# Epic AI-20 - AI Text Assistant

Project: Stream AI
Status: in development
URL: https://community.openproject.org/work_packages/AI-20

## Detailed specification

### Motivation and background information

### Acceptance criteria

### Out of scope

| Attribute | Value |
| --- | --- |
| Mockups |  |

## Planning

| Attribute | Value |
| --- | --- |
| Roadmap | 2026 Q4 |
| Date | 2026-10-23 – 2026-11-03 |
| List | RM AI |
| Target versions | 18.0.0 |
| Relevance |  |

## Subitems

| ID | Type | Subject | Status | Assignee | Target versions | Project | Priority | Spent time |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| AI-3 | Feature | Add LLM connection settings to admin AI configuration | in test | Ihor Khomenko | 18.0.0 | Stream AI | Medium |  |
| AI-98 | Feature | Admin UI: manage description assistant actions | in test | Ihor Khomenko | 18.0.0 | Stream AI | Medium |  |
| AI-101 | Feature | CKEditor dropdown to trigger AI-Actions | merged |  | 18.0.0 | Stream AI | Medium |  |
| AI-126 | Feature | Show results of AI-Actions in AI result popover | merged | Jonas Jabari | 18.0.0 | Stream AI | Medium |  |
| AI-156 | Feature | Documentation link at the work package type template setting | new |  | 18.0.0 | Stream AI | Medium |  |
| AI-166 | Feature | Text actions enterprise banner in admin settings | merged |  | 18.0.0 | Stream AI | Medium |  |
| AI-206 | Feature | Remove toggle in text assistant admin settings | merged |  | 18.0.0 | Stream AI | Medium |  |
| AI-96 | Implementation | Prompt collection: data model & seeded default actions | closed | Jonas Jabari |  | Stream AI | Medium |  |
| AI-100 | Implementation | Text transform API (list & execute) | in development | Jonas Jabari |  | Stream AI | Medium |  |
| AI-102 | Open Point | EE Feature? | decided |  |  | Stream AI | Medium |  |
| AI-198 | Implementation | Optimize Result Pan Run Robustness | specified |  |  | Stream AI | Medium |  |

## Bugs and other work packages

| ID | Type | Subject | Status | Assignee | Target versions | Module |
| --- | --- | --- | --- | --- | --- | --- |
| AI-205 | Bug | LLM Settings caption changes | merged |  | 18.0.0 | Artificial Intelligence (AI) |
| AI-204 | Bug | "Remove key" button next to the stored API key field on the LLM connection page looks out of place | merged |  | 18.0.0 | Artificial Intelligence (AI) |
| AI-155 | Bug | Rename Text transform actions to Text assistant in the administration | merged |  | 18.0.0 | Artificial Intelligence / AI text helper (AITH) |
| AI-102 | Open Point | EE Feature? | decided |  |  |  |
| AI-97 | Task | Define default prompt set & align with openproject-llm-evals | new |  |  |  |

## Details

| Attribute | Value |
| --- | --- |
| Module |  |
| Enterprise plan | Enterprise / Professional (PRO) |
| Labels |  |

## People

| Attribute | Value |
| --- | --- |
| Assignee | Dominic Bräunlein |
| Accountable | Dominic Bräunlein |
| Product manager | Dominic Bräunlein |
| Designer |  |
| Developers | OP Dev AI team |
| Interest by |  |

## Estimates and progress

| Attribute | Value |
| --- | --- |
| Work | 60d 0h · Σ 60d 0h |

## Description

# User Problem

## User

*   Product owner and managers
    
*   Project members
    

## Problem

*   Project management is only as good as the data put into the system. Planning, reporting, and collaboration all depend on work packages being complete, well-structured, and understandable. But entering that data can be slow and effortful, so sometimes it doesn't happen.
    
*   Users might have already unstructured input and need to bring it into a well-structured work package that follows the company template. The effort of structuring input is a barrier: the harder it is, the less and the worse the data that ends up in the system.
    
*   Many users have to write descriptions in a language that is not their mother tongue. Formulating clear, correct text costs them extra time and the results are harder for others to work with.
    
    <br>
    

## Pain

*   Users copy text into external AI tools (e.g. ChatGPT) to rephrase, then paste back which breaks the flow, loses formatting, and raises data privacy concerns.
    
*   Reviewers correct grammar and structure by hand in follow-up edits instead of the author getting it right the first time.
    
*   The most common &quot;workaround&quot; is silent: work packages get created with thin or empty descriptions, and the missing information is compensated later through meetings, chat questions, and rework.
    

# Business Case

##   

## Solution

*   An AI writing assistant embedded in the work package editor that lowers the barrier to getting good data into the system: users can input a quick brain dump in whatever form comes naturally, and the assistant handles structure and formulation.
    
*   The Assistant can be extended by admins with their own functionality in form of prompts
    

##   

# Launch and Growth

## Messaging

<figure class="table op-uc-figure_align-center op-uc-figure"><table class="op-uc-table"><tbody><tr class="op-uc-table--row"><th class="op-uc-table--cell op-uc-table--cell_head"><p class="op-uc-p">Headline</p></th><td class="op-uc-table--cell"><p class="op-uc-p">Better text, right where you write.</p></td></tr><tr class="op-uc-table--row"><th class="op-uc-table--cell op-uc-table--cell_head"><p class="op-uc-p">First Paragraph</p></th><td class="op-uc-table--cell"><p class="op-uc-p">Good project management starts with good data. OpenProject now helps your team get it into the system faster.</p><p class="op-uc-p">Paste in a quick brain dump, an email or meeting notes, and the text assistant sorts it into your team’s own template. It also proofreads, rewrites, adjusts the tone, summarizes and suggests a subject, all without leaving OpenProject. Admins can add their own actions to fit how your team works.</p></td></tr><tr class="op-uc-table--row"><th class="op-uc-table--cell op-uc-table--cell_head"><p class="op-uc-p">Customer Quote</p></th><td class="op-uc-table--cell"><p class="op-uc-p"><br data-cke-filler="true"></p></td></tr></tbody></table></figure>

##