Model, determine and monitor the sentiment of customers across touch-points to identify general trends of satisfaction

Using cameras in, for example, retail sites to categorise potential customer's demographic data - typically sex and age. This potentially raises customer response, ethical and legal issues, unless carefully managed.

Determine the likelihood of reactivation and with what techniques.

Measure and better predict the effectiveness of different techniques to cross and up-sell (e.g. sales, marketing, merchandising) based on historic purchasing and browsing behaviour along with cohort analysis

Determine the likely lifetime value of customers to allow better customer management and engagement to increase revenues. It will allow a firm to better target its marketing resources to ensure that customer acquisition costs are managed appropriately by relevant segment.

Measure and better predict the characteristics of target customers for different products and service offerings.

Personalise loyalty programs and promotional offerings to individual customers - typically this will involve balancing availability of time-sensitive supplier capacity (e.g. airline seats) with customer demand and preference profile

Measure and better predict the characteristics of churners to allow for preventative retention actions. Identify most effecive retention actions by segment.

Enhance business relationship management through generating customer insight and personalised recommendations (improved information flow to operative through ML analysis / personalisation of product and proposition / recommendation - potential suppliers include Accenture, Squirro)

Measuring and better predicting the effectiveness of different sales and marketing techniques should lead to improved customer conversion metrics.

Identify most profitable consumers through data analytics

Automating the collection of customer data including product preferences, demographics, and personal information. This data may be collected under the guise of a loyalty program or app, with AI being used to extract the relevant information.

Capturing data on key life state events - from ageing or birthdays to more profound events such as house moving or bereavement - enables companies to target specific products or services. This data can be uncovered from multiple sources (within relevant legal constraints) and then the running of specific campaigns can be automated.

Automate sales voice conversations through a conversational agent enabling potential for simplified, high-volume sales. If inbound this can be a valuable consumer experience reducing waiting time but outbound and unsolicited it can be legally questionable (and a most unpleasant experience).

Automate sales text conversations through a chatbot enabling potential for simplified, high-volume sales

Emails that are tailored to individual behaviour

Leverage machine vision to tag brand-related and product images taking into account users' preferences and relevant media channels.

Enable Augmented Reality (AR) sales propositions by geo-locating product against a moving image - often generated via a camera phone. The AI places the product in the image - for example furniture in a room or make-up on a face. This may often be a 3D image.

In the race between online marketers and people keen to avoid being marketed to neural networks have been deployed to subvert ad blocking software (by, for example, working out which images might avoid being blocked). Ad blocking software has increasingly deep penetration amongst consumers which poses a challenge to online publishers. Subverting ad blockers is unlikely to be the last lap in this particular race.

Predict email open rates and click rates

Use customer behaviour pattern data to predict and recommend media content that customers will want to consume. This may be further optimised to business imperative - for example what will enhance likelihood of subscription renewal rather than instantaneous gratification.

Optimise online search results - possibly the most frequently experienced form of AI for consumers. Frequently the optimisation is focused on maximising commercial marketing returns.

One of the fastest ways to build revenue is to increase average shopping cart size and to cross-sell to exisiting customers with whom you have a relationship, data and a brand affinity. Also called recommendation engine or recommendation system, these systems leverage customer data to ensure that existing customers purchase more.

Retargeting customers who have already expressed interest in products or services. Typically the drop-out rate whilst filling in personal details (e.g. payment card information) is high - but these are the most valuable of convertible leads for most online organisations.

Optimise website experience and engagement to improve online shopping cart stickiness and to maximise wallet size. Typically this works best at scale and using enabled data analytics platforms.

Personalise search results for individual's preferences. A key question is who sets these - ideally it is the individual's needs that do so rather than the need to optimise the commercial serving of advertising.

Net Promoter Score is one technique to measure way of measuring the likelihood that individuals will act as advocates for a given brand - and building them up is one critical way to increase sales and reach exponentially through advocacy and referrals - but not all indivuduals have the same newtwork impact. Using AI to help identify, reach and grow these network nodes can help accelerate increasing customer growth via personal connections. Note that paid influencer networks (for example via Instagram) face ongoing challenges proving their relevance and minimising "fake follower" issues - as well as potential brand risk issues.

Delivery of personalised, individual messaging based on individual customer’s real time and historic behaviour.

Building subscriber income is a key part of many media publisher strategies - especially as print disappears, online advertising pricing pressure increases and 3rd party traffic generators like social media sites or search engines change their algorithms at will. Whilst these conversion rates are typically in low single percentage points any increase can drive significant revenue to the bottom line. Typically the models will score customer propensity to respond to different types of advertising and commercial propositions, perhaps at different times, and offer these up accordingly. For many sites this will also drive the cut-off moment for free access. This is a shift to a more flexible variant on traditional Freemium (some content is free, some paid for), Metered (a certain amount is free) or Hard (everything is charged for) models.

Understanding the link between digital spending and real world reactions by consumers - in this case seeing an online ad and then going in to a store - is one of the holy grails of advertising. Machine learning makes the inherent data scale and connections behind this manageable. However, getting the right partnerships and data capture approach as well as consumer acceptance and data privacy right will be the key to long term commercialisation.