Skerply — Recommendations Disclaimer

Effective: 21 May 2026. Last updated: 21 May 2026. Version 1.0.

This Recommendations Disclaimer governs the cost-saving recommendations produced by the Skerply Azure Managed Application (the "Service"), operated by GreyStack (Pty) Ltd (registration number 2026/016718/07) ("we", "us", "Skerply"). It is incorporated by reference into the Terms of Service. Capitalised terms not defined here have the meanings given in the Terms of Service.

A short-form version of this disclaimer is rendered in the dashboard alongside every recommendation.

1. Short-form disclaimer (rendered in the dashboard)

Advisory only. This recommendation is generated by software, may use AI, and has not been reviewed by a human at Skerply. You are responsible for deciding whether to act on it. Verify the impact on your workload, security posture, contracts and compliance before making any change. Skerply does not make any change to your Azure resources. See the full Recommendations Disclaimer for the long-form terms.

2. Nature of recommendations

2.1 Each recommendation produced by the Service is a suggestion for the Customer's consideration. It is not an instruction, not a representation that the proposed change is appropriate, and not an undertaking that the proposed change will reduce the Customer's Azure spend.

2.2 Recommendations are generated by a combination of (a) deterministic analysers, (b) data sourced from Azure Cost Management, Azure Resource Graph, Azure Monitor, Azure Advisor and the public Azure Retail Prices API, and (c) reasoning generated by large language models hosted in Azure OpenAI inside the Customer's tenant (ADR-006).

2.3 Although we apply schema validation, pricing-source enforcement, prompt-injection sanitisation and a multi-agent debate process designed to reduce error (ADR-014), AI-generated content can be incorrect, incomplete, out of date or otherwise unsuitable for the Customer's circumstances. We do not warrant the accuracy, completeness or suitability of any recommendation.

3. The Customer is solely responsible for action

3.1 The Service is read-only with respect to the Customer's Azure resources and will not make any change to any Customer resource (ADR-004).

3.2 The Customer alone decides whether to act on a recommendation. Before acting, the Customer should evaluate the recommendation against the Customer's:

(a) operational and performance requirements (including service-level objectives, throughput, latency and capacity headroom); (b) security and compliance posture (including data residency, network segregation, encryption, identity, audit and regulatory obligations); (c) contractual commitments to third parties whose workloads or data may be affected; (d) disaster-recovery and business-continuity requirements; (e) financial commitments (including reservations, savings plans, sponsored credits or other prepaid commitments which may alter the actual saving); and (f) change-management process, including testing, peer review, backout plans and stakeholder communication.

3.3 The decision to make any change, the implementation of that change, and the operational outcome of that change are the act of the Customer.

4. No advice of any kind

A recommendation does not constitute, and must not be relied upon as, financial, tax, legal, regulatory, engineering, security or operational advice. The Customer should obtain advice from suitably qualified professionals where appropriate.

5. Pricing data and projected savings

5.1 Projected savings shown alongside a recommendation are estimates based on:

(a) public retail pricing returned by the Azure Retail Prices API at the time the recommendation was generated; (b) the Customer's then-current resource configuration and usage pattern; and (c) assumptions documented within the recommendation.

5.2 Actual savings may differ materially because of (without limitation) Azure pricing changes, the Customer's enterprise agreement or reserved-capacity discounts, currency movements, hidden costs of the recommended SKU (for example, egress, IOPS, transactions), or changes in the Customer's workload between the time of the recommendation and the time of action.

5.3 The fact that a saving is projected does not entitle the Customer to any rebate, refund or adjustment of fees if the projected saving is not realised.

6. AI-generated content

6.1 Some or all of the reasoning text in a recommendation is generated by a large language model. AI-generated content can produce plausible-sounding statements that are incorrect ("hallucinations"). The Customer should not rely on the model's reasoning without independent verification, particularly where the statement appears to assert a fact about the Customer's specific workload, contractual position or regulatory environment.

6.2 The Service rejects any model output that fails strict typed-JSON schema validation. The Service also enforces that pricing figures used in a recommendation must come from the Azure Retail Prices API call performed for that recommendation, not from the model's training data (ADR-005). These safeguards reduce, but do not eliminate, error.

6.3 Where the Customer's jurisdiction imposes additional disclosure or transparency obligations on AI-generated output (including, where applicable, the EU AI Act, UK AI regulation and analogous laws of other jurisdictions), the Service surfaces a per-recommendation provenance record (model name, model version, pricing source, schema version) inside the dashboard so that the Customer's compliance team can satisfy those obligations.

7. No liability for outcomes of acting on a recommendation

7.1 To the maximum extent permitted by applicable law, we are not liable for any loss, damage, cost or expense suffered by the Customer or any third party arising out of, or in connection with, any action the Customer or its personnel takes (or omits to take) in reliance on a recommendation, including (without limitation):

(a) loss of data, loss of availability, loss of performance, loss of feature parity or loss of compliance status arising from the Customer's implementation of a recommendation; (b) breach by the Customer of a third-party contract caused by the implementation of a recommendation; (c) increased — rather than reduced — Azure spend resulting from acting on a recommendation; or (d) the operational outcome of any change made by the Customer's personnel or contractors.

7.2 The liability cap and carve-outs set out in clause 10 of the Terms of Service apply to all liability under this Disclaimer.

8. Snoozing, dismissing and acknowledging recommendations

8.1 The dashboard allows the Customer to mark a recommendation as Acknowledged, Actioned, Resolved, Dismissed or Snoozed.

8.2 None of these states is an admission, warranty or representation by us:

(a) Dismissed does not mean we agree the recommendation was wrong; (b) Actioned does not mean we have verified the change was implemented correctly; (c) Resolved does not mean the projected saving has been achieved.

The state of a recommendation is metadata for the Customer's own workflow.

9. Changes to this Disclaimer

We may update this Disclaimer from time to time. Material changes will be notified to Customers in the same way as material changes to the Terms of Service (clause 13 of the Terms of Service). The current version is always available at this URL and in the dashboard.

GreyStack (Pty) Ltd Registration number 2026/016718/07 Incorporated in the Republic of South Africa