Cases

Projects that move the needle.

A selection of recent work — each one written up as Problem, Solution, and Measured Impact.

AI #rag-knowledge-base

1. The Problem: The internal teams lost hours per month searching docs on various scattered knowledge bases.
2. The Solution: A RAG system on Azure OpenAI with an evaluation pipeline, source citations, and a feedback loop for continuous improvement.

Tech used

3. The Impact

3.4× Support team load reduction

AI #document-processing

1. The Problem: Repititive manual work to extract and process information from complex documents (invoices, CVs) with high accuracy requirements.
2. The Solution: An information extraction API endpoint using OpenAI, with a human-in-the-loop review interface and integration into existing workflows.

Tech used

3. The Impact

70% less manual work

AI #multi-agent-farm

1. The Problem: Re-occuring complex workflows that require multiple agents with different capabilities and connected data sources, leading to inefficiencies and manual work.
2. The Solution: A framework and Teams app to easily consume and interact with multiple agents, each with different capabilities and connected data sources, to automate complex workflows across systems. Centrally managed and monitored with Langfuse.

Tech used

3. The Impact

70% less manual work

AI #jobsearch-agent

1. The Problem: Users only had one traditional way of searching for jobs, which was time-consuming and sometimes ineffective in finding relevant opportunities.
2. The Solution: An agentic job search assistant, balancing speed and accuracy, that interacts with users in a conversational manner, understands their preferences and qualifications, and proactively searches and applies for relevant job opportunities across multiple platforms.

Tech used

3. The Impact

10% More monthly applications.

AI #prescreening-agent

1. The Problem: Prescreening candidates for job openings was a time-consuming process that often resulted in qualified candidates being overlooked due to the high volume of applications.
2. The Solution: An agentic prescreening assistant, interacting with candidates in a conversational manner, understanding their, performing the task of prescreening by asking relevant questions, and syncing everything back to the CRM. This assistant helps to quickly identify qualified candidates, ensuring that they are not overlooked and improving the efficiency of the hiring process.

Tech used

3. The Impact

91% Time reduction between application and prescreening completion.

AI #creative-structured-writing

1. The Problem: Various use cases across the organization require generating content with a specific structure, content and tone-of-voice, but often resulted in a lot of manual iteration or repetitive and unaligned output from existing process.
2. The Solution: Structured content generation assistants, tied to internal tools and databases, that can generate content with specific structure, content and tone-of-voice requirements.

Tech used

3. The Impact

91% Time reduction between application and prescreening completion.