LexConnectPro. Operated by Tom Street & Co. Solicitors Limited
Last updated: 19 March 2026
Effective from: 19 March 2026
This policy applies to all personal data held on the LexConnectPro platform as of the effective date, including data processed by AI features before this policy was published. The AI processing activities described in this policy reflect how AI has been used on the platform since its launch.
1. Our Commitment to Transparency
Tom Street & Co. Solicitors Limited is committed to the transparent, responsible, and ethical use of artificial intelligence within LexConnectPro. This policy explains what AI we use, how it works, what decisions it influences, what safeguards are in place, and what rights you have.
We believe AI should improve access to justice, not replace professional legal judgement. Every AI feature on LexConnectPro is designed to assist, not to make final decisions about your legal rights.
2. Scope
This policy applies to all AI-powered and automated features within LexConnectPro, the Tribunal Claim Solicitors website (tribunalclaimsolicitors.co.uk), and any associated tools or services operated by Tom Street & Co. Solicitors Limited.
This policy should be read alongside our Privacy Policy, which explains the lawful bases for processing your data and your data protection rights.
3. What AI Features We Use
The following features on our platform use artificial intelligence. For each, we explain what it does, what data it processes, what role it plays in decision-making, and what human oversight is in place.
3.1 Lexi: AI Chatbot
| What it does | Lexi is an AI-powered chatbot that answers questions about employment law, guides users through common processes, and assists with initial enquiries on the Tribunal Claim Solicitors website and within LexConnectPro. |
| Data processed | The text of your conversation with Lexi, including any personal or case details you choose to share during the chat. |
| AI providers used | Anthropic (Claude) for the website chat. OpenAI for the in-app AI assistant chat. |
| Decision-making role | Informational only. Lexi does not make decisions about your case, determine eligibility for services, or provide formal legal advice. It provides general guidance and information. |
| Human oversight | Lexi’s responses are clearly labelled as AI-generated. Users are advised to confirm any information with a solicitor. Chat transcripts are available for review by our legal team where relevant to case management. |
| Limitations | Lexi may produce incorrect, incomplete, or out-of-date information. It cannot access your case file unless this is part of a logged-in session where you have consented to this. It is not a substitute for legal advice from a qualified solicitor. |
3.2 Case Assessor
| What it does | The case assessor collects information about your employment situation through a structured questionnaire and uses AI to provide a preliminary indication of whether you may have a viable employment claim, what type of claim it might be, and an estimated settlement range. |
| Data processed | Your answers to the assessment questionnaire, which may include employment dates, salary, reason for dispute, employer details, and details of the treatment you have experienced (which may include special category data such as health information or protected characteristics). |
| AI providers used | OpenAI (GPT-4o-mini) for initial case scoring. Anthropic (Claude Sonnet) for full case assessments with settlement range estimates. |
| Decision-making role | Assistive, not determinative. The case assessor provides a preliminary indication only. It does not determine whether we will take your case, make promises about outcomes, or constitute legal advice. All case assessments are reviewed by a member of our team before any decision is made about whether to offer legal services. |
| Human oversight | Every case assessment is reviewed by a qualified person before any action is taken. The AI output is used as a starting point for that review, not as a final determination. If the AI assessment indicates a potential claim, a solicitor will review the details before any advice is given. |
| Limitations | The assessment is based solely on the information you provide and cannot account for nuances, context, or evidence that would emerge in a full consultation. An indication that you may not have a claim does not mean you definitely do not. We encourage you to seek a solicitor consultation if you are unsure. |
3.3 Tribunal Document Drafting
| What it does | Generates draft tribunal documents based on case information held on the platform. Documents include particulars of claim, witness statements, schedules of loss, chronologies, lists of issues, skeleton arguments, letters of claim, document bundles, case summaries, cast lists, and preliminary hearing agendas. |
| Data processed | Case details, employment information, salary and financial data, correspondence history, and any documents uploaded to the platform that are relevant to the draft being produced. This will routinely include special category data. |
| AI providers used | Anthropic (Claude Sonnet) for most tribunal document types. Anthropic (Claude Opus) for skeleton arguments and other premium documents. |
| Decision-making role | Drafting assistance only. The AI produces a first draft. It does not determine legal strategy, decide what arguments to advance, or make submissions on your behalf. |
| Human oversight | Where documents are generated under a signed Retainer, all AI-drafted documents are reviewed, edited, and approved by a qualified solicitor before being used. Where documents are generated through the free self-service Tribunal Preparation tool, the user is responsible for reviewing all outputs before use. All self-service documents are clearly marked as drafts that have not been reviewed by a solicitor. |
| Limitations | AI-generated drafts may contain inaccuracies, miss important legal arguments, or fail to reflect the full nuance of your situation. AI-generated content may include factual errors (sometimes called “hallucinations”) which are a known limitation of current AI technology. The solicitor review process (where applicable) or the user’s own review (for self-service documents) is designed to catch and correct these issues. |
3.4 Document Analysis and Classification
| What it does | Analyses uploaded documents to classify their type, extract text via OCR (optical character recognition), and identify relevant information for case management. |
| Data processed | Documents you upload to the platform, which may contain personal data and special category data. |
| AI providers used | OpenAI (GPT-4o) for OCR extraction and document classification. Google Gemini (via Lovable gateway) for document analysis with vision capabilities. |
| Decision-making role | Processing assistance only. These features organise and extract information from documents. They do not make decisions about your case. |
| Human oversight | Extracted data and classifications are presented for user or solicitor review. Automated classifications can be corrected manually. |
| Limitations | OCR and document analysis may produce errors, particularly with handwritten text, poor-quality scans, or complex document layouts. Users should verify extracted information against the original documents. |
3.5 Settlement Agreement Scanner
| What it does | Analyses settlement agreements to identify key terms, potential issues, and areas that may require solicitor attention. |
| Data processed | The text of settlement agreements uploaded to the platform, which will typically contain personal data including names, employment details, financial terms, and potentially special category data. |
| AI providers used | Anthropic (Claude Sonnet). |
| Decision-making role | Analysis assistance only. The scanner highlights points of interest. It does not advise you on whether to accept or reject a settlement agreement. |
| Human oversight | Settlement agreement reviews are always conducted by a qualified solicitor under a signed Retainer. The AI analysis is a starting point for the solicitor’s review, not a substitute for it. |
| Limitations | The scanner may miss unusual clauses or fail to identify jurisdiction-specific issues. Solicitor review is essential. |
3.6 Voice Transcription and Text-to-Speech
| What it does | Transcribes telephone calls and audio recordings into text. Generates voice greetings and spoken responses for call handling. |
| Data processed | Audio recordings of telephone calls (which may contain personal data and special category data discussed during calls) for transcription. Text content for speech generation (typically short greeting scripts). |
| AI providers used | Deepgram (Nova-2) for speech-to-text transcription and voice agent interaction. ElevenLabs for text-to-speech voice synthesis. |
| Decision-making role | Processing only. These features convert between audio and text formats. They do not make decisions about your case. |
| Human oversight | Transcriptions are available for review by the legal team. Voice greetings are configured by staff. |
| Limitations | Transcription accuracy depends on audio quality, accents, background noise, and technical terminology. Transcriptions should be treated as approximate and verified against original recordings where precision matters. |
3.7 Internal and Administrative AI Features
In addition to the user-facing features described above, we use AI for internal administrative functions. These include generating system update summaries, formatting solicitor comments, extracting structured data from internal notes, enhancing content descriptions, and generating bug report titles. These functions primarily process internal operational data rather than client personal data. Where client data is involved (for example, when extracting data from case notes), the same data protection safeguards apply as described throughout this policy.
4. AI Providers We Use
LexConnectPro uses the following AI providers. We select providers based on capability, data protection practices, and contractual commitments regarding data handling.
| Provider | Service | Data Training Policy | Data Retention by Provider | Location |
|---|---|---|---|---|
| Anthropic (Claude) | Large language model API. Primary provider for tribunal document drafting, case assessments, legal chat, and transcript extraction. | API inputs and outputs are not used for model training under commercial API terms. | Inputs and outputs retained for up to 30 days for abuse monitoring, then automatically deleted. | USA |
| OpenAI | Large language model API. Used for AI assistant chat, case scoring, document classification, OCR, legal Q&A, and data extraction. | API inputs and outputs are not used for model training under commercial API terms. | Inputs and outputs retained for up to 30 days for abuse monitoring, then automatically deleted. | USA |
| Google (Gemini, via Lovable gateway) | Large language model API with vision capabilities. Used for document analysis, content suggestions, and chronology assistance. | API inputs and outputs are not used for model training under paid API terms. | Prompts and responses retained for up to 55 days for abuse monitoring, then automatically deleted. Lovable gateway logs retained for up to 90 days. | USA/EU |
| Deepgram | Speech-to-text transcription and voice agent. | Audio data is not retained after transcription is complete. | Audio processed in real time and not stored beyond the transcription session. | USA |
| ElevenLabs | Text-to-speech voice synthesis for call greetings and voice responses. | Data is not used for model training unless explicitly opted in via account settings (we have not opted in). | Generation history retained by default. Deletable via API, with backup data retained for up to 30 days after deletion. Enterprise Zero Retention Mode available. | USA |
We review our AI provider arrangements regularly and will update this policy if providers change. All providers are subject to data processing agreements or commercial API terms and the international transfer safeguards described in our Privacy Policy.
Important note on provider data retention: When your data is sent to an AI provider for processing, the provider may temporarily retain your inputs and outputs for their own abuse monitoring and policy enforcement purposes. This retention is a condition of the provider’s commercial API terms and is separate from the data we store on our own platform. We do not control these provider retention periods. After the stated period, the data is automatically deleted by the provider and is not used for model training. Full details of provider retention periods are set out in the table above and in our Privacy Policy.
5. Principles Governing Our Use of AI
Our use of AI is guided by the following principles, which align with the UK government’s five cross-sectoral AI principles and the ICO’s guidance on AI and data protection:
5.1 Safety, Security and Robustness
We implement technical safeguards to ensure AI features operate reliably. All AI API calls are made server-side through Supabase Edge Functions via encrypted connections. We do not expose API keys or model configurations to end users. We monitor AI outputs for quality and accuracy and maintain the ability to disable any AI feature immediately if safety concerns arise.
5.2 Transparency and Explainability
We clearly label all AI-generated content on the platform. Users are informed when they are interacting with an AI system (such as Lexi) rather than a human. This policy provides detailed information about how each AI feature works, what data it processes, and what role it plays in decision-making.
5.3 Fairness
We are aware that AI systems can produce biased or discriminatory outputs. We take steps to identify and mitigate bias in our AI features, including reviewing AI outputs across different case types and protected characteristics. If you believe an AI feature on our platform has produced a biased or discriminatory output, please contact us at mail@tomstreet.co.uk.
5.4 Accountability and Governance
Tom Street, as Data Protection Officer and principal of the firm, has overall responsibility for the governance of AI within LexConnectPro. Decisions about which AI features to deploy, which providers to use, and how AI outputs are integrated into legal services are made at the leadership level. We maintain records of our AI processing activities as part of our GDPR Article 30 records of processing.
5.5 Contestability and Redress
You can challenge any AI-generated output or decision. If you disagree with an AI assessment, you can request human review. If you believe an AI-related error has caused you harm, you can raise this through our complaints procedure.
6. Your Rights Regarding AI and Automated Decision-Making
Under UK GDPR Article 22, as amended by the Data (Use and Access) Act 2025 (in force from 5 February 2026), you have enhanced rights regarding automated decision-making:
6.1 Right to be Informed
You have the right to know when AI or automated processing is being used. We provide this information through on-screen notices within the platform, this policy, and our Privacy Policy.
6.2 Right to Meaningful Information About the Logic
You have the right to understand the logic involved in automated processing. Section 3 of this policy provides this information for each AI feature. If you require further explanation about how a specific AI output was generated, you can request this from our DPO.
6.3 Right to Human Intervention
You have the right to request that a human reviews any AI-generated output or assessment. For documents generated under a Retainer, all AI features already include mandatory solicitor review. For the free self-service Tribunal Preparation tool, you are the reviewer, but you may request additional human review by contacting us.
6.4 Right to Express Your Point of View
You have the right to express your point of view regarding any automated processing and to have that view taken into account before any decision is made.
6.5 Right to Contest
You have the right to contest any decision that has been influenced by automated processing. You can do this by contacting our DPO at mail@tomstreet.co.uk or through our complaints procedure.
7. AI and Legal Professional Obligations
As a firm regulated by the Solicitors Regulation Authority, we are bound by professional duties that apply to all of our work, including work assisted by AI:
- Duty of competence (SRA Principle 2): solicitors reviewing AI outputs are responsible for ensuring the final work product meets professional standards
- Duty to act in the best interests of clients (SRA Principle 7): AI is used to improve efficiency and access to justice, not to replace the professional judgement that clients are entitled to
- Duty of confidentiality (SRA Principle 6): all data shared with AI providers is subject to our confidentiality obligations and is governed by data processing agreements or commercial API terms that prohibit use of data for model training
- Duty not to mislead (SRA Code of Conduct, para 1.4): we do not present AI-generated outputs as the work of a solicitor without appropriate review, and we are transparent about the role AI plays in our services
8. Data Protection Impact Assessment
UK GDPR Article 35 requires a Data Protection Impact Assessment (DPIA) where processing is likely to result in a high risk to individuals’ rights and freedoms. Given that LexConnectPro processes personal data (including special category data) through AI systems in a legal context, a DPIA is required.
We are in the process of completing our DPIA. It will cover the processing of personal data (including special category data) by all AI providers listed in Section 4, the risks associated with automated decision-making in a legal context, the temporary retention of data by AI providers on US-based servers, and the safeguards we have put in place to mitigate those risks.
Our DPIA will be reviewed and updated whenever we introduce new AI features, change AI providers, or make significant changes to how existing AI features process personal data. Once completed, a summary of our DPIA findings will be available on request from our DPO.
9. Accuracy and Hallucination Risk
Large language models can produce outputs that are plausible but factually incorrect (sometimes called “hallucinations”). This is a known limitation of current AI technology. We manage this risk through:
- Mandatory solicitor review of all AI-generated legal documents before use (where generated under a Retainer)
- Clear labelling of all self-service documents as drafts that have not been reviewed by a solicitor, with the user responsible for checking accuracy
- Prompt engineering designed to ground AI responses in relevant legal frameworks and factual data
- Date accuracy systems to prevent common AI errors such as incorrect year references in drafted letters and documents
- Ongoing monitoring of AI output quality across different case types
- User feedback mechanisms that allow users to flag AI outputs they believe are incorrect
We do not guarantee the accuracy of any AI-generated output. All AI outputs on this platform should be treated as drafts or preliminary guidance, not as final legal advice or verified legal documents.
10. Future Changes
We may introduce new AI features or change the AI providers we use. Where such changes materially affect how your data is processed or the role of AI in our services, we will update this policy and, where appropriate, notify you directly. We will not introduce AI features that make determinative decisions about your legal rights without your knowledge and without providing the rights set out in Section 6.
11. Contact
If you have any questions about our use of AI, wish to exercise your rights regarding automated decision-making, or want to report a concern about an AI output, please contact:
Tom Street (Data Protection Officer)
Tom Street & Co. Solicitors Limited
Maidsgrove Farm, Frome, Somerset, BA11 2PY
Email: mail@tomstreet.co.uk
Telephone: 020 3835 3940