We have been speaking recently with the team at RosterLab (https://www.rosterlab.com/) and they are very keen to hear from those of you working in the NZ healthcare system involved in the rostering process. I have included the summary of their tech below so let me know if you are interested in connecting!
Rosterlab has developed an AI and web app so that you no longer have to make Rosters on Excel. The AI creates rosters faster than people can, saving 90% of the admin time involved. Not only that but it can create better rosters than any human can, leading to better staffing numbers and skills coverage. The AI considers all the nuances of rostering such as availability, preferences, and skills so there is no compromise, only benefits for the users. In considering all preferences Rosterlab believes we can provide better rosters for the staff too, leading to happier employees and lower turnover. We have been helping many nursing and doctors’ units to produce their rosters. We are looking for more people who are facing the rostering problem and are willing to trial, and hopefully adopt AI rostering.
Happy to echo your endorsement. We at ADHB have engaged with them as well and @sarv is helping lead validation studies with some wards at ACH. From what I’ve witnessed, the RosterLab solution really stacks up, the UI is simple and easy to use and there are integration options*. I’ve recorded a session that I’ll post up here soon. We are cognizant that Mahi ē Taea (a UKG Workforce Dimensions product) is also being rolled out and it’s important there’s alignment.
*Conceptually we’ve looked at RPA (and/or embedding a RESTFUL API into RPA) for fast/cheap integration and workflow automation i.e. taking org data from WFC, passing these to RosterLab, retrieving the roster data, and consuming this in WFC.
Several years back I put together a rostering solution for our local anaesthetic department (Dunedin). Was classic shadow IT, and predictably fell over approx 2 years after I left - and to my delight was replaced with a ‘real IT’ solution.
One thing that struck me whilst doing that was just how complex rostering is. Some bits were easy to automate (e.g. document presentation and communication), but much was not and required some seriously high level domain and interpersonal knowledge and communiation skills to manage effectively. The data to train AI to perform that task simply does not exist, at least not in a busy tertiary anaesthetic unit!!! I’m sceptical of their claims for anything but the most general of rosters.
Fair points and I initially thought they used supervised learning too (and hence a bulk of training data!), but I’ve reconfirmed:
It’s a mix of unsupervised learning and integer linear programming. We don’t need any external data because we don’t believe the data we are looking for exists, as all health rostering in NZ appears to be done manually.
@sarv has since heard back from SLT and given there’s already a rostering product in tow, there’s little appetite to fund a validation study for another - this is understandable but I’d hoped there are other DHBs that would put RosterLab up to the challenge! $16 per roster is what they quoted us.
I encourage others to think about meeting with RosterLab to see if it is something your organisation might be interested in looking at. Great people and great work.