AI and machine learning are unmatched in their ability to extract insights from colossal data sets. But, in most cases, humans remain far better at understanding the context of those insights, and the subtle nuance of their meaning. By bringing the two together, human-machine collaboration marries deep human experience with the scope and scale of analysis delivered by AI.
What is it?
Put simply, human-machine collaboration just means having humans and automated technology working alongside each other to achieve a shared goal.
That can take a huge range of forms. Most commonly it involves having human experts review and verify the outputs of AI or machine learning models, adding their own knowledge and expertise, filtering out bad outputs, and helping to train AI to deliver better outputs over time.
In process terms, it can be used to refer to operations that involve both humans and automated tools. This could be something physical like an assembly line that involves both human and automated touchpoints, or a business intelligence process that combines AI and human analysis.
What’s in for you?
Human-machine collaboration enables you to look at automation in an entirely new way. Rather than simply automating work completed by human teams and cutting costs, it uses technology to augment human intelligence — prioritizing high-value outputs over immediate cost savings.
It lets you make the most of the expertise and knowledge your team have developed over the years — freeing them from routine, mundane tasks and empowering them to focus their time and effort where it’s most valuable to the business.
For now, there are limits to what AI and machine learning alone can deliver. Human-machine collaboration adds the all-important human verification layer that’s essential for ensuring that whatever conclusions your AI draws from the data you feed it, they’re properly translated into the right outputs for your customers.
What are the trade offs?
The clearest trade-off with human-machine collaboration is cost. When a process is fully automated and no humans are involved, you save on labor costs. When it’s completely human-led, you don’t incur any of the costs associated with building, training or operating AI or machine learning models.
With human-machine collaboration, you get neither. You incur both AI, and human expert costs. But, what you get in return is very high quality outputs that neither humans or AI could have delivered in isolation.
For that reason, human-machine collaboration is primarily used in high-value exploratory or innovation projects — the kinds of use cases that demand both large-scale analysis, and the depth of expertise held by human teams.
How is it being used?
In Sweden, distiller Mackmyra recently used human-machine collaboration to create a new single malt whiskey named “Intelligens”. Working in partnership with data-driven consultancy Fourkind — now a part of ThoughtWorks — the distillery used AI to crunch through billions of recipe component combinations, alongside sales and customer preference data to create new recipes that customers should love.
The distillery’s master blender then selected the best AI-generated combinations based on her own experience and expertise. The result was a whiskey that proved hugely popular — selling out extremely quickly.
That’s just one recent use case. But most human-machine collaborations will look something like that — human expertise augmenting AI outputs to deliver high-value results.