Organisations need to urgently take action to address the communication and skills gap between their corporate leaders, key decision makers and AI developers, in order to reap the full benefits of AI innovation and transform into an AI-competent organisation.
What lies at the root of the gap between the promise of AI and the practice of an AI–based strategy? As recent evidence-based inquiry suggests1, companies widely report that the adoption and use of AI techniques significantly lag the promise they were led to believe AI holds for making work more efficient and productive. The answer is not technical. It is organisational and cultural: A massive skills and language gap has emerged between key organisational decision makers and their ‘AI teams’. It is a barrier to innovation in the workplace that promises to stall, delay or sink algorithmic innovations for the next decade or more. And it is growing, not shrinking.
The Skills Gap. Here is the crux of it. The skill sets of those in the upper echelons of organisations are out of sync with those creating ‘AI solutions’:
• Executives know how to talk to other people. (See figure 1) They have complex and well-honed abilities for listening, empathising, deliberating, energising and de-energising meetings, emoting and reading others’ emotional landscapes and adapting their ways of being to seemingly intractable social situations.
• Those who develop machine learning solutions to business problems know how to talk to machines. They write pseudo-code and code, develop large scale platforms that scale to millions of users, aggregate data in multiple formats from multiple sources, specify and code interfaces for users that incentivise them to interact with the machines they build via combinations of words, images, colors, haptics, and action prompts.
• Developers want clear, precise instructions that are easily translatable into code or pseudo-code; but –
• Business development executives provide them with stories and anecdotes. This lack of computational savvy gap near the top of hierarchical organisations has been a problem for every ‘IT’ wave in business dating back to the 1990’s – but the widespread use of ML algorithms working on large data sets exacerbates this problem and brings it to a boil.
The algorithmic skills gap arises because people belonging to these two groups cannot speak to one another in productive ways. They aim differently, see differently, think differently and feel differently:
• Machine learning programmers want clearly specified cost functions they can use to train their algorithms; but –
• Chief strategy officers and business development executives supply them with aspirational goals phrased in the fuzzy language that coders routinely call ‘corporatese’.
• Big data, multi-user platform developers want clear allocations of decision rights among the end users of the platforms that specify who gets access to what information when and who gets access to information about the identities of users having access to information and the specific levels of user privacy that are achievable given the precision and reliability of the statistical analyses these data are used for; but –
• Clients will only talk about broad principles of fairness, diversity and inclusivity that should be used to design the platform they are contemplating purchasing, but do not specify these concepts to levels of precision that makes them amenable to algorithmic implementations.
[ms-protect-content id=”9932″]
We need to bridge the skills gap that leads to this self-perpetuating communication debacle. Organisations need people who can talk to both people and machines and people who can talk to people who talk to machines to inhabit their upper echelons. Key to competent communication and collaboration is a common language and pattern of reasoning that makes financiers able to communicate to engineers and marketeers – which I have called Intelligent Artificiality.
The current lingua franca of business is a significant part of the problem. The proliferation of economists in business school faculties since the 1960’s has contributed to the production of a common language system (‘cost benefit analysis’, ‘competitive mapping and simulation of competitors’ responses, marginal cost and rates of substitution analysis, portfolio planning, … ) in which executives plan their actions and justify their decisions. In the age of fast algorithms working on distributed data sets, this language system is outclassed and thus dated. It needs to be replaced.
By what?
Communication Codes and Protocols for the AI-competent Organisation As Stephen Wolfram and Jeanette Wing have argued, computational thinking needs to be proactively expanded beyond the current reaches of computer science departments and technical teams. Wolfram points out that, for any field of human endeavour X (from linguistics to architecture, from logic to music and from plasma physics to dance ethnography) there is now a specialised field of computational X: e.g. computational analysis of discourse, computational historical research, etc. Businesses have been too slow to get with the computational wave, and are paying the price.
What to do?
• Computational modeling should complement causal (physics, thermodynamics) and teleological (economics, parts of psychology) modeling in everyday business discourse – as well as in business. Computational strategy, computational marketing, computational logistics and computational hiring should come next. The language of business needs to meet the language of computer science on an equal playing field;
• Intelligent artificiality – the discipline of specifying business problems and challenges at the algorithmic and computational levels including the procedures, routines, data sets, objective functions and tolerable error rates for possible solutions – should complement and in some cases replace the standard conversational capital imprinted from economics, finance and accounting.
• The basic operations of defining and structuring problems, enumerating and evaluating solutions, designing solution algorithms, iterating on a solution of the requisite accuracy and reliability, and evaluating the complexity of a problem before trying to solve it – should become the lingua franca of executive dialogue and of design – and solution-oriented conversations between executives and their technical teams;
• Practice-guided training – helping executives turn ‘business problems’ into ‘computational problems’ intelligible to coders and scientists – should be deployed at scale to help executives appropriate and master language for designing solutions;
• Those trained in talking to machines – researchers and developers – must be helped to develop ways of broadening the domain to which their current patterns of speaking and referring apply: not just sterilised data sets already parsed into the requisite variable fields stored in the right database formats, but real organisational entities like people, tasks, roles, decision rights, incentives, expertise, product features and the architectural topologies or products, services, value linked activity chains and organisational influence networks;
• Organisations must develop the relational and communicative skill base of their technical team members. Functioning competently in a top management team or board meeting is about much more than accurate reporting, valid reasoning, ‘critical thinking’, ‘making decisions’ and the production of proofs of optimality or uniqueness of a solution. It is about finding successful modes and means of expression, choosing language to match context, and producing patterns of facial, vocal and gestural expression that evince the right level of conviction, responsiveness and trustworthiness;
• So-called ‘soft skills’ are among the hardest to develop and wield. They are best thought of as ‘hot skills’: you need to deploy them in emotionally hot states and they are ‘market-hot’. It is time organisations recognised them for what they are and took on the challenge of developing them in those that have been relegated to the category – or is it a dust bin of ‘autistic technical experts’?
The Reset. ‘AI strategies’ fail because AI is a means, not an end – which is what ‘AI strategy’ misleadingly implies. The daftness of the ‘Do you have an AI strategy?’ question jumps out when you ask, “Do you have an Excel strategy?” This is a daft error that leads to half-baked algorithmic solutions looking for real business problems, inaccurately reported technical advances using specious examples to make exaggerated claims aimed at garnering corporate resources managed by misinformed executives, and venture capitalists and entrepreneurs slapping ‘AI’ on their funds and products even when peddling technology developed in the 1960’s to lower their costs of capital and attract new funds.
This is a degenerative state of affairs. It is unfortunately fuelled by a coterie of AI pundits and economists that have never written a line of commercial code, designed an algorithm, or built a platform that uses machine intelligence or a business that sells it. It is time for CEO’s and CTO’s to take control of the so-called ‘AI agenda’ and speak this truth to their teams: “We do not – and never will – have ‘an AI strategy’. That is as absurd as having a PowerPoint strategy. We will come up with solutions to business problems using methods congenial to computational scientists, researchers and coders, and thus bring their skills to bear on producing the best products in the industry.”
The tools for doing so are at hand. And the resources currently being wasted on ‘AI strategies’ are misspent. Let us reallocate them smartly.
[/ms-protect-content]
About the Author
Mihnea Moldoveanu is Desautels Professor of Integrative Thinking, Professor of Economic Analysis and Vice Dean of Learning, Innovation and Executive Programs at the Rotman School of Management, University of Toronto, where he is also the Founding Director of the Mind Brain Behavior Institute and the Desautels Centre for Integrative Thinking. He is the Founder of Rotman Digital, the Rotman Self Development Laboratory and the Joe Weider Foundation Leadership Development Laboratory. He is past Founder, CEO and CTO of Hefaistos, Inc. (designer manufacturer of ADSL modems) and of Redline Communications, Inc. (TSX: RDL) a leading manufacturer of cellular base stations and broadband wireless networks. He is a Senior Advisor to the Boston Consulting Group and a member of the global advisory board of McKinsey Academy. A Top 40 under 40 for his contributions to business and academia, Moldoveanu is the chief architect of 3 machine learning platforms for increasing the effectiveness of skill development in business and higher education.
