Using AI to build the first chronology in an employment investigation

What each timestamp means for the chronology

Sorting these records by their visible times produces an order, but leaves the decision sequence unresolved. The draft's timestamp must retain its stated meaning; it does not automatically establish when approval occurred or when the change was first proposed. The complaint email records a communication to human resources, without establishing the operations manager's knowledge at 10:15. Publication establishes which roster was made available at that point. The effective date records the intended start of the reassignment, while evidence of actual implementation remains missing.

Design the chronology around decision stages. Include columns for event date and time, what the timestamp represents, decision stage, actor, neutral event description and exact source reference. For communications, record sender, recipient and sending time where supplied. Preserve source versions and a page, message or other location reference so each entry can be traced back. Add an unresolved timing column for unknown approval dates, uncertain event times and missing evidence of receipt or reading. An empty field should remain visibly unresolved rather than acquiring a plausible answer.

The extraction instruction and the events it supports

Give AI a bounded extraction instruction: "Separate proposed changes, recorded decisions, communications and actual implementation. Identify the source, version and relevant passage for each event. Distinguish the document date from the date of any event it describes. Record sender and recipient only where identified. Do not infer reading from sending or attribute motives to participants. Flag missing links explicitly." Ask it to preserve ambiguous wording as well. The phrase "as discussed" should remain available for examination without becoming an invented meeting, approval date or account of who attended.

Walk through the extracted events. Version 1 supports an entry showing a proposed reassignment. The 09:00 email supports a complaint communication. The 10:15 message records an instruction to the scheduler, but does not identify the earlier discussion's time or the roster version under discussion. Publication of version 3 needs its own entry. The 8 June date belongs in an entry labelled scheduled implementation, with actual implementation unconfirmed. A recorded approval cannot be inserted simply because the published roster must have emerged from some decision process.

Competing readings and targeted evidence requests

Two competing readings now become visible. The reassignment may have been settled before the complaint, with the draft reflecting an existing decision. Alternatively, the draft may have been an option that remained open and was confirmed or changed later. The four records do not resolve that difference. Missing version 2 matters because it could help explain changes between proposal and publication. Its absence from this packet does not itself establish concealment. Even matching assignments in versions 1 and 3 would not, by themselves, establish when the change became final.

Turn the gaps into targeted evidence requests. Who had authority to approve the roster, and how was approval recorded? Is there a record of the discussion referenced at 10:15? What did version 2 contain? When did each relevant participant learn about the complaint, and through which communication? A later reply discussing the complaint's contents could support knowledge by the time of that reply, while leaving the earliest point of knowledge uncertain. Attendance or shift records could help establish actual implementation. Each request should identify the particular uncertainty it is intended to resolve.

The next legal questions can then be framed with greater precision. Which action requires assessment: the initial proposal, approval, publication or implementation? Who participated at that stage, and what did they know? What explanation for the reassignment was recorded at the time? Those factual questions help the lawyer identify the applicable legal issues and procedures without treating a short interval as a finding about motive. Temporal proximity makes the sequence worth investigating. On this packet, the first useful deliverable is a chronology with explicit gaps and focused follow-up questions.

Standing instructions for chronologies in Wisanna

In Wisanna, a firm administrator can define standing AI instructions that apply across the firm's work. The distinction illustrated here provides a concrete rule a firm could adopt: "In chronologies, separate proposed changes, recorded decisions, communications and actual implementation; preserve source versions and flag unknown timing." This is an example of an instruction the firm could define, without assuming a dedicated investigation feature. For this assignment, the rule keeps a consequential factual question visible: at what point did the proposed reassignment become a settled decision?

Wisanna is a private and secure legal-AI workspace built for lawyers and legal professional work. AI-assisted outputs require lawyer verification and professional judgment, including checking whether the chronology has mistaken a publication date for an approval date or a scheduled shift for a completed one. The intended result here is a traceable sequence that makes the next factual inquiry specific enough to pursue. Request a Wisanna demo to discuss how standing instructions can support focused legal assignments such as this first investigation chronology.