In this article, we discuss how does a SolvedBy.Ai proof of concept work and answer some of the common questions we get about the subject.
SolvedBy.Ai has a number of AI-powered products that can help organisations obtain more value from the data they have. We started in HR or WorkTech and have products that predict the likelihood of someone leaving and build the optimal rota. However, to deliver some of the products, such as our rota Scheduling.Ai, we needed accurate inputs that relied less on our HR skills. For instance, we had to get very good at forecasting sales and other factors that influence staffing levels, and we had to do it for thousands of different stores in 15-minute segments for a two-week period in advance.
So we developed more AI products and more core AI competencies. And because we are dealing with multiple different organisations, there was an element of bespoke work that each customer wanted from our AI products. When speaking to potential customers, one of the first questions we would get asked is often a version of “how do I know your AI will work with my data.”
So to reduce the implementation risk for the customer, we came up with a proof of concept approach that allows customers to see a bespoke version of our product work with their data.
What goes into a SolvedBy.Ai proof of concept?
There are several phases in a SolvedBy.Ai POC. These include:
- Discovery phase
- Accessing data
- Cleaning and processing the data ready for training the AI
- Training multiple (often 1,000s) AIs
- Marking the AIs using test data and selecting the most successful one
- Reporting on progress, obtaining additional data sets and retraining
- Producing a final report
What types of AI can SolvedBy.Ai work with for a proof of concept?
We are happy to explore any use case that uses our existing AI pipelines and falls within our core competencies. If you have an AI that you think we can help you with, try us, and we will soon tell you if it is “above our pay grade.”
What are the SolvedBy.Ai core competencies?
Our data science team have a lot of experience and skills in the following areas:
- Optimisation and resource allocation
- Predictive modelling
- Robotic process automation
- Data engineering
What happens if the POC is a success?
Usually, we do a proof of concept with a specific aim, ideally providing the final AI throughout an organisation. For example, if we were proving our ability to build accurate sales forecasts, we would be doing so with the intention of providing a production solution that automates the creation of forecasts for every department or venue within an organisation every week and delivers it to wherever it is needed.
How much does a proof of concept cost?
The cost depends on the complexity of the implementation and the size of the proof of concept. The typical cost ranges from £5,000 – £30,000. A forecast proof of concept where we already have an integration working with one or two of the data providers will be on the lower end of the scale. A full auto-scheduling POC where we need to create forecasts from your data and extract business logic from a new workforce management software will likely be on the higher end of the scale.
Does SolvedBy.Ai provide free proof of concept?
We don’t offer a free proof of concept. There are costs involved; these include access to our most valuable resource, our PHD and MSC data scientists, plus cloud server costs as we experiment with hundreds of different AIs. Both parties benefit from doing the POC, so we always ask for some payment to cover some of our costs and ensure both parties have some skin in the game.
We don’t profit from a POC, we need long-term contracts to make a profit, so we are highly incentivised to make the POC a success.
What happens if the pilot is unsuccessful?
The customer has the right to terminate the contract and walk away. No money is refunded, and both SolvedBy.Ai and the customer would have had a commercial loss. However, the customer would have learned a lot about the quality of their data, what is possible with that data and possibly have a road map to improve their data collection and management further to take advantage of AI in the future. The customer will also have avoided signing a long-term contract for an AI solution that does not meet their needs.