Innovation with AI in Uncertain Times

Why Innovation Matters More Than Ever

Innovation has always been tough but now it has got a whole lot tougher, however there are ways to be more effective at delivering AI/ML innovation so there is more buy-in from the start.

As we navigate new ways of working with different business demands and market uncertainty, the focus on innovation for many has taken a back seat. The paradox to this way of operating is that the companies that are keeping their innovation tracks in place will come out of this crisis all the stronger. Where once it was a luxury to innovate it is now going to be a necessity as there has been a seismic shift in the way businesses are operating and consumers are spending which will have repercussions for the long-term and only those companies adapting to that change will survive.

It is unsurprising that those with automation and optimisation artificial intelligence (AI) initiatives are the companies that have continued to invest in these as ultimately these projects will enable operational efficiencies that can be saved or those funds can be pumped into other areas of the business to see if through this season of change.

Often articles focus on the what and why of AI case studies but here we thought it would be more useful to focus on how to get an AI innovation project off the ground and prove the results so you can set yourself up for success from the start.

‍

Identifying a use case

Let’s take customer services and email responses as an example where machine learning can have a huge operational impact.

‍

‍

Areas 1 and 2 can be handled by Machine Learning solutions freeing up more resource to handle point 3 or requiring less overall resource in situations where it is not available.

‍

What Data Do You Need?

For this use case training data will be required in the form of incoming emails and the (correct) responses. For a proof of concept you will require a subset of these emails e.g. can we train a ML model to respond to questions about where is my delivery. Ideally for a good model, we will want 1000 examples of that question but we have been known to train a model on a lot less.

‍

Who Needs To Be Involved?

• Project Lead - could be you, the Innovation lead or the Domain Expert.
• Domain Expert - Customer Services Manager
• Data Scientist - if you do not have one Kortical has Data Scientists or we can train an experienced Data Analyst / Statistician
• Data Extractor / IT - someone who can get you the data
• Snr Stakeholder / Budget Owner - who owns this pain right now - Customer Services VP

‍

Proof of Concept (POC) Process

Four weeks is all you need if you deliver an AI/ML project with AutoML which goes from raw data to data cleaning, 1000’s of models tried and tested, full explainability, a production-ready ML model to be implemented at the end of the POC. Along with a business case and insights to help drive your business forward.

‍

What You'll Get at the End

You will know that you can automate x% of emails and you can then build the business case around that e.g. when did an email automation solution for a UK bank we were able to automate 57% of the emails that came with a 95% accuracy which meant a saving of £750k p.a. And being able to respond to customers 3 x faster.

Thus the innovation project goes from an idea to a clear path to delivering true ROI for the business and having a business case means senior stakeholders can quickly sign off the pilot as there is a clear benefit and plan laid out for optimisation.

‍

Other High Impacy AI Use Cases

Efficiency and optimisation are the key AI areas that we are seeing getting the most traction right now from logistics and inventory forecasting to free up working capital, to AI driven marketing for personalised content and offers for higher conversion rates.

Get in touch with us if you would like to find out more about how Kortical can deliver innovation that makes an impact, with you.