Our hypothesis was that genAI could be used to provide effective, tailored assistance and support to every employee, regardless of risk level, location, or language, in a way that could scale cost-efficiently for large organisations. 

We validated the proof-of-value and immediately got to work on a proof-of-concept with a single topic: Conflicts of Interest, a key part of an organisation's Anti-Bribery and Corruption policy. 

We were effective in creating a human-like exchange to ascertain whether something was a conflict of interest, the type and severity of that conflict and how it could be mitigated. This proved our hypothesis and gave us the confidence to build a broader product covering all of the compliance and Code of Conduct topics. 

Ombo, was born, in terms of product and proposition, to seamlessly integrate compliance into every employee’s daily work, helping large organisations across the world reduce compliance risks using advanced technology but with a natural language approach. 

Building advanced AI-powered compliance software starts with assembling a highly skilled, cross-functional team. Our team of software engineers, analysts, and AI specialists utilised an iterative agile approach, allowing rapid prototyping and fast-paced iterations to refine and improve the solution.

To create the product, we deployed a modern software stack, leveraging NextJS for front-end development and Vercel for hosting and continuous deployment. The AI core of our solution is powered by OpenAI's GPT-4o, enabling human-like interactions for compliance support.

Our AI development process was underpinned by a robust training suite that included prompt engineering, version control, AI-assisted training, and simulation tools. Automated testing was seamlessly integrated to ensure high standards of reliability and performance.

Key features include:

• Retrieval-Augmented Generation (RAG) with a reranking mechanism to optimise response accuracy.
• Model Fine-tuning for improved contextual understanding and compliance-specific responses.
• Eval-Driven-Development to consistently improve on underlying vector DB, prompt, and reranking optimisations.
• Use of multiple models to optimise performance and control costs.
• Integrations with document sources of truth and organisational Single Sign-On (SSO) systems, facilitating smooth adoption within existing corporate ecosystems.

‍

This combination of technology and agile delivery enabled the creation of a scalable, AI-powered compliance tool that seamlessly integrates with employee workflows.

User Experience

After a period of user research, we learned that the existing methods of finding information buried in a company intranet or having to remember an eLearning course that may have been completed months in the past weren’t effective at helping employees understand the ethical standards and compliance rules that they needed to follow. What worked effectively was when an employee had direct access to a compliance subject matter expert to speak with. They asked their questions and got clear guidance and advice that they could follow, as well as the reassurance that they were doing the right thing. 

We modelled our AI-powered conversational interface on a modern chatbot to ensure a seamless learning curve for employees, including non-tech-savvy ones. As well as our own web app interface we built an integration with Microsoft Teams to make it easy for employees to directly interact with Ombo on their own Teams account which employees at large organisations are already very familiar with. 

In addition to the user interface for employees, we created interfaces for compliance leaders to see data and insights from employee usage, from product metrics to specific topics of conversations and risks.

As well as accessing data and analytics, we created administrative interfaces for non-technical compliance subject matter experts to provide additional training to guide the AI following a logical feedback loop process, where feedback is turned into suggested improvements and implemented. 

These allowed the compliance and IT stakeholders to better understand how Ombo was being used and continue to improve it over time, without the need for software engineering support. 

Ombo has a rich product roadmap, from enterprise integrations to more advanced generative UI interfaces and key automations to carry out burdensome tasks. 

After recently winning the Tech Innovation of the Year Award, we’re looking to pilot Ombo with more global enterprise organisations over the coming months. Watch this space and learn more on ombo.ai.

Learn more about our approach to AI software development.