Screenshots of an actual application interface at https://www.prprophet.ai
Prophet’s SAAS is targeted towards public relations specialists that are looking for journalists who could publish their news. With old platform features, the product failed to provide relevant results so the users did not convert to paid customers after the trial period. Prophet’s leadership recognized two major high-priority business requirements:
Provide superior matching capabilities with AI for the users;
Drastically improve the platform’s performance;
The product version 1.0 was built as a POC. Naturally when the development backlog started shifting towards a more complex and elaborate version 2.0, the original team struggled to keep up with increasing technical debt around the most core components.
This in turn caused the product feature release schedules to fall behind. And for a fast growing startup, that had to meet the market, investor and management expectations this wasn’t a situation that would somehow sort itself out.
When Horion’s tech team analysed PRophet’s development backlog and code base key patterns emerged:
Features that were planned ahead weren’t possible on their current MVP code base. Code refinement and refracturing had to be done;
On monolithic system it wasn’t possible to accommodate new features, so microservicing was needed;
Product features such as matching algorithms weren’t functioning as expected;
Release schedules were very short, and development team could not keep up with it;
Technical debt discovered after a thorough code review;
Performance issues were present;
No established data engineering standards and full use of data lake benefits;
Absence of data science standards. Data scientists weren't able to experiment with algorithms within the given infrastructure.
Saige, a current PRophet 2.0 product development provider partnered with Horion digital. Horion’s Senior solutions architect (SA) along with Back-end (BE), Front-end (FE) devs and DevOps engineers worked with Stagwell Tech product owners and project managers to develop a strategy on tackling the previously mentioned challenges.
We implemented AI models into the platform. It is capable of understanding the context of the user’s press release pitch.
We have used AI to contextualise any journalist and his/her articles.
Our solution identifies best suited journalists for the user’s pitch. It is then able to generate a journalist-specific email to grab their attention. The AI model also explains why it chose this specific journalist.
We have made the product 6 times faster by refracturing the code and drastically reducing the technical debt.
Together with Saige we helped Prophet to become a truly AI driven product, with features that save hours of time for platform users. Every day with our solution the Prophet contextualises hundreds of thousands journalists and their articles to provide up to date information for the companies that are looking to publish their news and find the right journalists for that.
The following results were delivered with the cost of ownership and future proofness in mind:
Code refractured and split into separate modules and developed new features;
Replaced algorithmic matching with machine learning (ML) based matching;
Performance issues solved and customer experience improved with 6 times better loading times;
Implemented data lake architecture for big data computation;
With the correct infrastructure changes now data scientists can pop the hood and play with the algorithms using live data;
Introduced a delivery framework with standardisation and scalability in mind. CI/CD pipelines and code quality and structuring standards.