How AI and Workforce Management work together?

Mar 28, 2023

Workforce management has been operating in the cloud for ten years; over that time, software vendors have collected a lot of data. This equates to billions of shifts with a lot of associated business rules and data. This data is ideal for developing new functionality to automate many laborious tasks that managers find tedious or difficult to get right. 

In this blog article, we discuss how AI and Workforce Management work together and outline ten use cases for Ai in workforce management. These all replace tasks previously done semi-manually, in spreadsheets or with non-AI algorithms. In 2023 all of these are live and in use somewhere and outperforming the old method and helping organisations increase revenue and save time and money. 

What is the relationship between AI and a WFM platform, and what part does each play?

A workforce management system includes everything needed to complete the task that the AI is being asked to do but often requires manual intervention and decision-making by a manager. When adding AI, there is no need to replicate any of the components of the workforce management software, which acts as a repository for the data and business logic related to workforce management. WFM components include:

  • Business logic: Such as Pay rules, leave and compliance rules.
  • Data: Including staff contract details, locations, and budget levels.
  • Processes: Such as leave approval and overtime confirmation
  • User Interfaces: Including staff app, clock-in app and managers console. 
  • Reporting: Such as Actual v Budget, under contracted, outstanding leave.

The AI needs to automate human processes and do it quicker and to a higher quality, outperforming the key metrics of an experienced manager.

How AI and Workforce Management work together

“The AI” is actually a collection of “APi first” tools and processes that together sort and analyse data, make decisions, optimise and then load that data back into the relevant part of the workforce management system. Those tools and processes include:

  • Data Engineering: These tools extract, transform and load data to enable the AI to be trained. They might be used to calculate the data they extract so that the AI can be trained, for instance, by calculating the cost of a shift, based on pay rate and hours worked to make it easier for the AI to learn about costs.
  • Robotic Process Automation: A series of “software bots” that collect and sort data replacing human managers, such as extracting data from spreadsheets or web pages to support data engineering.
  • Machine Learning and AI models: These models are often publically available via cloud-based suppliers such as AWS and Google, and your software vendor will have developed a few specialist models of their own. The AI is trained on the data that has been engineered for it using many models. Once training is complete, another algorithm is used to pick the best model, which is used to create the output. In some of the use cases, multiple AIs and models are used, and a number of stages of data engineering and training are required to create the output. 

Using these tools allows bespoke AIs to be run for each organisation and each department. Platform-wide AI’s do not train AIs just on the specific data required for your unique store or department. To do this, the AI needs to be trained on the data that relates to that store or department, which makes the data engineering aspect of AI’s in workforce management much more important if we want great results.

The 3 big use cases for AI in WFM – “right people, right place, right time”

When AI emerged as a viable option to automate tasks and decisions in a workforce management system, demand-led scheduling was the most obvious first choice. Workforce management platforms and employee scheduling software nearly always included a forecasting tool, a demand module and some sort of auto scheduling algorithm. The issue has always been that the forecasts were inaccurate, and the schedules often weren’t popular with staff, so more than 50% were manually corrected to keep staff happy. These three tools are now all available using AI. They comfortably outperform a human manager such that organisations that use them are reporting increases in sales by 3%-5% and low percentages of manual changes required by a manager. These three AI use cases really do ensure that you have the right people working in the right place at the right time:

1. Forecasting: Using AI to predict sales by revenue, item and transactions, as well as footfall and other metrics which correlate to staffing levels, is a key WFM function. Using historical sales data, online orders, reservations, plus 3rd party data such as weather and sporting events allows dedicated forecasts per department per week. Across a multi-site retailer, there are requirements for 1,000’s forecasts a week, and all bespoke to each department or store. Only AI can deliver the required accuracy at this kind of scale.

2. Demand: Once you know how much of something you are going to sell, an AI can work out how many staff of what role you need every hour. When plotted on a curve, it is called a demand curve, and it looks something like this.

A

labour demand forecasting for WFM

This shows the number of staff per hour and per day. The demand AI that creates this curve considers historical staffing levels for different revenue levels and data such as preparation time. 

3. AI Auto Scheduling: Auto scheduling which performs better than an experienced manager, has always been a target of workforce management software vendors. The auto-scheduling tool has to consider many factors, such as demand, staff preferences and compliance (we won’t go into them all here because we have written a separate article on that subject). The completed schedule or rota should offer shifts that staff are happy to work and which ensure customer demand is met and revenue increases. The image below shows two seven-day demand curves for the same rota. The top one was scheduled by a human manager and the bottom one by an AA. The colour coding is designed to show managers where too few or too many staff are scheduled with green being an exact match to demand. You can see at a glance that the AI meets demand more often than a human.

ai schedule versus human schedule

7 further use cases for AI in Workforce Management

As AI has started to deliver real benefits for organisations based on demand-led scheduling, people started asking what else AI and Machine Learning can do to improve workforce management. Here are 7 more great use cases that are all available now from WFM vendors that will automate tasks and decisions in your workforce management system:

1. Retention: Using the data collected in a WFM system, an AI can be trained to predict how likely someone is to leave a company in the next 30, 60 or 90 days and the most likely reasons. Armed with this information, an HR manager can intervene and work through staff issues to improve staff churn. 

2. Task allocation: In this use case, tasks are work elements that are worked within a shift. Take a supermarket; an employee could work an 8-hour shift and during this 8 hours work the checkout, stack shelves, fulfil internet orders or prep sandwiches for the deli. The typical supermarket has around 70 such tasks, and they are often time and skill sensitive. In fact, the auto-scheduling AI needs to be clever enough to ensure there are enough people on site to cover each task when required. Tasks can be a form of demand – if you know you need 3 people to unload a delivery and what time it is coming in, it is no different than needing 3 staff to deal with customers at the same time. With the task allocation, the Ai recommends who does what task throughout the day, a sort of auto-schedule within an auto-scheduled day.

3. Break allocation: A simple one for an AI but a time-saving tool for managers that ensures fairness and compliance with working time directive, allocating breaks to staff during their shifts.

4. Budget creation: Creating a staffing budget is a difficult task often considered “a chicken and egg” problem. As discussed above, the best way to calculate the correct number of staff is a mixture of demand forecasting and task-based scheduling. However, companies often set a budget a year in advance for each department’s staff. Yet staff in some organisations can often be effectively a cost of sale, i.e. they are directly proportional to the sales volume, and if you lower the staffing level, you can lower the sales. So you can see how setting the wrong budget and sticking to it could lead to lower sales. This AI takes a demand forecast – even based on a 52-week AI sales forecast- and then calculates the optimal spending level for that demand forecast. That way, the business gets a more accurate forecast of how much it is likely to spend. The 52-week forecast, however, is not ideal because it is too far in advance to consider factors such as weather and changes in the economy. This Ai can be re-run every week to cover the forecasts that are driving demand and update the expected salary cost.

5. Timesheet approval: This AI approves lateness and overtime against pre-set criteria and reduces the number of exceptions referred to a manager (this kind of AI can also approve leave requests).

6. Resource allocation: superficially similar to demand, however this AI uses completely different modelling techniques and is used to optimise shop opening hours and other resources that are profit sensitive. For shop opening hours, the AI will produce a matrix of all possible shop opening hours and using a forecast of “possible sales”, which isn’t limited to existing opening hours, the AI will return a revenue forecast for each hour, highlighting which hours could be profitable.

7. Holiday allocation:  Not a regular use case by any means, but we have built customers an AI to allocate and schedule unused leave that hasn’t been taken by staff after a deadline. It is an auto-schedule for unused leave, and the AI is much fairer than a manager.

Conclusion

AI is automating many of the tasks previously done via spreadsheets or manual processes, and this list of ten use cases is only the starting point for how AI and Workforce Management work together. If you would like to know more about Ai, Machine learning and workforce management contact us for a demonstration of any of these use cases.

Related articles

What is Workforce Management software?
How does Workforce Management software work?
What are the types of workforce management software?
How much does Workforce Management software cost?
What are the 10 biggest issues when implementing WFM software and how do you address them?

What To Know More About AI and Workforce Management?

Get in touch with us today for a quick exploratory chat with one of our consultants.

solutions circle

Share:

Want to more about this topic or would like to speak to us regarding anything Workforce Management related? Just get in touch with us below and we'd be happy to help.

Latest Blog Posts

Loading...