Hi there,I am Ng'ethe

I design & develop
data products

Schedule a call View Projects

Latest Products

Product Updates

Discover the latest enhancements and features, inspired by user feedback.

August 19, 2024

Stay up to date with Legal Fabric

Legal Fabric is a powerful new feature designed to keep lawyers up to date with the latest legal notices and statutory changes. This tool provides detailed outlines of modifications in legal texts, including amendments, deletions, repeals, and additions. Legal Fabric automatically highlights the affected Acts or Statutes, the specific type of legal effect, a clear description of the changes, and the source document for reference. By ensuring that legal professionals are always informed about the most recent legal developments, Legal Fabric enhances their ability to stay compliant and provide accurate, up-to-date advice to their clients.

Infrastructure

No changes in this version.

Impact

No measurable impact in this version

Knowledge

We introduced practice knowledge that assists the model to classify legal updates based on type of legal effect

Impact

The model accurately classifies the effect of the legal notice while sharing the appropriate reference in under 3 seconds.

Look & Feel

View past legal notices by month

Impact

Review the impact of legal notices at your leisure to improve your firm's overview of Kenya's legal landscape.

Model

Automatically processing, outlining and classification of affected Acts or Statutes, while offering detailed descriptions of these changes along with their source documents.

Impact

Lawyers can stay continuously informed on the latest legal developments, ensuring compliance and delivering accurate, up-to-date advice. This feature significantly enhances their ability to swiftly respond to legal landscape changes, improving client service and reducing the risk of oversight.

This is very useful as our current process is very manual and clanky.

July 12, 2024

Case Summaries and Multipage Layout

In less than 2 seconds, the client can now summarise cases using a standard Kenyan format and benefit from responses that are relevant in the Kenyan context. Furthermore, we have reduced the blank canvas problem by separating the product core features in a multipage application.

Infrastructure

No changes in this version.

Impact

No measurable impact in this version.

Knowledge

Our team added 2 more guidelines to improve response quality.

Impact

The case summary is twice as relevant as non-grounded LLM.

Look & Feel

Separated legal research and case summary functions into 2 separate pages.

Impact

The separation of activities improved ease of use and reduced the "empty canvas" problem.

Model

We equipped the model with a guidelines to produce insightful case and legal report summaries.

Impact

Lawyers can prepare a case summary that adheres to the Kenyan context and laws within 1 - 3 seconds.

Indeed, the model provides responses that cite better than Bard and ChatGPT.

July 3, 2024

IRAC Formatting and Enhanced Value Addition

The legal research model now boasts 20% more resources, doubling relevant insights and covering a broader legal spectrum. The chat input area is repositioned for easier navigation, and the IRAC framework boosts evaluation speed by 70%, enhancing productivity and decision-making.

Infrastructure

No changes in this version

Impact

No measurable impact in this version

Knowledge

The model has access to 20% more legal resources including best practices and opinions.

Impact

The client can gain twice as many relevant insights across a broader spectrum of the law compared to Co-pilot and ChatGPT.

Look & Feel

The chat input area is now positioned at the bottom.

Impact

The client scrolls half as much to view the response.

Model

The model develops opinions using the IRAC framework

Impact

The legal researcher is 70% faster at evaluating responses.

The response has captured a broad set of statutes. I had not thought of that angle.

June 25, 2024

Lower costs and enhanced interpretation of Kenya Laws

Our model now uses AWS S3 and Langgraph, cutting costs and improving output. With 30+ new statutes ingested, it offers 20% more relevant responses. A custom UI channel enhances accessibility, and deeper understanding of Kenya's laws boosts draft opinion accuracy.

Infrastructure

Access to the knowledge base on AWS S3. Changed model orchestration from CrewAI to Langgraph.

Impact

Cost reduction - Reduces the application resource utilisation on EC2 instances. Predictability - Consistent agent output because of increased level of control.

Knowledge

We ingested over 30 new statutes and notices.

Impact

Detail - The model can now provide 20% more volume of relevant responses to the user's question.

Look & Feel

Added a custom corporate UI channel

Impact

Improved accessibility - Users can procure a dedicated workspace to develop case projects.

Model

The model can now understand the hierarchy, dependency and principles of Kenya Laws across multiple sectors.

Impact

Accuracy - Improved detail in the draft opinions because of an advanced understanding of Kenya's legal environment.

Great understaning of our context and the level of detail is impressive!

September 7, 2024

Engage with your readers by improving the persuasive quality of your idea

Gain a deeper understanding of how your ideas resonate with both the intellect and emotions of your audience through a rhetoric analysis. By evaluating the logical structure, emotional appeal, and credibility of your work, you can identify areas for improvement and refine your message. This approach helps ensure your communication is clear, engaging, and authoritative, with a focus on actionable insights.

Infrastructure

No change in this version.

Impact

There is no measurable impact in this version.

Knowledge

No change in this version.

Impact

There is no measurable impact in this version.

Look & Feel

We introduced a visualisation of the rhetorical analysis and nested the evaluation in a high-contrast container.

Impact

The user can pick up useful insights 10 times faster.

Model

We equipped this model with the analytical capability to assess the different rhetoric components of a text.

Impact

By leveraging the automated insights. you can now accurately improve your idea's level of engagement and persuasion.

The rhetoric analysis and recommendations provide clear steps for enhancing the persuasion of my ideas.

August 28, 2024

Gain a 360 Degree view of your ideas

BROAD enables you to gain diverse perspectives on your ideas by leveraging large language models, writing frameworks and an evolving natural language analytics model. In this version, the writer can evaluate the core idea of their write-up and the different perspectives on the same. Each perspective includes a book, article and quote to encourage the writer to enrich and challenge their idea and the application takes about 1 - 3 seconds to develop the result. There are still issues regarding the accuracy of the link but this is a work in progress for the next version.

Infrastructure

Developed using the Streamlit framework for fast development

Impact

Depending on the complexity, the user's bug reports and feature requests are resolved within 1 hour to 3 business days.

Knowledge

The model is currently trained on writing frameworks and a few tenets of perspectives to create the different categories.

Impact

The categorisation yields both supporting and divergent set of concepts that enrich the writer's understanding of the landscape of their idea.

Look & Feel

The evaluation uses a cardinal framework that categorises the perspectives as "Ideas Originating from the Article," "Ideas Influencing the Article," "Ideas Opposed to the Article," and "Ideas Similar to the Article."

Impact

The reader can scan the evaluation 5 times faster than a standard paragraph text.

Model

Extract the core idea, four different perspectives and supporting material in under 3 seconds.

Impact

The writer can evaluate the perspectives that can challenge and support their idea and benefit from the proposed reading material.

Put me on this because I like the idea of seeing the opposing perspectives

🤔We think that:

Including a poll to a news update within a specific theme improves engagement within online communities

📊The data shows that:

Groups that received a poll after the news update had 200 more instances of conversations as compared to groups that received no poll.

😶So what?

Follow up a news update with a poll to prompt community members to perform a simple, low cost initial action.

Sample Size:

586 Whatsapp Users

Experiment #:

6

🤔We think that:

Adding links to LLM responses improves credibility

📊The data shows that:

18 of 20 respondents found the responses much more credible and useful because they could follow the link.

😶So what?

Enforce web browsing and link referencing for AI assistants/ agents to improve the user's capabilities to solve their problem.

Sample Size:

20

Experiment #:

5

🤔We think that:

Using AI assistants to progressively review AI document assessment

📊The data shows that:

8 out of 10 times, the report was voted as 'highly useful' and 'highly actionable' as compared to a simple completion prompt.

😶So what?

For tasks that involve more than one human contributor, use simple completion prompts to protptype assistants.

Sample Size:

10

Experiment #:

4

🤔We think that:

Preceeding time-boxed conversations with a timed poll on Whatsapp can increase engagement

📊The data shows that:

3 out of 3 participants engaged thrice with the topic and responses in a span of 7 minutes.

😶So what?

Expand the sample size of the test and include a control group to verify the results.

Sample Size:

3

Experiment #:

3

🤔We think that:

Whatsapp users are more likely to follow a link to an article if you precede the link with a short excerpt from the page

📊The data shows that:

Yes, thrice as many more recipients from the test group opened the link as compared to the group that received a link without a description.

😶So what?

Identify statements that resonate most with the target group and include it in a link before sharing. Additionally, include quotes in the sharing feature of your client's website.

Sample Size:

235

Experiment #:

2

🤔We think that:

Does a full screen hero image with a call-to-action improve user engagement?

📊The data shows that:

8 out of 10 viewers did not scroll past the full screen hero image despite the included call-to-action button.

😶So what?

Design the hero image to occupy 50 - 70% of the screen

Sample Size:

20 website visitors

Experiment #:

1

Ideation Sessions

Book a free 1-hour session below:.

The result of the session should be a measurable experiment that you can run with your team.

Articles

Designing a Research Dashboard

February 5, 2024

|

This is a highlight of a UX project to design and develop a research dashboard for a Research and Development company.

Externalizing Mental Frames

June 15, 2022

|

What if we could digitize our perspectives and identify areas of growth and more excitedly areas of commonality. Enabling Frames is a digital application designed to collect and digitise perspectives.

Afrikan Mbiu - A challenge-based community

June 15, 2022

|

Tribes can be challenging, better yet they can be centred on creating and taking challenges. This is my experience combining UX and psychology theories such as Self Determination Theory to achieve this goal for my client.

Using Crazy 8s for a Co-creation Session

April 22, 2022

|

Unorthodox Change Management

April 22, 2022

|

Using design thinking techniques to improve change management within a development team.