Understanding the Building Blocks of Work

Currently each of us has a job, a role. This typically involves a series of goals that we have to achieve in order to fulfil our responsibilities. In turn each of these goals consist of a series of tasks we need to perform, or execute, to fulfil each goal. In effect, this is what every job description is specifying.

As technology develops it can perform 'some' of our Tasks. But Goals and Roles are still overseen by Humans.

Over time technology will enable a series of Tasks, that make up specific Goals, to be performed entirely by an 'AI Agent'. These will be discreet mini applications that are tasked with performing X, and provided with the necessary capabilities to do so.

This means that some tasks will remain as a collection of tasks performed in part by humans, and in part by technology, whereas other goals will be possible to achieve solely through the application of technology. That goal can then be removed from the ‘job description’ and handled separately.

What might also occur is that multiple goals can be fulfilled by pulling from a repository of ‘AI Agents’ that can be combined in ways that enable them to have utility across multiple domains. Maybe with 30 ‘AI Agents’ we can deal with 50, or 100 goals.

Think of it like Lego; with the same pieces we can combine them in different ways and create a multitude of different things.

Let me bring this to life with a practical example from the marketing world, where we're already seeing the early signs of this transformation. Imagine a Product Marketing Manager's role today. Their goals typically span market research, competitor analysis, content creation, campaign management, and performance tracking. Each of these goals encompasses dozens of individual tasks.

Now picture how this unbundles and rebundles with AI agents: One agent might continuously monitor competitor websites, social media, and pricing, synthesising changes into actionable insights. Another could generate first drafts of marketing copy across multiple channels, maintaining consistent brand voice. A third might analyse campaign performance in real-time, automatically adjusting parameters for optimal results. Together, these agents could handle what previously required a team of specialists.

But here's where it gets truly fascinating: These same agents could be recombined to serve entirely different goals. That competitor monitoring agent? It could also feed insights to product development. The copy-writing agent could support customer service responses. The analytics agent could inform inventory management.

This is where the 'rebundling' becomes transformative. Our Product Marketing Manager isn't replaced – they're elevated. Instead of being caught in the weeds of daily execution, they're now orchestrating these agents, focusing on strategy, creative direction, and the deeply human aspects of brand storytelling that no AI can fully grasp. They're identifying new opportunities for agent collaboration that we can barely imagine today.

What must also be noted though, is that for a fixed amount of work, fewer such well equipped humans will be required.

With that caveat this transformation of the Product Marketing Manager's role illustrates a broader pattern we're going to see across knowledge work: roles will be decoupled, tasks will be redistributed, and entirely new forms of value will emerge. But crucially, this isn't just about efficiency – it's about unleashing human potential in ways we're only beginning to grasp.

So how do we navigate this transformation in our own work? How do we ensure we're architects of this change rather than merely responding to it? The path forward requires both systematic thinking and creative imagination.

Ideally, like this:

Taking Action: Your Personal Job Audit

First, you need to go through this process of breaking down jobs and processes into their component parts. I advise you to do this for your own job. Map out (ChatGPT and markmap.js.org are great for this) the goals you are tasked with then think about everything you need to do to get them done. Often it is worth digging a layer or two further; what are the sub-tasks, and then how are they achieved? As seen in the mind map below, here’s how a common goal like 'Identifying Optimal Locations' can be broken into its component tasks.

Once you’ve defined these, you can identify which tasks can be automated, augmented by AI, or should remain fully human.

Now you’re getting to the upside - the rebundling process.

You can start reconstructing your workflows: rebuilding processes with optimal human-AI collaboration.

And thinking about how you redistribute skills: where do you reallocate human skills so that they add the most value.

Now clearly there is a lot of devil in the detail with all of this, but I hope the abstract principle is clear that we’re moving very much to a ‘Human + Machine’ world and we need to redesign the work we do accordingly. Processes and workflows can, and need to be, reworked to optimise our capabilities, and those of ‘the machines’.

Learning from History: The Electricity Parallel

It took 40 years for electricity to transform factories. And the productivity gains (which were dramatic) only occurred when the form factor of steam powered factories was completely overhauled. From being one monolithic machine, driven by a central drive shaft with chains and pulleys, to a patchwork of interlinked but separate processes each with their own, electric, power.

We are at a similar inflection point, where to achieve the gains we have to jettison the past, and build for the future. This time though I expect the timeline to be compressed. Sure, every such change takes a while to permeate through society, often longer than the optimists expect, but the gains to being an early adopter with AI are such that progress might well be 10X faster this time around. After all we are building on an installed base of 50 years of computing and 30 of the Internet.

The Human Factor: MIT's Surprising Findings

There is an interesting twist though to achieving strong productivity gains with AI, a peculiarly human one. Researchers at MIT recently published the results of a meta study of 106 individual papers discussing the successes of human + ai collaboration. It turns out that In many cases the AI performed best when left alone and often when humans and AI worked together it performed worse than either could have achieved on their own. Rather amusingly it seems that we humans often think we know best when we don't and are minded to overrule our AI helpers.

Where real human + AI 'Synergy' occurred, where the combination produced better results than either could apart, followed four patterns, and for these reasons:

Four Patterns of Successful Human-AI Collaboration

• When Tasks Have Both Pattern AND Exception:  AI excels at recognising patterns, while humans are better at handling exceptions. Together, they create robust systems neither could achieve alone.
• When Scale Meets Judgment:  AI processes vast amounts of information; humans bring contextual wisdom. Combining these strengths leads to smarter, more nuanced decisions.
• When Creativity Needs Structure:  AI generates countless variations; humans curate and refine. This synergy drives more effective innovation.
• When Analysis Meets Intuition:  AI finds correlations; humans provide causal understanding. Together, they solve complex problems that neither could tackle independently.

The Human-AI Symphony: Our Path Forward

We need to look for workflows where AI and humans can each play to their strengths while compensating for each other's weaknesses.

I think we all have a natural aversion to thinking machines can be smarter than us (whether or not we believe it to be possible), whereas we are happy if we can retain agency over what matters to us. So if we can find ways to work with technology without our egos being impacted, we are far more comfortable letting 'whatever will be, be'. In the end, mastering this technological revolution demands as much wisdom about human nature as it does technical understanding. The winners in this transformation won't just be those who adopt AI fastest, but those who learn to collaborate with it most intelligently. There is a phrase going around about how ‘Language models are the revenge of the humanities graduate’ and this resonates with me. There is something strange about interacting with them. You know it is science, but it feels like Art. What it isn’t is more of the same.

‘The times - (most definitely) - they are a-changin'

Your Next Step:

This week, pick one goal from your role, break it into tasks, and identify where AI could amplify your impact. You'll be surprised by what you discover.