What work works...and can I get a robot to do it for me? (2/5)

How much of what you do in the name of work actually matters? And just how scared should you be that a robot could do it better/faster/cheaper?

The pop-science answer to the first question is that work matters when it does something valuable. That could be growth – in revenue, profits, customer/user outcomes, market share – or it could be better-managed risks – reduced costs, environmental or reputational damage limitation, unlocking key business drivers and blockers, etc. Good work makes a difference your employers and you care about.

The popper-science answer is that if you haven’t been fired yet and you don’t wake every morning dreading the day ahead, you’re probably doing okay.

As for the second technophobic question, the pop-science answer is anything between Very and A Bit, depending on nothing more than sloppy statistical extrapolation and/or overactive inference. 

The popper-science answer is: run for the hills.

Enough panic. This series is here to separate the wheat of reliable science from the chaff of too much current business practice. Let’s look at what we can say with accuracy, reliability and relevance about work:

A century’s productive work and it’s still not finished

Productive work is an area of research where we have huge and detailed datasets stretching back to the Victorian era, from Frederick Taylor Winslow’s time-and-motion studies in the nineteenth century to the electronic-tagged workforce of our present day. As ever, some of this research is deeply flawed; some has yet to be replicated; and some is very specific to a particular role, industry or time, but thoughtful meta-analyses – comparisons of multiple research studies – reveal strong and consistent themes across jobs, gigs and projects.

One of those revelations is that there are only three worthwhile things anyone does in the name of work:

·      Process work – getting stuff done

·      Problem work – finding ways to get better results more easily

·      People work – cooperating, managing, inspiring, serving others

In practice, most gigs, jobs and project roles contain a mixture of at least two of the above, but research into the drivers of performance suggests that in just about every case there is one predominant type of work and that focusing on that core work is the best way to improve results.

So let’s look at what we know about each type of work – and how vulnerable each is to those pesky robots:

Which part of “do that” did you not understand?

Over 2/3 of what people do for pay is Process Work – everything from following sets of procedures to filling in forms to performing heart surgery to flipping burgers. The bulk of Process Work tends to be pretty unskilled (think: roadsweepers), though some can be highly skilled (think: minesweepers).

Meta-analyses of what drives successful Process Work come to one big conclusion: the trick to Process Work is efficiency: work steadily and methodically, use the fewest possible resources, pay attention to minor snags and replicate, replicate, replicate. If you can increase quality at the same time, great, but Process Work is much more about good-enough predictable consistency, not best-possible breakthrough – you need truckers to reliably deliver the goods from A to B every day, not to put all their energies into driving the perfect lap. 

Sounds easy, but in the real world Process Work can be so mind-numbing that it’s easy to slack off into chattering to co-workers, checking your Twitter feed, staring out the window or watching paint dry.

Which is where the robots come in!

Robots are great at Process Work. They follow instructions without interrupting to ask irritating questions or skipping steps or deciding to do things a funny new way they just thought of; they stick to the task in hand and don’t get distracted; they work 24/7 without whining about tea breaks or holidays; once they get set up they’re generally cheap to run and do what they’re told. For bounded, replicable, logically-structured tasks, they’re unbeatable. They’re cheaper/faster/better.

Which is why Process Work is often the first part of a job or industry to be automated - challenging a parking ticket, taking payments, filling in forms.  Especially when the non-automatable bits don’t need paid humans but can be done by someone else for free – think of all the time you spend typing your credit card details into websites, or shifting groceries through those annoying automated tills. Increasingly, it’s not just administrative Process Work that gets done by silicon and steel: combining advanced mechanics with machine learning has led to burger-flipping robots and all kinds of industrial line work, though industrial automation has not necessarily resulted in job losses. 

What nobody has fully worked out yet is how to make the remaining non-robot parts of Process Work fulfilling – or even endurable. Just look at the grim faces of former cashiers who now have to trouble-shoot automated tills. Or talk to senior lawyers who worry how trainees who now rely on text comparison software will develop the deep expertise that leads to partner-pay-level judgement. Process Work that is fiddly is the least enjoyable type of work out there. And if there’s one universal truth about human behaviour, it’s that people tend to do most what they enjoy most.

I’ll return to the dilemma of how to handle human-robot cooperation in a future post. For now, let’s race on towards the less-obviously automatable areas of work: Problems and People.

Sorry, what was the question again?

Problem Work is all about new things – analysing and solving difficulties, spotting opportunities, finding better ways of working, inventing, repairing, innovating and adapting.

Great Problem Work is essentially great thinking – coming up with creative approaches, finding ways to compare disparate data, having the patience to work through analyses and dig deep into data, experimenting and learning – plus the ability to present the results of that thinking clearly and compellingly enough for people who aren’t as expert as you to take on board what you’ve discovered. For most Problem Work, there’s a combination of micro-analysis – a plumber meticulously tracking the flow of water through a house’s piping – and big-picture visioning – that same plumber working out that it’s a more powerful boiler, not a rerouting of the pipes or the fitting of a pump – that will give you enough hot water to take a shower when some inconsiderate family member has run a bath, loaded the washing machine and started the dishwasher ten minutes beforehand. Increasingly, Problem Work requires the ability to work with probabilistic as well as concrete data, extrapolating and inferring and determining significance. Such complex problems present challenges not just in solving the problem, but in putting together a workable way to effect that solution – real world Problem Work, not ivory tower abstraction. 

Very little Problem Work – far less than Process Work, to most people’s surprise – can be done in isolation. Even work like building data models, which typically requires long periods of concentrated focus, also involves negotiating with others over data access, factor choice and weighting; discussions with other data experts; briefing software engineers on how to put the model to use; reviewing potential solutions; and agreeing parameters and project structure. Two heads, so long as they’re differently-thinking heads, are better at Problem Work than one.

But getting two different and capable heads isn’t easy. Problem Work is supply-constrained – there aren’t enough people around who combine high-enough levels of reasoning ability and other problem-solving capabilities with the motivation to stick with this kind of highly demanding - ok, exhausting and frustrating - work. Automation is rapidly taking over some of the more linear and bounded Problem Work, but the Problem Work that remains – the multi-factor, creative, unbounded messy stuff where rules are ambiguous or unknown and success factors vary as much if not more than the data – is growing in importance and, unless AI makes the leap from Specific to General Intelligence (nowhere on the horizon at present), that work will have to be done by humans or, more likely, humans augmented by machines.

Again, cooperation’s a topic for future posts. Back on the topic of work in general, the biggest problem with Problem Work is quantity. Even if huge numbers of us got hugely smarter (something that actually may be happening), there just isn’t enough Problem Work around to replace the huge amounts of Process work that have/are/ will be being lost to automation.  What remains for the bulk of human workers is…

French existentialist hell

People Work does what it says on the can: managing others, persuading them, communicating, working collaboratively. A former colleague once cynically summed it up as how you get other people to do your work for you.

Oddly given the fact that their job is largely to get other people to work effectively, many senior leaders still consider People Work as more of an art than a science – some workers are just natural People People, that kind of evidence-lite assumption.  Sure, People Work obeys few Newtonian laws, and there has been less and less-rigorous research in this area compared to other types of work, but there is a fair amount we know for certain, a great deal that we know within limits (e.g. it’s true within certain populations or contexts) and a growing number of plausible hypotheses that technology and renewed interest in and funding for research might soon validate.

One of the things we know for certain is that different people get great results from doing People Work very differently. Introvert salespeople will show more modesty and humility in the face of clients, extravert salespeople will be readier to share compelling stories – each a highly effective persuasive technique that wins and keeps business. Overall, Emotional Intelligence – not just how you empathise but how you influence others, understand and manage yourself – is a robust predictor of and guide to effective People Work. Thinking about the areas of Limited Knowledge, we know a great deal about the persuasive aspects of People Work – thanks to the huge amount and easily-accessible nature of the data on salespeople’s performance and behaviours – and rather less about the human systems – teamwork, for example, though initiatives like Google’s Project Aristotle are rapidly increasing our knowledge.

But the biggest problem with understanding People Work isn’t the lack of research, it’s the fact that humans prefer making up stories to observing reality. Ask people to identify the drivers of success in People Work, then look at how those factors actually map onto performance, and you’ll see bigger gaps than with Process Work or Problem Work. Tools that have been clearly and repeatedly proven to be little better than snake oil – Myers Briggs and Belbin’s Team Roles, I’m looking at you – remain popular with too many people and institutions who should know better. So it’s in People Work, even more than other areas, where I see the biggest gains to be made from making decisions on the basis of robust research rather than trusting hunches, individual examples or the lattest glittery received wisdom.

I know, I know, robust research is thin on the ground. And there are lots of false starts and wastes of time and money out there. Too much investor (and therefore startup) attention has gone into the easily-measurable and popular with clients stuff (tracking physical movement around the workspace, for example, or frequency/quantity of communication with colleagues) rather than focusing on deepening and sharpening our understanding of already-proven predictive factors (e.g. psychological safety). Neuroscience offers fascinating potential, but so far even fMRI machines can essentially only detect the strength of neurocortical arousal – it can tell that I’m deeply engaged by what you just said, but not whether I love it or hate it. Some of the research where I see biggest opportunity is mapping the various and varied paths of human-machine interaction, personality and motivational drivers and, above all, human system dynamics, but while these fields have produced some very interesting hypotheses – which I’ll return to in detail in a future blog – it’s still early days.

But we’d better hurry up with that research, because People Work is a growth area. Biggest growth, as ever, is in relatively low-level jobs – care worker roles, for example – many of which are not just low-paid, low-security and low-autonomy but are too often experienced as draining, unpleasant and unrewarding. But it looks very much like these negative aspects might not be inherent to People Work so much as a reflection of how People Work is managed and organised – care workers in self-managed organisations such as Buurtzorg report high levels of satisfaction as well as delivering impressive business performance and growth.

We need to start learning faster and better about how to do better at People Work because it’s the one area of work where robots are close to nowhere. AI is great at depressing performance by making us addicted, distracted and less-nuanced, but we have been far less successful in using technology to drive improvements in productivity or embed the ways of working that have been shown to result in great People Work outcomes. Full disclosure: I’ve spent the past twenty years working in this area, and while I’m proud to say that there’s some great proven-effective People Work tech out there, the robots are nowhere near beating humans at People Work, as appalling chatbots show us almost daily.  The core of People work is eliciting and reinforcing emotional response, and that’s enormously easier for humans and other mammals to do compared to anything made up of silicon and steel. 

Work all done?

That’s it for work! Master Process, Problems and People and we can all go home and put our feet up. Exactly how we do that – what are the best ways of doing each type of work, and how we can individually know what we’re good at, how we can learn fastest and what we should leave to others or the robots – will be the focus of the next blog in this series.