AI Transparency

Last reviewed: 24 September 2026

This page sets out where ScheduleLens uses AI, what the AI receives, what it cannot do, and how to switch it off. It is written for firms that must assess the tools they use, including RICS members and regulated firms under the RICS professional standard Responsible use of artificial intelligence in surveying practice (1st edition, effective 9 March 2026). The sections below answer, in writing, the questions that standard asks firms to put to a supplier. ScheduleLens is not affiliated with, endorsed by or certified by RICS, and this page is not a statement that any firm's use of ScheduleLens complies with the standard; that assessment is yours.

In short

The analysis uses no machine learning and no language model. Every finding, figure, score and chart is computed by rule-based analysis code, and the same file always produces the same figures. The RICS standard adopts the OECD definition of an AI system, which is broad enough to include rule-based planning algorithms, so whether our analysis engine counts is your firm's call. If you treat it as one, the answers on this page cover it too: it was built by ScheduleLens and has no training data.
AI is used in two optional places: writing the explanatory prose in reports, and the subscriber AI assistant that answers questions about a completed analysis.
The AI works only from our computed results. It never receives your schedule file.
Every report says whether AI was used and which sections it wrote, in its Methodology section. Where prose and a table or chart differ, the table or chart governs.
You can switch the report AI off for one upload, or for all uploads under Account > Report preferences. When it is off, no AI system receives any data from that analysis. The AI assistant only runs when you ask it a question.

The AI system

Type: A general-purpose large language model, used only to write prose.
Model: Qwen 3.6 27B (model id qwen3-6-27b), an open-weight model published by the Qwen team at Alibaba Cloud. Its extended reasoning mode is switched off for this use.
Run by: Venice.ai, through its API. Venice lists this model as "private": prompts and responses are processed for inference only and are not retained after the request completes.
What it receives: A structured summary of results our engine has already computed: headline figures, the tables shown in the report (which include activity IDs and names), the findings and the comparison summary, plus short excerpts of the published standards those findings cite.
What it never receives: Your schedule file, the full activity list or the logic network. A test in our codebase lists every field the AI may be sent and fails if a new one is added without review.
Trained on your data?: No. ScheduleLens trains no models, and Venice's model providers are prohibited from using prompts or outputs beyond returning the response.

How it works, and its limits

The analysis engine runs first and computes every figure. The model is then given those results and asked to write the explanatory paragraphs for particular sections of the report. It does no analysis of its own and cannot change a figure in a table or chart.

Known limits of any language model apply here. It can restate a figure inaccurately, for example describing a recovery as a slip, or describe an event more strongly than the figures support. For that reason:

each report's Methodology section names the model and the sections whose prose it wrote, so a reviewer knows exactly which text to check;
the tables and charts are authoritative, and the report says so;
prose that uses a phrase we have retired or banned (for example a figure we no longer report) is discarded automatically and replaced with standard template wording;
prose that quotes a figure absent from the computed results it was given is discarded automatically and replaced with standard template wording, and each case is counted for our review;
where a figure's direction is easy to misread (a negative "slip" is a recovery), the report states the direction in words, which is what the model reads.

Reports, including any AI-written prose, are analytical output, not expert opinion or legal advice, and have not been reviewed by a qualified delay analyst.

Supplier due diligence: written answers

The six questions below are those the RICS standard (section 4.1) asks firms to put to a supplier in writing. Where our answer is limited, we say why, so the gap can be recorded in your risk register.

1. Environmental impact of the AI system

We cannot give a measured energy or carbon figure: neither Venice nor the model's publisher publishes one for this model. What we can state: the AI is used only to write short paragraphs, and a typical health or comparison report makes five model calls, reading about 20,000 tokens and writing about 600 (measured on our sample reports, September 2026). The analysis itself uses no AI. Switching the AI off removes this use entirely.

2. Stakeholders involved in developing the AI system

The language model was developed and published by the Qwen team at Alibaba Cloud and is run by Venice.ai. ScheduleLens did not develop or train it. ScheduleLens's analysis engine and its use of the model were designed and built by Robert Kay, trading as ScheduleLens (ABN 41 580 709 632).

3. Compliance with data and confidentiality law

Our handling of personal and project data is set out in our Privacy Policy. In summary: an uploaded schedule file is deleted once it has been processed, unless you add it to a tracked Project, in which case an encrypted copy is kept until you delete it. Only computed results are sent to the model, and nothing is sent when the AI is switched off. If you use the AI assistant, an encrypted copy of the computed results (not the file) is kept for about three days so it can answer follow-up questions. Venice's commitment not to retain prompts or responses is contractual; it cannot be independently verified. Venice does keep operational metadata (for our calls, our server's address and our account, not yours). The application and its database run on Fly.io in Virginia, USA; stored reports and retained files are held in Tigris object storage, which chooses the region automatically. Our Terms of Service are governed by the law of Western Australia.

4. Permissions where personal data was used

ScheduleLens used no personal data to develop anything, because it develops no models. What personal data the model's publisher used in training is set out, to the extent published, in the Qwen model documentation; we have no further information. At run time the model may see activity names from your schedule, so if your activity names contain people's names, those will be included.

5. Accuracy, relevance and diversity of the training data, and known bias

We did not train the model and have no information about its training data beyond what its publisher documents. This is a limitation to record. We manage the practical risk at the output: the model writes prose only from our computed results, the tables govern, and each report names the sections it wrote. In our most recent test (September 2026, 48 report prompts), none of the model's paragraphs stated a figure absent from the results it was given.

6. Type and extent of our liability

Our liability is set by our Terms of Service. Our total liability for any claim is capped at the fees you paid in the 12 months before the claim. We are not liable for indirect or consequential loss, or for decisions or actions taken on the basis of report output. Reports are analytical output, not expert opinion, expert determination or legal advice, and have not been reviewed by a qualified delay analyst. Obtain independent professional advice before relying on them for contractual decisions, disputes or formal submissions. This applies equally to the AI-written prose.

For your AI register

If you record ScheduleLens in a register of AI systems (RICS standard, section 3.2), this entry can be copied and completed:

AI system: ScheduleLens AI narrative and AI assistant (language model Qwen 3.6 27B, model id qwen3-6-27b, run by Venice.ai)
Supplier: Robert Kay trading as ScheduleLens (ABN 41 580 709 632)
Purpose: drafting explanatory prose in schedule analysis reports, and answering questions about a completed analysis, from figures computed by ScheduleLens's rule-based engine
Material impact on our service: [your determination and reasoning]
Date first used: [date]
Next review: [date]
Can be switched off: yes, per upload or as the account default (Account > Report preferences)

Telling your client

A sentence you can adapt for your terms of engagement or report covering letter (RICS standard, section 4.3):

We use ScheduleLens to analyse programme files. Its findings and figures are computed by a rule-based analysis engine. Where we enable it, a language model (Qwen 3.6 27B, run by Venice.ai) drafts explanatory prose from those computed results, and each report states which sections it wrote. We review that prose before relying on it, and you may ask us not to use the AI feature on your project.

Questions

What the standard asks of a firm that reviews programmes is covered in The RICS AI standard and programme review. For anything this page does not answer, contact us. We review this page whenever the model, provider or data flow changes.