top of page

b.woah! Where’d all this AI come from?

By Jack Crawford, b.world Engineer


With the truly overwhelming potential of generative AI, and in the context of ensuring responsible use, we had to ask a hard question:

How can social impact teams use GPT to increase their capacity to deliver on their mission?

Our answernow available in b.worldwas twofold:

  1. Help impact teams start projects faster by lowering barriers to entry through AI-assisted project creation and design.

  2. Improve data literacy by summarizing impact results to generate insights.

  3. Reduce the time to generate an impact story to seconds by powering b.world's storybuilder with generative AI.

Based on these answers, and in just the last month or so, we've AI enabled b.world with the following new AI assisted features:

  • Project Generation

  • Results Framework Design on the Whiteboard

  • Data Summarization

  • Story Generation

Each of these tools makes life in b.world easier, saving you time so you can focus on impact. But with all the questions around data and AI, it’s important to know exactly what goes where! So, let’s look at what parts of your data are used to power b.world’s new AI features.


Project Generation

Projects are the foundation of b.world, and setting up your new or existing projects on the platform should be as easy as possible. That made Project Creation the perfect jumping-off point for our GPT-powered features. Paste a few paragraphs about your project into the prompt, and hit Create Project. Using AI, we discern a good title, description, challenge statement, and proposed solution for your project, and spin it up for you!


This is the simplest one: new Projects are generated using only the contents of the text input. No other content from your b.world instance is included in the prompt.


IN: Project description paragraphs

OUT: b.world Project Overview, including Title, Description, Challenge Statement, and Proposed Solution



Results Framework Design

Whiteboards within b.world have always been a great way to brainstorm and outline results frameworks, then push those outlines into the Logframe to start measuring your project’s progress and impact.


New on Whiteboards: Draft with AI generates an outline of a results framework for your project, based on your Project Overview.


IN: Project Overview (Title, Description, Challenge Statement, Proposed Solution)

OUT: First draft of results framework, a goal, three objectives, three activities, and indicators at each level.



Data Summarization

Once you’ve recorded some measurements within an indicator, hit the magic Insights button to generate an analysis of the data.


IN: Measurement values and dates. Stakeholder names are not included.

OUT: Insight into measurement data trends and changes.



Story Generation

Creating impact stories has always been easy on b.world. After choosing a story template, drag and drop written content from throughout your project, then include data visualizations, maps and other media to underscore the narrative. The power of data-driven storytelling was already innate to the platform, so Stories were ready for AI generation right away. In fact, there are actually four types of generation within stories:


Story Title Generation

To get a new title for your story, hit “Title with AI” at the top of your story. GPT will read your story and decide on a compelling title.


IN: Written story content, not images, maps, charts or videos

OUT: Story title



Single Answer Generation

For any prompt within a story template, hitting “Draft with AI” will come up with an answer, taking relevant project data into account.


IN: Project Overview (outlined above), Logframe content (Goals, Objectives, Activities, Indicators) and Insights. Not included: Indicator measurements, Stakeholder information, Budgets, Tasks

OUT: Answer to story prompt.


Custom Story Answer Generation

For any text block within a custom Story, hitting “Draft with AI” will open a popover. Enter your own prompt or question about the project, and it will use the same set of information as a template prompt to generate a response.


IN: Project Overview (outlined above), Logframe content (Goals, Objectives, Activities, Indicators) and Insights. Not included: Indicator measurements, Stakeholder information, Budgets, Tasks

OUT: Answer to custom prompt.



Full Story Generation

Whether you hit the magic “Draft a Story” button anywhere in a project, or turn on “Draft with AI” when creating a new story, this feature will generate an entire story with responses to each prompt, as well as a relevant title based on those answers. This tool uses the same dataset as single-answer generation.


IN: Project Overview (outlined above), Logframe content (Goals, Objectives, Activities, Indicators) and Insights. Not included: Indicator measurements, Stakeholder information, Budgets, Tasks

OUT: An entire first draft of your story, ready for revision and expansion. Make sure to add some visual elements to create a deeper narrative!




Building these features was, and will continue to be, an educational experience. The biggest hurdle was the composition of effective and accurate prompts, especially for results framework generation on Whiteboard. That generation needed to be formatted properly (so that we could parse the response and turn it into shapes on a whiteboard) and more importantly, contextual: generated Objectives should make sense within their parent Goal, indicators should correspond to their parental elements, etc. Devising that prompt was a lengthy series of trials and errors, but we’re very happy with the results.


Go try the new GPT features on b.world today!

133 views0 comments

Comments


bottom of page