AI first - going beyond the first steps

A growing number of businesses are seeing the benefits of using artificial intelligence (AI). But many are simply implementing isolated pilot projects on an ad-hoc basis. The lack of a clear AI strategy may prove costly.

This insight builds on an article published earlier this year in disruption magazine by Tariq Khatri

Developing an AI strategy requires acknowledging the current limitations of AI as well as its strengths to identify where one can and cannot (or should not) exploit it.

This article is about the 'what' of an AI strategy rather than the equally important 'how'. We look at the various business areas where AI solutions are having an impact, try to characterise the boundary line of its applicability and talk to some of the areas of research that may shift that boundary to bring more of corporate business and operating models within the scope of AI solutions.

Room for opportunity

First, some terminology. We prefer 'machine learning' over 'AI' since what the machine learning community calls 'learning' generally means simply fitting a model to data. The remarkable achievements of machine learning (for example in real-time language translation) are all founded on the basic principle of probabilistic pattern recognition. Appreciating this becomes helpful to understanding where it can be used.

Researchers are busy extending the performance and scope of individual machine learning techniques. For example, natural language processing is now at or close to human level performance for language translation across a range of language pairs, but generating anything but the simplest conversation with a human is still a work in progress. While it is important to keep abreast of these specific developments (the Electronic Frontier Foundation tracks progress against a wide range of benchmark tests), considering the bigger picture can be useful.

Figure 1: 

Machine learning takes over where rules based approaches break down:

  • Excessive quantity and/or complexity of predictive variables
  • Rigidity of rules inadequate

Figure 1 considers the cost of making a mistaken decision against how different future experience is likely to be from past experience. Rules-based systems operate in the darkest shaded region – where the future is expected to be so similar to the past that simple rules will hold true for most circumstances. 

Machine learning takes over where the rigidity of these rules breaks down. But it can struggle if the future task environment differs too much from the one upon which the algorithm is based. The vertical axis is important as well. It’s not the end of the world if peanuts are mistakenly ordered when a customer has requested cheese through a voice-controlled grocery app (unless they have a nut allergy). Mistakenly telling someone that they do not have a life-threatening medical condition when in fact they do is more material.

As we move up the vertical axis, machine-learning deployments must rely more on the ability to exception manage – to default to human judgement when the machine declares failure. In some real-time, customer-facing, high-risk situations this may not be feasible. Businesses are exploiting machine learning in five different ways (see Figure 2):

Figure 2: Five areas where corporates are exploiting machine learning

1 2 3 4 5
Enhanced Prediction Automation New Propositions Commercialisation Disruptive Models

Using ML models to support mgmt. decision making, e.g.

  • Lead generation
  • Inventory management
  • Digital spend allocation
  • Credit risk decisioning

Company Examples:




Zest Finance



  • Adverse media screening
  • Contract information extraction

Or Revisionist

  • Swarmed robot warehouse automation

Company Examples:

Slaughter and May

Link laters

JP Morgan


New customer propositions made possible by ML, e.g.

  • Heavy equipment management
  • Precision agriculture
  • Intelligent receivables matching
  • Conversational banking

Company Examples:


John Deere

Bank of America


Of valuable proprietary datasets

  • Of valuable proprietary datasets

Of new techniques/solutions

  • e.g., video compression

Company Examples:


Orbital Insight

Twitter and Magic Pony Technology

New competitive models, e.g.

  • "perception- reality" arbitrage
  • Proactive asset gathering
  • Proactive asset gathering

Company Examples:

Accelerated drug discovery

Open Door

Benevolent AI

 1 Enhanced 'core business' prediction

Predictive analytics has been used for years to support many aspects of business decision-making – particularly in marketing and risk management.

Machine learning’s can accommodate vastly greater numbers of predictive variables and variable relationships, across both structured and unstructured data (eg text, voice) alike, and improve predictive power as new data is received. This means predictive analytics is being applied in new areas of business, such as evaluating the level of compliance risk posed by specific client interactions).

The intelligent enablement of (even formerly non-IP addressable) physical devices and equipment is bringing active end-to-end flow optimisation, fault prediction and root cause analysis to great swaths of heavy and light industry operations that hitherto have been operations management blind-spots. Business functions that historically may have had little ex-ante predictive analytics deployed for management purposes are now firmly within scope (eg supply chain management, finance and risk operations, health and safety, etc.).

2 Automation

At least three types of automation solutions are being deployed,

Extractive -  the automated extraction of structured information from unstructured sources. For example, automated searches for adverse media reports assessing new client risk.

Orchestrative - automating processes or activities where simple rules-based approaches break down. An example is automatically predicting the most appropriate clauses for a legal contract from a range of possibilities based on the requirements of the contract.

Generative -  generative models are being used to automatically “blank page” generate passable e-commerce product descriptions. They offer tremendous potential to excel in environments where creativity and stylisation meet structured constraint. Architecture is one example where generative design is beginning to have an impact. There will be many more.

3 New customer propositions

Machine learning is driving a wide range of proposition innovations. B2B examples include precision agriculture solutions to optimise the cost and effectiveness of pesticides (see, for example, John Deere’s acquisition of Blue River), intelligent receivables matching solutions for corporate banking clients (see, for example, Bank of America’s Intelligent Receivables solution), and data and technology providers building platforms for clients to make use of and/or develop bespoke machine learning solutions on top of their core service of data management and provisioning (for example, SAP’s Leonardo offering). B2C examples include Google’s pixel ear-buds permitting near real-time face-to-face language translation and the Amazon Go machine vision-enabled no-checkout shopping experience.

4 Commercialisation

As machine learning becomes more widely used, organisations need to be aware of the value of the data they possess. Investors in early stage machine learning start-ups routinely value the ownership of or privileged access to training datasets much more highly than they value the start-up’s machine learning algorithms. Opportunities may exist to build new businesses in their own right founded on machine-learning driven competitive advantage.

New techniques such as representation learning are making it more easier to integrate a business’s data asset with that of relevant third parties. That can help predictive performance in new non-core but potentially lucrative areas. As an absolute minimum, businesses must start to give consideration to IP ownership of third party models trained on their data.

5 Disruptive models

Machine learning has the potential to radically revise pre-existing business models. Ocado’s swarm robot warehouse automation approach – now fully implemented – is arguably one such example.

Order picking is fully automated. It is done so in a way that minimises the number of different types of components involved (so reducing the cost of maintenance and repair) and maximises cost efficiency through making full use of the 3D space of a warehouse. This and other innovations across its order to fulfilment chain are enabling Ocado to become the e-commerce fulfilment provider to other retailers in other countries.

Another disruptive example is premonition in legal analytics. This provides commissioning clients with performance prediction for specific lawyers in front of specific judges on specific case types – thereby bringing transparency to what has until now been a performance-opaque industry where brand has often had great weight.

Figure 3: Systematically identifying machine learning opportunities

Example: Identifying machine learning opportunities across the FS operating model

    Govern Design Distribute Operate Support
  Indicators of potential
  • Strategy
  • Financial control
  • Legal and compliance
  • Treasury
  • Product
  • Marketing
  • Risk policy
  • Channel management
  • Customer servicing
  • Application management
  • Risk operations
  • Collections and recoveries
  • Supplier management
  • HR
  • IT

1. Enhanced prediction

  • Large quantity of potentially important predictive variables
  • Limited model data richness/ breadth
  • Relevance of unstructured information
  • Existing models more than three to four years old






2. Automation

  • Repetitive information- based tasks (extractive or generative) with complex variations
  • Ability to model and simulate sequential tasks






3. Enhanced/ New Propositions

  • Existing customer propositions are information rich or could become so
  • Propositions could be made more granular
  • Customer pain-point characterised by other attributes in this list



4. Commercialisation

  • Existence of or potential to create dataset of commercial relevance - in whole or part
  • ML solution created of high value but yielding limited long - term competitive advantage





5. Disruptive Models

  • "Blank-page" reassessment of value proposition, revenue model, delivery model






The range of potential applications of machine learning is extremely broad. So it is importance to have a systematic and comprehensive approach to identifying them. Structured reviews of the organisation. (see Figure 3) undertaken by mixed teams of business practitioners and machine learning scientists are proving successful.

Making it happen

A machine learning strategy is no longer a ‘nice to have’. Here are five key considerations to think about when creating your AI strategy:

1 Determine where machine learning might be exploited for the benefit of your business and your customers
2 Set priorities 
3 Set aside resource/engage third party support 
4 Set up your govern, design, distribute, operate and support model
5 Develop and support your people to function in this more agile environment.

All figures taken from Machinable.

This insight builds on an article published earlier this year in disruption magazine by Tariq Khatri.