- The Accountancy Cloud is a fast-growing tech accounting start-up
- The objective was to automate the bookkeeping tasks traditionally completed in Xero
- The NLP models were trained on Plaid data and Xero classifications
- Using Kortical’s platform they were able to build a Machine Learning model in 2 weeks that was 95% accurate
- With the deployment functionality in the platform, end to end it took 6 months from data to live AI production model
Kortical AI as a Service Platform
The Accountancy Cloud team chose Kortical as the platform ensured that they would get quality results, with an ML model that they could deploy quickly as they wanted a tool that would enable them to work at the pace that they are used to working at. Kortical also provides data science support and strategic advice, which they also leveraged to ensure the best quality ML outputs at the end.
The first thing is the data - historical Plaid transactions with matched Xero account codes were uploaded to Kortical. The platform automatically:
- Cleaned the data
- Created the NLP embeddings
- Category encoding
- One Hot encoding
- Took 20% of the data for testing against
- Used AI to find the best model for that data
- Trained 1000’s of models
- Fully explainable AI models with high level and row by row explanations
- Created a fully scalable model ready for production via API
The results had to be the same or better than current methods of people applying those classifications and Kortical managed to achieve that: 98% accuracy
Production-ready Machine Learning
Having a great AI model is only part of the solution, the key is being able to integrate it into the application it needs to automate. This is where companies have found it hard to go from a great model to one that works at scale and live in production. However Kortical takes care of this process by making that model robust - with built in failover and redundency, and highly accessible via a REST API - with service level guarantees on uptime.
This meant that The Accountancy Cloud were able to go from data to live NLP model working in production in
6 months. Adding increased speed and consistency of service that only a machine learning model can achieve.
Kortical’s platform does not end with predictions, it is also learning with new data and as new clients are onboarded. This ensures that the models do not degrade but also that they can also improve over time, guaranteeing the highest quality machine learning at any given point in time.
This kind of success has been repeated with many different companies and different datasets, with not only great models but great ML that can be embedded into an organisations’ systems quickly to get value from AI fast.
If any of these points interest you, do get in touch with us via the form below. We would love to hear from you.
Get In Touch
Whether you're just starting your AI journey or looking for support in improving your existing delivery capability, please reach out.