As software consultants, we spend our days on site with customers listening to their problems so we can try and make their lives just a little bit better. Our current clients include customers in central Government, telecoms, banking, recruitment and siftie is based on problems that we’ve heard from customers in each of these industries.
What it does
Our remit for Siftie is to try and automate as much mundane work as possible leveraging Monday.com. Our initial focus is on matching, as it’s a common problem for our customers. They all have some processes that take a massive amount of man hours and just require people to read and check for keywords. This is a ubiquitous problem across industries.
A lot of these processes already lived in Monday.com, Kayleigh’s(who’ll you see in the video) team already managed their workflow in Monday.com so siftie was an obvious match. John Dunnet (we had to cut his interview from the video for time) who runs multi million pound bid tender processes for the government was not using Monday.com, but once we showed him what was possible with siftie and Monday.com he signed up and is now a paying customer.
The hard part is that data can live in lots of different formats, so we needed to use AI methods to pull out content from a variety of file types including, docx, PDF and images. We then needed to make sense of all this content, so we run AI natural language processing on the content from the documents to understand the context of the data and to build insights to help our users.
All of this data is fed back into Monday.com to use for further enhancements like giving a matching score for CV requirements and enabling advanced search of the contents of these documents without ever leaving Monday.com. This enables new intelligent workloads that were never possible before.
We’re user centric design experts, accessibility is core to everything we do, so we really wanted the tools to enable inclusivity and enable people with enhanced accessibility needs to do work that was not possible before, step 1 for this was to pull contents out of documents and images, but we also use AI machine vision techniques to explain what's in an image, further enhancing the accessibility of using a monday board as important context is no longer hidden away in images and documents.
How I built it
We utilised React and Redux to build our UI frontend service and we built a NodeJS backend with Express framework.
The backend is responsible for managing monday.com app subscriptions. To process a file, Siftie will first check the files format as each format has different processing steps. If text cant be directly extracted from a document, Siftie will use machine vision to perform ocr to get the transcription. Then we cache the document in a massively scalable index and we return the transcribed content back to monday.com via the integration sentence.
And lastly Siftie uses unique natural language processing to tag the transcription to facilitate more insightful searching/filtering.
Challenges I ran into
The main problem we ran into was breaking it down into a real MVP that would have enough functionality to be able to drive some of the powerful workflow features that we wanted to drive. The problem is conceptually the same across industries but what you do with the data that you unlock is extremely varied. We’re really proud that we were able to deliver something that demonstrates this well in the time.
Accomplishments that I'm proud of
With all the software we build, inclusivity is essential and Siftie is no different. The additions we have made for workflow is “game changing” as our users have stated, but our accessibility enhancements can be life changing for our users with accessibility needs as it gives them access to jobs that they would have been denied access to before. We have tested the product with people with accessibility needs with really positive results, the users were not comfortable being in the video so we have not included their interviews.
We’d be interested in working with Monday.com to integrate our accessibility features into Monday.com core so that more users can have access to these benefits by default.
What I learned
This experience has been really enlightening, we started out with a concept that was quite simple but in implementation these problems are really hard to solve which is why they require human engagement to solve.
Our goal is never to replace people’s jobs, but to enhance their workflow so that they can focus on the parts they like, we’ve been extremely successful in automating large chunks of mundane boring work, but there's so much more tools that we can build to further automate similar workloads.
We also learned that a lot of the functions we’ve been working on make a lot of sense to be part of monday.com’s core services, unlocking the power of people’s data brings incredible enhancements to Monday.com and would give Monday.com a strong competitive advantage. We’d be interested in discussing helping make this a reality.
It’s a great problem to solve and the improvements we can bring to our customers are incredible. We’re really excited to continue working on Siftie.
What's next for Siftie
Step one was always to expose data that was locked away in files and use them to showcase the kinds of functionalities and insights they could drive. The next step for us is to double down on our initial insights and start to provide real intelligence in insight and use these insights to drive process improvement automatically. We want the system to start proactively giving suggestions on what it’s learned from the documents and how users can improve their workflows. Our eventual goal is always to kill all of the mundane work like CV sifting and bid marking that people suffer through day to day and to enable our customers to focus on the parts they love about their jobs.
We are really pleased with where we got to in the end, we managed to prove our concept, there’s so much that can be done to further improve these workloads. We have a roadmap of enhancements we want to build that will further improve support and understanding of data for Monday.com unlocking the potential for Monday.com’s users.
We thought Siftie would be best suited to the Marketing and Creative category as our primary use case is advertising recruitment, but some aspects of Siftie could potentially be useful to the CRM category. Noting here as we were advised to do so by a competition manager on the discussion board.