Highly placed AI people have already bought apocalypse retreats
In an interview recently, a respected US technology author said: “.. it was alarming how many people I talked to who are highly placed people in AI and who already have bought retreats that are sort of 'bug out' houses, to which they could flee if it all hits the fan.”
That’s a comforting thought
So, is AI an existential threat to humanity or a transformational capability that delivers a better world for us all?
Does it change everything you ever believed about how businesses or organisations work or is it just a powerful-but-complementary technology for improving performance and enabling better experiences?
Like most developments in human history, I believe it comes down to ‘how’ it is used. We are right at the foothills on AI at present. Here are some of my perspectives on how AI can deliver value for organisations and also why I think most organisations are tripping up in the implementation of AI
Connected pervasive computing is driving our next industrial revolution
Connected pervasive computing is driving our next industrial – and social - revolution. Across the world, we face disruption, uncertainty, convergence and complexity
Connected pervasive computing (what many refer to as ‘Digital’ is creating winner-takes-all-economics.
It means that digital transformation is one of the biggest strategic issues for global business, education and government and analytics is the key enabler, particularly in the form of AI, or more accurately automated decisioning.
Just as psychologists cannot agree about a common definition of human intelligence, so there are many interpretations of what constitutes Artificial Intelligence. I like to define AI by its purpose; to make smarter, timely decisions in complex interactions, faster and cheaper than a human being. The word ‘decisions’ is the key one for me.
For humans, making decisions is not just a function of heuristics or empiricism. In making decisions, our brains are also using history and current context to calculate future probabilities. Unfortunately, we all succumb many cognitive biases that screw this up. Anyway, that’s another matter.
My point is that decision-making relies on probabilities. We live in a world not of self-evident truths but probabilities. Because humans prefer clarity to uncertainty, our policymakers, economists and journalists hubristically cast probabilities as facts. Saddam had weapons of mass destruction; the financial system was crash-proof; diesel cars polluted less. All were probabilistic judgments presented as facts; all were catastrophically wrong. Is climate change a fact? No, it’s probability. It may well be a very high probability but it’s nevertheless a probability.
And that’s essentially what AI is - predictive model management, the automation of mathematical probabilities. However, the huge increments in processing power and interconnectivity have dramatically accelerated the scope and scale of the application of predictive modelling into real-time, operational processes, be that Search, Recommendations, Robotics, Driverless Cars, Piloting Aircraft or managing customer service.
Consequently, for all organisations, I believe the effective adoption of AI into the decision-making process and systems across the enterprise will be the sine qua non of future survival, let alone success
Start-ups have opportunities and advantages, incumbents have to transform
For start-ups and new entrants, this upheaval should represent an opportunity. They can design customer propositions and business operating models mindful of the capabilities and advantages of AI to achieve creative disruption.
For incumbents, the task is one of business transformation, arguably a much more difficult challenge.
Either way, in my view, most organisations are struggling to land automated decisioning. As a consequence, they are not realising either the scale and timeliness of benefits from their investments and efforts.
At best, this represents a drag on performance; at worst, an existential threat.
Why implementations are failing
Digital has changed the process of decision-making for both consumer-citizens and the entities that serve them. And I don’t believe this fact is properly understood by organisations.
I believe there is still widespread ignorance about the topic, which leads to investment mistakes, poor implementation and disillusionment in what should be a transformative capability.
A senior executive said to me not so long ago: “so far our AI investments have just made our stupidity more scalable”. An executive colleague, on hearing about the automated decisioning capability we were planning told me: “AI means that the machine will just tell me the answer, without me having to ask”. And I have lost count of the times I have heard or read something along the lines of : “operators should learn from Facebook and Google, where data is king and every decision flows from what the data says”, which is naïve delusion, in my view.
In my view, there’s far too much obsession with acquiring technology and computer scientists and not enough thought and effort going in to transforming decision culture and business operating models.
You can’t just write a cheque for a new decision-culture and operating model, like you can with technology or data scientists. Changing decision-culture in a large organisation is particularly difficult, especially with executive preferences for managing by macro-decisions; heavy reliance on heuristics and hunch; decision-based fact-making and the financial, functional or product-based command-and-control models of management.
All of this runs counter to the effective adoption of automated decisioning.
Besotted by the magical promises of a world of AI and machine learning, many executives seem to have either ignored or down-played the human factor. But it remains absolutely essential
The ultimate target variable of AI is the human being; consumer, citizen, student.
Businesses and organisations will continue to be run by humans
Even if a machine makes a decision, that decision needs to be understood and controlled by a human (unless you really do sign up to the kind of dystopian future displayed in sci-fi films such as The Matrix and Terminator)
The US author Andrew Keen believes that the key relationships in the future will not be between people but between people and machines.
I think that’s right. The advantages will lie with those who have superior abilities to make the tech work for the human. That’s business enablement
So, the focus on technology and the blind spot about business enablement is, in my opinion, the biggest single reason why AI implementation fails.
The answer does not lie with Technology
AI changes everything, forcing CEOs to rethink how companies execute, with new business processes, management practices, information systems, proposition developments and the very nature of customer relationships
It requires a transformation in operating model to enable an enterprise to be suitably:
Agile - rapidly changeable to cope with new regulations or business conditions
Analytical – putting data to work improving the quality & effectiveness of decisions
Adaptable - learning from what works & what does not work to continuously improve over time
Automated decisioning, or AI, should greatly benefit in all three respects:
Agility – advanced automated analytical ecosystems can manage complexity, velocity and change with far greater effectiveness than a construct with just human employees can
Analytical – there are significant benefits from deploying advanced analytics not to just manage but optimise interactions throughout the enterprise model with an ability to calculate and trade-off billions of micro-decisions in a way that is beyond human staff. I believe the application of automated decisioning to the optimisation of the allocation of business resources is missed completely by most providers and users of AI.
Adaptable – AI also offers the benefits of evaluation objectivity, which is a valuable component of understanding what works, what doesn’t work and what would probably work better in the future to enable continuous improvement. It takes human subjectivity, self-interest and bias out of understanding the paths to better performance in a changing landscape.
However, the most important consideration is how AI is implemented and any organisation seeking to do so successfully needs to pay attention to three critical, inter-dependent factors:
1. Executive Decision Culture – the way an organisation determines decisions on business resource allocation to maximise its goals.
The decision culture required from AI is the opposite of top-down, fewer, bigger, better.
If executives believe they can maintain the same decision culture whilst just adding AI technologies, they will be sadly disappointed with results.
Executives need to be prepared to abstract themselves from the detailed decision processes and focus on what outputs they require and in what form, what constraints they have or wish to impose and what inputs they are willing to make
I think that executives also need to get their head around how to exploit the capability to optimise the billions of interactions in their business models. It’s a little like Sir David Brailsford’s marginal gains approach to UK cycling; making small improvements to a large number of factors that, taken together, add up to significant competitive advantage
For example, I have worked with a number of executive teams and it has usually been a big struggle to get them to change the way they see their customers. Getting them to work with simple customer segments rather than a homogenous base or via the lens of the products they provide is difficult enough, but it is far, far short the sophistication required to exploit AI. And there is much more value to be extracted from AI optimisation than segmentation.
2. Operating Model & Concomitant Organisation – the way an organisation designs and manages the accountabilities, roles and flows in its decision ecosystem for maximum effectiveness and efficiency.
You may have heard of Conway’s law, which was an idea put forward in 1967 by Melvin Conway, a US software engineer. He observed that the communication structure of an organisation structure was reflected in the design of its software. Over-simply, a disjointed organisation produces disjointed products. Transferring his law to this topic, I’d suggest one of the central problems is that an organisation will tend organise itself based on the decision capability it has today, not the decision capability it could, or should, develop.
The existing financial, functional or product-based model and structures may be the most suitable for human executives to manage but may not be the most suitable for automated decisioning deployment
In my experience, shoe-horning new technology and algorithms without modifying a traditional operating structure usually fails
The operating model and organisation need to be designed around the opportunities for AI to drive benefits. Essentially, this requires the extraction of humans and human decision-making out of the middle of certain enterprise processes and placed instead in the (usually head-end) part of the process where human skills, oversight and control are most suitable or prudent.
3. Business Adoption & Enablement– the way an organisation enables its human capital to effectively use its AI capability to drive performance improvements and business objectives
In my view, there is not nearly enough upskilling and development of employees to ensure these new capabilities can be exploited. I don’t mean the technical or mathematical skills.
The use of automated decisioning or AI requires different business approaches and techniques. It requires different thinking (and, in my view, we don’t adequately teach our children and our employees how to think well)
You cannot just dump new technologies or analytics methodologies on existing staff and expect a different result. We might all have smartphones, but a 16-year-old uses a smartphone quite differently from a pensioner, despite it being the same device.
I’ve the mistake of ignoring business enablement myself and seen the unedifying results. In the past I’ve put smart automated decision capability into large organisations and then found most existing business stakeholders just didn’t ‘get it’ and, therefore, didn’t use it properly. In one large organisation I worked for, my team used to joke with me that we’d delivered all this sophisticated analytical and decisioning capability ‘only so that the marketing department could send customers different coloured envelopes’.
Even worse to me are the Kool-Aid brigade. These are the employees – from many different disciplines - that have belatedly jumped on the Digital, AI, Machine Learning bandwagon and swallowed the text book whole. They surf the latest fads and waste the organisation’s time and money messing around with a miscellany of ‘cool’ technologies and algorithms that, in the end, don’t deliver benefit. I experience many organisations now where senior executives are seriously questioning why they have invested in all this technology and data scientists when they have very little value output to show for it.
If the wasted money wasn’t bad enough, the wasted time is another matter. Time is the one thing you cannot buy, and velocity is king in this emerging AI world.
The good news is that, with suitable focus, business adoption and enablement can be cracked, even with a major shift in upskilling. I’ve managed to make it happen in diverse areas such as targeted linear TV advertising, TV rights evaluation, retention and acquisition in subscription models, cross sell and up-sell.
Both the prize and the penalties are significant
The agility and adaptive advantages that AI offers in a winner-takes-all economics should be self-evident.
However, when it comes to the financial outcomes, I read and hear a lot about AI benefits that concentrate on cost out-take, essentially replacing expensive, awkward, difficult-to-manage human beings with machines.
Many executives I talk to in incumbent market positions look enviously at new digital natives who have designed-out the human cost in their business models. For example, avoiding costs of call centres, retail or installation.
I completely understand the attraction of AI implementation for capital and operational expenditure (and, therefore, margin) benefits, especially for the large incumbents in Banking, Retail and Telecoms. And worthwhile savings are indeed achievable.
The difficulty for me in this unidimensional focus is that no-one achieved growth through cost-cutting. The real prize for me is in the deployment of AI to make micro-improvements in every decision, whether machine-to-human or machine-to-machine in every interaction that, taken together, provide significant gains to growth
From my experience, deploying AI for decision optimisation can increase enterprise end value by over 30% over the most sophisticated analytical targeting. If your organisation isn’t using even basic analytical targeting, gains would easily be 60% or more. I’ve personally seen uplifts of hundreds of a percent when up-grading from the sort of faux behavioural decisioning one often finds in popular digital platforms to advanced real-time, dynamic decisioning, or AI
There is also one more collateral benefit that I’ll mention. It may seem bizarre, but there are often large improvements in value to be gained just in the re-modelling of a human-based process into a computer-based one. That’s because many enterprise human-based processes have unseen entropy. In other words, what you think happens, doesn’t actually happen in practice. I have seen tens of millions of pounds of value appear when implementing automated decisioning even before switching the clever AI part on.
Don’t be blind-sided by the snake oil
AI is coming of age and has now powered a few businesses we’d never heard of a decade or so ago to be the biggest companies in the world
A lot of AI doesn’t yet work very well but that will change. I also believe that many future applications touted for AI may, in fact, not be either desirable or viable. Along with environmental sustainability, that is going to be one of the biggest debates we have as a species.
Don’t be blind-sided by the snake-oil. Vendors, Consultancies, Journalists and your own staff may tell you it’s all about this or that cool technology (which seems to change every six months, by the way) or the large data science team you need to recruit.
It’s not.
My advice is not to obsess about the technology or expend your energies finding that apocalyptic bolt-hole in case it all goes wrong.
To be agile, analytic and adaptive you’ll need to pay close attention to the human factors, in both customers, citizens, employees and stakeholders
You will need to get enterprise decision culture, operating model and business adoption right to take advantage of the opportunities in the time you have available to grasp them
Get that wrong and the penalties for the losers could be lethal.
Get that right and the prizes for the winners are significant
Comments