Best Practice AI Use Cases

Identify root causes for low product yield (for example input, tool or machine specific issues) in manufacturing. This analytical work will likely require working closely with human analysts.

Context helps with marketing. So emotional states triggered by relevant content will impact on marketing effectiveness - so football on TV attracts specific product categories for example. What should go where can be micro-optimised using AI.

Combine product information from disparate sources, structuring the information centrally for internal use to improve product discoverability and supplement decision-making.

Automate building systems to reduce energy costs through AI-powered adaptive temperature and energy control. The challenge is to justify the investment costs against actual energy savings.

Optimise sales resource assignments, product mix, commissions modelling and accounts

Visually recreating people in detail to show them moving and/or speaking in video. The application can have positive results, such as bring artists who have passed away back in life in a digital form, but also negative ones, such as deepfakes, who have been used for spreading fake news.

Provide access to digital information to illiterate folk through conversational agent tools / chatbots. Voice interaction enables new forms of communication - for example low / free marginal cost access to healthcare advice or other information. Access to digital devices, Internet and electricity may be factors limiting deployment.

Identify performance and risk for pilots through flying patterns and other data. Similar analysis can be done on road drivers although the data is inevitably more likely to be complicated by ground-level interactions that would be more traceable in the air (e.g. other vehicular movement).

Predicting drug demand in different geographies for different products. This will potentially be driven by factors from disease outbreak vectors through to socio-economic indicators or even media-driven consumption demand trends.

Predict energy demand trends based on data sources ranging from sensors to social media. In a competitive market the more varied and unique the data sources accessed the more refined (if not always more accurate) the model can become.

Analyse social, traffic data to measure potential impact of new building developments. Modeled outcomes might indicate hazards (e.g. fire evacuation risks) or design features for consideration (e.g. siting of facilities etc)

Employees in conversation record an account of their concerns and issues which is automatically captured and transcribed. HR gets a detailed report and can conduct follow-up interviews through the bot. Potentially, employees can stay anonymous if they so choose.

Detect potentially fraudulent or nefarious users (e.g. individuals under sanctions, investigation etc) through pattern matching of structured and unstructured data on transactions from different sources, such as phone numbers, addresses, company directors and news reports (HSBC, Quantexa, Silent Eight)

Streamline insurance application processes - examples would include pre-populating forms or automatically linking with relevant information sources (e.g. public registers etc).

Automate aggregation of personalised property listings for real estate customers

Identify drug compounds with current regulatory approval which could be used in new ways to treat other conditions. Machine learning assists by searching through existing research literature for known and inferred relationships

Assist physicians by providing the latest medical information and research to inform proper patient care - this may be from specialist providers and cover the sorts of rare ailments that medical staff may not encounter on a regular basis

Detecting, identifying and monitoring animal populations is especially useful when they are in the wild and potentially at risk. Typically it requires a network of static camera traps or acoustic recorders but can also be extended to include mobile sensors.

Software which automates the work of human moderators to identify abuse and misconduct in online messages between video game players on community boards.

Use video, images or robots in the retail area to audit and analyse shelf space usage. Optimising shelf space drives retail sales, can reduce risk of running out of stock and it can also be use to prioritise shrinkage risk areas.

Automate collection of banking data from non-standard documentation (e.g. in shipping paperwork) - this will typically be turned in to standardised banking documentation

Successful social media sites tend to have more content available for users than all but the most addicted (and socially restricted) could possibly read. This means that they have to use multiple indicators to power the algorithms that rank the content displayed. In this they are in a constant battle to stay one step ahead of the content marketeers trying to game the system.

Construct detailed map of farm characteristics based on aerial video or image capture. This can be of assistance for a series of tasks ranging from planning to valuation to deployment of automated farming tools.

Create more personalised insurance pricing based on actual monitored customer behaviour. Examples would include 'black box' functionality in the car or wearable sensors on the body . Essentially the customer agrees to swap hitherto private data in return for what they expect to be lower pricing as the insurer can better measure real risk - and potentially the exact circumstances around an insurance claim.

Create contact lists for outbound communication through automated web scraping (e.g. of social network traffic, databases or websites) or filtering based on response and sourcing analysis. Note that this is increasingly encountering regulatory challenge (eg GDPR in the EU).

Conduct "digital human" conversational interface with customers in physical locations (e.g. banks such as Natwest, soul machines). The aim is to provide effectively online functionality via some form of automated, robotic interface.

Identify target patient subgroups that are underserved or not diagnosed to support creation / deployment of a mitigation strategy

Students expressions and movements are analysed to check that they are paying attention in class. The system will be able to tell if students are reading or listening – or napping at their desks. Students will get a real-time attentiveness score, which will be shown to their teacher on a screen.

Replicate (and replace) human-made decisions in control room environments to reduce cost and human error

Modify driver experience based on emotion tracking through facial recognition system. This will for example lead to different music being played, lighting set up and driving advice offered.