Productivity Hopes and the Solow Paradox

Away from the breathless excitement about GPTs and their latest impressive content creation tricks, there has been some thoughtful discussion recently about the impact of AI on work and the organisations we build to coordinate it.

In the Financial Times, Eric Brynjolfsson talks about the likely macro economic impact of AI and guesstimates that it could help boost productivity gains to around 3% per year in the current decade, despite the difficulty of attribution, which we saw with the rise of personal computing and the so-called Solow paradox. Genuine benefits often don’t show up in the statistics, and sometimes we might see a productivity J-curve, with initial flat or mildly negative effects until the real gains kick in. He also likened this new ‘general purpose’ technology to the rise of electrification in industry, where there was a gap of several decades between the emergence of the technology itself and the re-tooling and re-organising of factories and work sites to take advantage of it.

Amidst all the urgency to do something obvious and visible, like appoint a Chief AI Officer or announce a partnership with an AI tech provider, it is worth asking how exactly conventional organisations could re-tool or re-organise to maximise their preparedness for AI, and make the most of it as it becomes available.

person writing on white paper


Automating Tasks and Workflows

In a recent Diginomica piece, Stuart Lauchlan quoted Sanofi’s CEO Paul Hudson from a WEF panel event in Davos about the impact of AI on jobs:

Our people move from doing analytics to working with insights to doing something to have impact. …We have no objective to reduce the number of people because AI can do that. We have a big objective to increase productivity, and we have an even bigger objective to establish more insights that can lead to more valuable delivery of healthcare for patients. That’s so exciting and most of these things can’t be done by a human being not even one good at Excel or PowerPoint

Yes, there will be some low-hanging fruit that can be picked by using Gen AI to summarise research or create content, and if we can address potential privacy and security concerns, this will save a lot of time; but the next stage will be about real automation and augmentation of all kinds of tasks and workflows. AI is most likely to automate individual tasks rather than entire roles, so we need to think more flexibly and creatively about the unbundling of tasks, roles and jobs in the workplace.

But we also need to think about how we augment peoples’ work with autonomous agents and personal smart tech that goes beyond single Gen AI prompts, and helps people chain together agent actions to address their own very specific workflow needs.

The consulting group BCG announced an ActionKit framework this month that is intended to help simplify this process for people by using routing agents to decide which AI agent is best placed to perform a task, and then combining the various workflow steps into an action plan.

                                     OpenAI video generated from a text prompt using Sora

Re-inventing Middle Management

This sounds rather like the work a conventional middle manager might perform in a bureaucratic organisation, so re-thinking the role and focus of mid-tier management roles will also be an interesting area of enquiry, and perhaps this is where we will see AI replacing jobs, unless that redefinition can find a role focus that is greater than just micromanaging process steps.

Mindaugas Petrutis wrote about this recently:

Now imagine a future where middle management as we know it is reimagined, or even phased out, influenced by concepts like Scott Belsky’s ‘collapsing the talent stack’.

In such a world, AI doesn’t just automate tasks but also blends roles, making each team member’s contributions more interdisciplinary. This concept, seen in startups and increasingly in larger companies, suggests a shift where middle managers might need to evolve into roles that amalgamate leadership, product development, and direct customer engagement.

This change, driven by the need for agility and faster decision-making, could see middle managers becoming more like versatile leaders, adept in various aspects of business from technology to design thinking.


Catalysing Previous Waves of Change

Of course, the introduction of enterprise AI is not happening in a vacuum. It is happening in an environment where previous waves of change are still sloshing around, and might combine with or catalyse some existing forces of change – most obviously data and analytics, but also the rise of social computing and the connected company. For AI to really show what it can do, we cannot continue to divide work into vertical silos. We need connected services, structures and data.

Creating a common data layer in the organisation will make it possible to derive insights from across various sources that might not be obvious within a silo structure. Turning process work into services that can be combined into different workflows will help accelerate automation, and give agents something to work with. But also, let’s not forget online collaboration, knowledge sharing and online-first ways of working.

If our goal is to create smart organisations, we will need all of this to come together with AI and automation as a force multiplier in augmenting human capabilities and collective intelligence, whilst making it much easier to get the basics done.

As the AI pioneer Andrew Ng put it in his recent conversation with Azeem Azhar about AI in business:

be it human intelligence or artificial intelligence. The world has huge challenges, even major challenges like climate change and pandemics. But then also I think today hiring intelligence is very expensive. Only the wealthiest among us can hire huge amounts of intelligence to advise on the medical problem or understand a two-third child because human intelligence is expensive. So I think that I look forward to a world where you can hire intelligence inexpensively so that you no longer have to worry about the huge medical bill for falling sick and going to see a doctor or for sending yourself or your child to school. And that’s the future I’m excited about building, but it will take years and figuring out many use cases collectively to apply AI to all the places it can, that intelligence can apply.

That sounds to me like a great guiding principle for AI in the enterprise – not cost reduction, job cuts or turning people into machines, but making collective intelligence accessible and affordable to all.