability and support: intuitive role-based interface; standard browser access; minimal training; defined-response user support.

Methodology must support both shadow mode (live RFQs running in parallel with current process) and direct processing for selected RFQs, per Rand Water's inception decision.

No explicit experience or qualification thresholds (years of experience, reference projects, turnover) are stated in the annexures provided. The bidder's capability is demonstrated through:

• Proposed solution architecture and integration approach.
• AI model disclosure and governance controls.
• Security certifications (e.g. ISO/IEC 27001, SOC 2) and latest independent security test date.
• Detailed programme showing realistic delivery by December 2026 for RFQ results and six-month completion for RFB use case.
• Commitment to key personnel named in Schedule B, with no replacement without Rand Water approval and replacement at equal or better qualification/experience.

Refer to the RFQ form (RW SCM 00016 F) for any minimum experience, turnover or personnel qualification requirements.

Hosting and environment:

• Pilot deployed in Rand Water's Microsoft Azure environment, separate from production.
• Any component hosted outside this environment (including AI model endpoints) must be disclosed, justified, and approved in writing by Rand Water.

Data residency:

• Rand Water data should be stored and processed in South African Azure regions.
• Any processing outside South Africa must be disclosed in Schedule C and approved by Rand Water.

Security:

• Zero-trust principles: Microsoft Entra ID single sign-on, multi-factor authentication, role-based access control with least privilege, encryption and key management for data in transit and at rest.

SAP integration:

• Standard, supported interfaces without modifying SAP core.
• Compatible with future SAP S/4HANA direction.

Interoperability:

• Open, documented APIs and open-format data export for future integration and exit without lock-in.

Pricing is deliverable-based against 12 pricing items linked to 11 deliverables (D1–D11). Licence fees are a separate recurring cost.

Pricing structure:

• Item 1: D1 — Project plan, risk and issues register, fortnightly progress reporting.
• Item 2: D2 — Inception and baseline report; D3 — Solution design.
• Item 3: Licence fees — monthly in arrears from D4 commissioning date.
• Item 4: D4 — Environment commissioning report.
• Item 5: D5 — SAP integration and reconciliation test report; security test report.
• Item 6: D6 — Configured RFQ workflow, controls and dashboards; D7 — RFQ interim results report.
• Item 7: D8 — RFB pilot configuration and results report.
• Item 8: D6 — Linked to RFQ workflow deliverable.
• Items 9, 10: D9 — Control narratives, process documentation, training material and records.
• Item 11: D10 — Integration and architecture roadmap; scale-up business case including life-cycle costing.
• Item 12: D11 — Final evaluation report, presentation, close-out report, data return and deletion certificate.

Payment terms:

• No advance payments.
• Payment against deliverables accepted in writing by the Rand Water project manager.
• Licence fees monthly in arrears from environment commissioning (D4).

Bidder must include life-cycle costing in the scale-up business case (D10).

Payment terms:

• No advance payments.
• Payment made against deliverables accepted in writing by the Rand Water project manager.
• Licence fees (pricing item 3): payable monthly in arrears from environment commissioning date (D4).

Pricing items linked to deliverables:

• Item 1: D1 (Project plan, risk and issues register, progress reporting).
• Item 2: D2 (Inception and baseline report), D3 (Solution design).
• Item 4: D4 (Environment commissioning report).
• Item 5: D5 (SAP integration and reconciliation test report; security test report).
• Item 6: D6 (Configured RFQ workflow, controls and dashboards), D7 (RFQ interim results report).
• Item 7: D8 (RFB pilot configuration and results report).
• Item 8: D6 (linked to RFQ workflow deliverable).
• Items 9, 10: D9 (Control narratives, process documentation, training material and records).
• Item 11: D10 (Integration and architecture roadmap; scale-up business case).
• Item 12: D11 (Final evaluation report, presentation, close-out report, data return and deletion certificate).

Financial requirements beyond the pricing schedule are not detailed in the annexures. Refer to the RFQ form (RW SCM 00016 F) for bid security, guarantees, insurance or turnover requirements.

Pilot cost commitments:

• Bidder must produce a costed architecture roadmap and scale-up business case including life-cycle costing (Deliverable D10).
• Earned value management required for pilot delivery reporting.

Pre-qualification:

• Signed Schedule B (Data Use, Confidentiality and AI Model Training Undertaking) — mandatory under RFQ 3.1.5.

Security and certifications:

• Compliance with Rand Water's ICT and information security policies.
• Security testing completed before any live data use.
• Security incident reporting within 24 hours of detection.
• Security certifications to be listed (e.g. ISO/IEC 27001, SOC 2) with date of most recent independent security test.

Data protection:

• Compliance with Protection of Personal Information Act (POPIA).
• Signing of Rand Water's non-disclosure agreement and operator agreement (where processing personal information on Rand Water's behalf).

AI and data governance:

• Rand Water data must not be used to train, fine-tune or improve any AI model outside the pilot environment without prior written consent.
• All sub-processors must be disclosed in Schedule C.
• No new sub-processors or processing outside Rand Water's Azure environment or outside South Africa without written approval.
• Data return and deletion within 30 days of pilot close-out in open, usable format; signed deletion certificate required.

Key personnel:

• Named key personnel in Schedule B may not be replaced without Rand Water's prior written approval.
• Any replacement must be of equal or better qualification and experience.

CSD, tax clearance, B-BBEE, CIDB, CIPC and local content requirements are not stated in the annexures provided. Refer to the RFQ form (RW SCM 00016 F) for these requirements.

No occupational health and safety requirements are stated in the annexures provided.

Special Conditions of Contract (take precedence over General Terms and Conditions where conflicting):

• Pilot only: no commitment to further procurement, licence or rollout; licences do not renew automatically.
• Ownership: all Rand Water data including derived data remains Rand Water property. Rand Water owns reports, roadmaps, control narratives, configurations and documentation produced specifically for it. Bidder retains pre-existing intellectual property and grants Rand Water use rights without further charge.
• Confidentiality: mandatory signing of Rand Water's non-disclosure agreement; confidentiality obligation survives the pilot.
• POPIA compliance: mandatory; operator agreement required for personal information processing.
• AI model training prohibition: Rand Water data must not be used to train, fine-tune or improve any AI model outside the pilot environment, or be disclosed to third parties, without prior written consent. All sub-processors must be disclosed.
• Data return and deletion: within 30 days of close-out, all data returned in open usable format and permanently deleted from all systems; signed deletion certificate required.
• Information security: compliance with Rand Water's ICT and information security policies; security testing before live data use; 24-hour incident reporting; security certifications and latest independent test date to be listed.
• Key personnel: no replacement without prior written approval; replacements must be equal or better qualified.
• Exit and transition: orderly exit support including configuration export, data export and documentation handover to prevent lock-in.

Schedule B (Data Use, Confidentiality and AI Model Training Undertaking) is a pre-qualification requirement (RFQ 3.1.5).

See contractual_terms section for full special conditions.

Evaluation support covers responsiveness, functionality, price and specific goals, with scoring worksheets and full audit trail. Final evaluation decisions remain with Rand Water officials.

RFB use case: bidder must propose which RFB stages can realistically be demonstrated within the pilot period; complete outcome may extend into 2027.

No production SAP changes permitted; integration uses non-production (development/QA) SAP environments and standard interfaces.