How to Use AI to Pull Your Work History From Slack, Docs, and Tickets

The opportunity

Most people do not need to remember everything from scratch. Their work history is already sitting in systems they touch every day: Slack, Linear, Jira, GitHub, Notion, docs, and meeting notes. An AI assistant can review those systems and collect the raw material much faster than a human doing manual archaeology.

The key is asking the model for raw wins with evidence, not a vague recap. The model should retrieve and package source-grounded material, while Tally handles the heavier career-story shaping afterward.

What to ask the AI to find

Your prompt should bias the model toward items that matter later in reviews, promotion cases, and resumes:

• Shipped work and launches
• Measurable improvements or business impact
• Cross-functional coordination
• Leadership, ownership, or decision-making
• Customer impact, escalations, or incident handling
• Process improvements and mentoring

That keeps the artifact focused on material that actually helps the user later.

What output format works best

Do not ask for prose. Ask for a simple JSON dump of raw wins plus evidence. That keeps the model focused on retrieval and makes the result much easier to process inside Tally.

At minimum, each raw win should include:

• A title and concise summary
• Date or date range
• What happened
• Any metrics mentioned in the source material
• People involved
• Source references back to Slack, docs, tickets, or PRs
• Quoted snippets when useful

Sample prompt

This is a strong default prompt users can paste into Claude, ChatGPT, Gemini, or an internal agent that has access to their systems:

You are helping me gather raw evidence of my recent work so I can paste it into Tally.

Review any sources you can access from the last 90 days, including Slack, docs, tickets, PRs, and meeting notes.

Your job is to identify meaningful wins, projects, launches, milestones, and notable contributions. Focus on retrieval, not interpretation. Merge related evidence into a single raw win when it clearly refers to the same project or result.

For every raw win:
- include a concise factual summary
- include dates or an approximate date range
- include what happened
- include any metrics mentioned in the source material
- include people involved
- include source references
- include short quoted snippets when useful

Do not invent facts. If data is missing, leave the field null instead of guessing.

Return ONLY valid JSON in this shape:
{
  "employee_name": "string",
  "time_window": {
    "start": "YYYY-MM-DD or null",
    "end": "YYYY-MM-DD or null"
  },
  "raw_wins": [
    {
      "title": "string",
      "summary": "string",
      "date_start": "YYYY-MM-DD or null",
      "date_end": "YYYY-MM-DD or null",
      "what_happened": "string",
      "why_it_might_matter": "string or null",
      "metrics_mentioned": ["string"],
      "people_involved": ["string"],
      "systems_checked": ["string"],
      "evidence": [
        {
          "type": "message | doc | ticket | pr | note | other",
          "label": "string",
          "date": "YYYY-MM-DD or null",
          "reference": "string"
        }
      ],
      "source_quotes": ["string"]
    }
  ],
  "open_questions": ["string"]
}

Privacy note: users should not send confidential company strategy, regulated data, customer secrets, legal material, or anything their employer does not permit them to share with third-party AI tools. When in doubt, they should redact names, metrics, and links first.

Example artifact

Here is a small example of the kind of raw output we want:

{
  "employee_name": "Alex Santos",
  "time_window": {
    "start": "2026-01-01",
    "end": "2026-03-31"
  },
  "raw_wins": [
    {
      "title": "Coordinated onboarding simplification launch",
      "summary": "Worked across product, design, and engineering on onboarding simplification for new users. Source material shows launch coordination and post-launch follow-up.",
      "date_start": "2026-01-12",
      "date_end": "2026-02-28",
      "what_happened": "Coordinated launch-related work and decision tracking for the onboarding simplification effort.",
      "why_it_might_matter": "Appears tied to improved activation and a smoother first-run experience.",
      "metrics_mentioned": ["Activation improved from 38% to 52%"],
      "people_involved": ["Growth Engineering", "Product Design", "Lifecycle Marketing"],
      "systems_checked": ["slack", "docs", "linear"],
      "evidence": [
        {
          "type": "message",
          "label": "Slack launch thread",
          "date": "2026-02-21",
          "reference": "slack://growth/onboarding-launch-thread"
        },
        {
          "type": "doc",
          "label": "Launch plan",
          "date": "2026-02-10",
          "reference": "Onboarding v2 Launch Plan"
        }
      ],
      "source_quotes": [
        "Alex is owning launch coordination and final decision tracking."
      ]
    }
  ],
  "open_questions": [
    "Confirm final activation percentage from analytics dashboard."
  ]
}

Why this is better than asking for a summary

A loose recap usually sounds polished but loses the details that matter most later. It might say, "You supported onboarding improvements this quarter," which is pleasant and almost useless. A raw artifact preserves the parts that actually compound:

• what changed
• what evidence exists
• what metrics were actually mentioned
• which people and systems were involved

Once those fields exist, Tally can help turn them into performance-review inputs, promotion evidence, and resume bullets without forcing the user to dig all over again.

Guardrails to recommend to users

• Start with a limited date range like 30 to 90 days.
• Ask for source references and quoted snippets.
• Tell the model to leave unknown fields null rather than hallucinating.
• Do not send private company information unless the user is explicitly allowed to share it with that AI tool.
• Have the user review the artifact before import, especially if sensitive channels are involved.
• Run a second pass only to fill obvious missing references or dates, not to do more interpretation.

The positioning for Tally

The AI does the retrieval. Tally does the organizing, categorizing, and reusing. That is a clean story: users can leverage the tools they already have access to, then bring the output into a product built specifically for career evidence.

That makes this a strong content angle for acquisition too. People are already searching for ways to use AI to summarize Slack, export work history, and prepare for reviews. We can meet that intent with a practical workflow instead of a generic "AI for careers" post.