Our recent blog explored how FYLD’s machine learning, or artificial intelligence (AI), works in practice for field teams and how the platform’s predictive capabilities help managers and fieldworkers to proactively manage sites, teams and workflow. You can read that article here.
We’d like to go a little deeper and explain how simple it can be to maximise the machine learning element of FYLD to deliver meaningful benefits to your daily operations.
So how does machine learning work in FYLD?
At the heart of FYLD’s Visual Risk Assessment (VRA) process is the AI-driven engine that is continuously learning from each interaction with the platform.
Every time a FYLD user accepts or modifies a recommendation generated by the platform, the AI engine learns from that interaction and adds that information to its operating model. Accepting a FYLD generated hazard or control measure reaffirms the validity of the model, conversely, when a user makes an adjustment or declines a suggestion, the AI model is refined. Little by little changes are made to the model and the platform’s suggestions become more accurate. Be reassured that it takes several iterations of learning to change the outcome of the model to prevent inaccurate results.
One of the powers of the platform is that when field workforces proactively engage with safety, their behaviour changes. For this reason, FYLD always presents users with hazards that they may not have identified, but that the algorithm thinks may be on site. This ensures the overall safety environment is thoroughly well managed.
Machine learning depends on an accumulation of data. Volume is therefore a key driver in how quickly the AI-model will be enhanced. One user performing one VRA per day will give 20 new data points in a working month. However, expanding the user base to 100 users will give 2,000 data points in the same time. A large number of users will enhance the AI-model much faster than a small sample size.
FYLD deployments commence with a minimum of 100 users to ensure that the model will learn from the interventions made by fieldworkers and remote managers rapidly, and so our customers see the benefits of FYLD.
FYLD’s AI engine enables the platform to combine multiple forms of rich data to generate suggested hazards and controls in response to visual risk assessments.
For example, a user may record a point-of-work visual risk assessment of their site to document a pump replacement job. The old pump is visible in the video, but the user fails to mention that this pump is a heavy object, which could cause injury when lifting. The AI-model may not generate heavy lifting as a risk for this job, and the user may or may not remember to add it manually.
If this scenario gets repeated several times, the AI-model will learn that pump replacements could involve heavy lifting, due to the subsequent logs of requiring supplementary equipment to assist. FYLD will then suggest this is a risk and may require additional support automatically, even though it was not mentioned in the video. In this way, FYLD will warn users of a potential risk that they failed to notice – adding great value to the video risk assessment and helping to prevent injuries to workers.
FYLD stands alone as an overarching digital solution for managing field operations and digitising paper-based risk assessments. Offering the tools to capture high quality, digital risk assessments, improve cross team communications, evidence work and increase site visibility for remote managers – FYLD is all backed and improved by its AI-engine.
To find out more, visit: https://www.fyld.ai/bespoke-solutions/