The problem
Employment tribunal bundles are notorious. In anything other than the most straightforward case, you are looking at hundreds of pages: emails, policies, meeting notes, disciplinary records, grievance correspondence, contracts, payslips, medical records, and witness statements, often poorly organised and sometimes scanned at an angle. Reading through all of it carefully, extracting the relevant facts, identifying inconsistencies, and building a coherent chronology is one of the most time-intensive tasks in employment law practice.
The problem compounds when bundles arrive late, as they frequently do. A bundle landing on your desk on a Thursday for a hearing on Tuesday creates real pressure. Every hour spent indexing and summarising is an hour not spent on strategy, witness preparation, or drafting submissions. Junior fee earners can help but they need supervision, and the risk of missing something important in a large bundle is a genuine professional concern.
Beyond the time cost, there is a quality issue. When you are reading under pressure, things get missed. A single email buried on page 387 can be the most important document in the bundle, but when you are working through hundreds of pages in a hurry, it is easy to skim past it. A systematic AI-assisted review does not get tired, does not skim, and can be asked to look specifically for particular types of documents, dates, names, or themes.
The system
Step 1: Prepare and upload the bundle (NotebookLM)
The starting point is getting the bundle into a format AI can work with. If you have a PDF bundle, upload it directly to NotebookLM. If the bundle is in separate documents, upload them individually or compile them first. NotebookLM allows you to upload multiple sources and then ask questions across all of them.
Before uploading, note the key facts of the case: the claimant's name, the employer, the dates of key events, the nature of the claim, and any particular areas of dispute. This context will make your subsequent queries much more targeted.
Initial prompt example: "I have uploaded an employment tribunal bundle for a case involving unfair dismissal and discrimination. The claimant is [name], a [job title] employed by [employer] from [date] to [date]. The dismissal took place on [date]. The stated reason for dismissal was [reason]. Please confirm you can access all uploaded documents and give me a brief overview of the main document types present."
Step 2: Extract the chronology (NotebookLM)
Ask NotebookLM to build a chronology of key events from the documents. Be specific about what you want captured.
Prompt example: "Please create a chronology of key events in this case in date order. Include: all formal meetings, any disciplinary or grievance proceedings, key correspondence between the parties, any references to performance issues or complaints, medical certificates or occupational health referrals, and the dismissal and appeal process. For each entry, note the date, what happened, and the document reference where possible."
Review the chronology carefully against the bundle. NotebookLM draws directly from the documents you uploaded, so it should be accurate, but errors do occur, particularly with poorly scanned documents or ambiguous dates. Correct anything that is wrong before using the chronology in any proceedings.
Step 3: Identify key documents and inconsistencies (NotebookLM)
Once you have a chronology, use NotebookLM to interrogate the bundle for specific issues relevant to your case theory.
Prompt example: "Based on the documents, what evidence is there that the claimant raised concerns about [specific issue] before the dismissal? Please quote directly from the relevant documents with page references."
Further prompt example: "Are there any apparent inconsistencies between the employer's account in the ET3 response and the documentary evidence in the bundle? Please identify them with document references."
This targeted questioning is where AI adds the most value. You are effectively conducting a document review that would normally require hours of careful reading, in a matter of minutes.
Step 4: Draft the case summary and key facts document (Claude)
Take the output from your NotebookLM analysis and use Claude to shape it into a polished case summary and key facts document suitable for use in preparation for the hearing.
Prompt example: "I am preparing for an employment tribunal hearing. Below is my working chronology and a list of key documents and inconsistencies I have identified. Please draft a structured case summary with the following sections: (1) Background and key facts, (2) The claimant's case in summary, (3) The respondent's case in summary, (4) Key documents and what they show, (5) Points of dispute and evidential issues, (6) Areas where the claimant's position is strongest, (7) Areas requiring further investigation or attention. [Paste your working notes]"
Edit the output against the bundle. Add your own analysis and strategic judgement. The AI produces a useful first structure, but the final document needs your professional assessment of the evidence.
Step 5: Prepare hearing questions from the bundle (Claude)
Use Claude to draft suggested cross-examination questions based on the inconsistencies and key documents you have identified.
Prompt example: "Based on the following inconsistencies and key documents, please suggest cross-examination questions for [witness name], who was the decision-maker in the dismissal. The questions should explore the inconsistency between [X] and [Y], and test whether the stated reason for dismissal was the genuine reason. [Paste relevant extracts]"
These will need editing to match your style and the specific dynamics of the case, but having a structured starting point significantly reduces preparation time.
The results
Before: Reviewing a 400-page employment tribunal bundle from scratch typically took an experienced solicitor around six to eight hours to read, index, create a chronology, and produce a case summary. For a junior fee earner, the same task often took longer and required supervision.
After: The same process, using NotebookLM for document interrogation and Claude for drafting, typically takes two to three hours. The chronology is produced in minutes rather than hours. Inconsistencies that might have been missed under time pressure are surfaced systematically. One employment solicitor reduced their bundle review time by 60% across a three-month period, while reporting greater confidence that nothing significant had been overlooked. The time saved went directly into witness preparation and submissions, improving the quality of the advocacy rather than just the efficiency of the admin.