Use your deep mind: AI for thinking better, not just finishing faster.
Technology should be designed with human needs in mind, but our needs extend beyond merely automating tasks. We must use AI to augment, not shorten, human effort.
In the recent Economic Index report, Anthropic refers to “automation” and “augmentation” modes to distinguish between doing and learning modes of using AI. As automation is surging ahead of augmentation, we are in danger of “using AI to solve the problem of thinking”, says Advait Sarkar, a researcher.
From jobs to slop, AI over-automation is at the heart of all our fears. But how do we ensure AI augments our thinking, and how does that differ from mere task acceleration?
Use your deep mind
Microsoft’s Tools for Thought project tries to unpack the problem. Focusing on both purpose (“work over time”) and action (”work in the moment”), the team explores ways to use AI to deepen our thinking, not outsource it.
The key advice is to use AI to understand the task at hand better, rather than to complete it faster. We can all spend more time refining our questions and creating higher-quality work and less time churning out unoriginal stuff. To use Anthropic’s terms, the aim is not to “automate” what we know, but to “augment” our sphere of knowledge.
Like most technologies, AI adoption will begin with streamlining existing workflows. But as adoption spreads widely, more and more companies will need to pull on three levels to encourage “augmentation” workflows: Guidelines, Education and Design.
The three levers: Guidelines, Education, Design
Start with guidelines. Many companies are now issuing acceptable use policies for AI systems, often aligned with regulations or accreditation standards. That will help avoid data leaks and security risks; however, companies should also encourage teams to use it widely, at a pace that allows exploration and learning.
As adoption spreads, education is becoming increasingly important. Understandably, the initial focus is on efficiency: how to design prompts to get to the desired outcome faster. However, training must also cover how we collaborate with AI to challenge assumptions and reframe the problem before we rush into the solution.
And finally, AI is a tech that begs to be designed. The prompt-response “chat” interface is overly simplistic and lacks depth. To encourage human-AI collaboration, we must embed it in all our tools and allow it to use context from different workflows. Some pushback would be helpful too: I would rather my prompt be challenged than labelled invariably “terrific”.
Smarting up with AI
As adoption spreads, companies will inevitably use Guidelines, Education and Design to ensure teams use AI safely and productively. Some, recognising the subtle difference between automation and augmentation, will go further, helping their teams use AI to learn, challenge and advance their thinking.
AI is powerful, but it is no magic wand. To create value, professionals should use it to reframe the problem, not just produce unoriginal, bland results at speed. Augmenting human thinking is the ultimate transformation.