The 5 Business Values of AI - Part III of The Business Leaders Guide to AI

A weekly series of 5min reads on AI. No buzzwords. No hype. No fluff.

Last week, in Part II, we discussed how AI creates business value by optimizing any business decision. AI can't work miracles - there are no perfect answers in business - but AI can better inform any implicit trade-off or explicit business decision to generate more revenue, capture higher margins, or reduce expenses.

To further help make sense of AI's wide benefits, I've broken them down into 5 common types that apply to every company in every industry:

The 5 Business Values of AI

1. Grow Revenues by Increasing Personalisation
2. Grow Revenues by Creating New Products
3. Reduce Costs by Improving Business Processes
4. Reduce Costs by Automating Manual Tasks
5. Reduce Costs by Limiting Future Uncertainty

1. Grow Revenues by Increasing Personalisation

Personalised customer experiences are a key driver of conversion rates, average order sizes, and repeat business (retention). "Acquiring a new customer is anywhere from five to 25 times more expensive than retaining an existing one" (HBR), underlying the importance of personalisation for retention.

But personalisation requires learning a lot about a customer: how do they communicate, when are they looking to buy, what do they care about, why do they use your competitors, etc.

Traditionally, answering these questions would be prohibitively expensive and time consuming. But because AI can analyze much more information, brands can now deeply personalise their customers' experiences in a feasible, cost efficient manner.

Examples Include: personalised marketing campaigns, product recommendations, promotions, or even user experience (e.g. the order and arrangement of content on your LinkedIn feed), and more

Detailed Example: We helped one of Asia's largest grocery store chains deeply personalise their email marketing campaigns and monetise their customer loyalty card purchase data. Our Machine Learning algorithms analysed past purchase data, store location data, product data, and even calendar and recent weather data, to learn what, when, and how each customer liked to buy. Our ML algorithms then generated a personalised arrangement of products, at a personalised time, with a personalised promotion, for each and every customer. The result was a nearly 30x increase (3,000%) in email marketing conversion rates.

2. Grow Revenues by Creating New Products

New products are a key driver of new customers, overall sales, and market share. Yet creating the right new products requires anticipating industry trends, analysing customer feedback, and monitoring competitor activity - all combined with a dash of human creativity.

Traditionally, this means human merchandisers and designers try to consume and analyse more information than humanly possible - trade magazines, blogs, competitor websites, social media, and more... Yet with AI, human product creators can be superhuman.

AI can generate entirely new product designs, which human designers use as inspiration to further refine, or AI can identify trends that warrant investigation by human merchandisers. AI does what it does best (analyse lots and lots of data), and human experts are able to focus more of their time on the high-value, specialised and creative tasks they do best.

Examples Include: identify industry, competitor, and consumer preference trends, generate new sample designs or mockups, suggest product ideas, and more

Detailed Example: We worked with the world's largest health and beauty retailer and their eCommerce division to improve their new product creation. First, we implemented automated scripts to programmaticly source unstructured data from tens of thousands of sources such as leading blogs and social media influencers. Then we applied several Machine Learning techniques such as Natural Language Processing and Unsupervised Learning to identify and rank the top key themes such as "vegan supplements" and "CBD consumer goods" for human merchandisers to investigate. As a result, merchandisers were able to consider 100x-1,000x more data points before investing in new product development.

3. Reduce Costs by Improving Business Processes (Legacy Analytics)

In many business processes, from assortment planning to resource allocation to marketing, employees have leveraged data and analytics to make better decisions for years now.

But not all analytics are created equal. In fact, "two-thirds of the opportunities to use AI are in improving the performance of existing analytics use cases" (McKinsey). Many existing analytics are built on legacy technology, make over-simplifying assumptions (for example that factor A and factor B are linearly related), and are simply unable to account for the full range of circumstances and complexities involved.

Because AI techniques are more powerful, and can leverage more data types than most legacy analytics (eg. images, natural text, video, etc), AI can greatly improve the accuracy, flexibility, and usefulness of existing analytics and business processes.

Examples Include: optimizing customer service, marketing budgets, and resource allocation (logistics), better modeling real-world constraints in ERP decision, improving human employee efficiency or accuracy with intelligent nudges and suggestions , and more

Detailed Example: We helped a large multinational logistics provider, which delivers 100 million units of consumer products each day, to improve their fleet allocation and management. Their existing system involved human planners and legacy enterprise software and analytics. Due to limitations of the systems though, simplifying assumptions had to be accommodated, such as imposing discrete 'Zones' - this simplified the constants, but greatly reduced efficiency. Using various internal and external data sources, our ML solution enabled human planners to work with an AI's more accurate insights, resulting in a 40% reduction in route costs.

4. Reduce Costs by Automating Manual Tasks

How much time today did you spend on a new task you've never done before? How much time did you spend doing something you barely had to think about for the umpteenth time? For example you may have driven the same route to work, again, - did you think about every turn you had to make? No. (Likely, the same is true of organizing your email. Or resending that weekly report. Or some other work tasks...)

Every business has critical tasks that must get done to keep the lights on - but as critical as they may be, they don't require much mental effort to repeat. For example, it's expensive when sales staff isn't spending their time on selling, but on filling out forms, processing completed forms, and responding to boilerplate customer inquiries. It doesn't take much mental effort at all to decide to copy and paste from one application to another, but it can take a long time to repeat it again and again. And again...

Traditionally companies have had to pay high labor costs or accept high rates of human error. But with AI, clearly defined, manual tasks can be completed in less time, with lower cases of error, and at much lower cost. Well known Stanford AI researcher Andew Ng has said that virtually any small mental task can be automated using AI.

Examples Include: chatbots replying to customer inquiries, image recognition verifying identification or paperwork, anomaly detection flagging suspected fraud, and more.

Detailed Example: We were contracted by a leading financial custodian to reduce the overhead and costs of error in their customer on-boarding process. We designed several solutions which automated multiple tasks within identify verification, form processing, multiple data source triangulation, and more, cutting time to complete the on-boarding process in half.

5. Reduce Costs by Limiting Future Uncertainty

Planning for future uncertainty is one of the most expensive activities businesses engage in. Whether it's deciding when to do equipment maintenance, how much inventory to order, or where to invest in infrastructure, being overly aggressive or conservative can entirely wipe out margins and invite leaner competition to grab market share.

Traditionally companies have been forced to live with over-simplifying assumptions and high costs. Predicting the future is notoriously difficult. And while no AI can ever perfectly forecast the future, because AI can consider vastly more information and recognize more complex trends and patterns than humans, AI can play a hugely important role in reducing business uncertainty.

Examples Include: better managing inventory levels (inventory forecasting), demand planning, anticipating supply chain shocks, predictive maintenance, predicting future infrastructure strain, proactively identifying at risk customers (customer churn), and more.

Detailed Example: We worked with a large supply chain management company to reduce their expenses wasted on reserving excess container space (ie. excess inventory of shipping capacity). To mitigate future uncertainty, they were chronically reserving excess capacity, representing a substantial cost once aggregated across each segment and shipping lane. Using internal order data and external shipping lane and real-time forward pricing data, we used multiple supervised and unsupervised ML/AI techniques to reduce wasted expenses by more than two-thirds.

Looking Ahead

AI isn't magic or a silver bullet. Just like every other technology AI creates business value in specific ways to i) grow revenues, ii) increase margins, or iii) reduce costs.

And just as every modern business needs to share information via the internet to remain competitive, businesses in every industry increasingly need to analyze information via AI to work smarter, run leaner, and keep up with the competition.

Now that we've discussed AI's benefits from a business value lens, next week we'll look at AI's capabilities from a high level technological capability point of view to understand more concretely how AI helps. Don't worry, no technobabble or jargon needed.

-----

This is part of a weekly LinkedIn series, The Business Leaders Guide to AI. If you found this helpful or thought-provoking in any way I'd love to hear from you in the comments.

Even more importantly, if you have suggestions, topic requests, or found this unhelpful in any way I'd especially love to hear from you. Please drop a comment (was it too theoretical for you? too simplistic? too technical? not technical enough?) or, if you prefer, message me on LinkedIn.

Thanks for reading.