Remember software dark factories? A new approach to software development where code is not written by a human and code is not reviewed by a human. Maybe the approach will catch on, maybe it won't. At the same time, it is becoming pretty clear that we will produce so much code that people won't be able to read it anyway.
Meanwhile, we are getting requests from companies that are backing the use of AI across entire departments. They already have the licences. Now they want to change how the work gets done.
Dealing with this too? Let's talk.
What stops a dark department from appearing next to the dark factory? A unit that delivers something and has no people? Or an entire dark company? Before we can talk about that, we need to turn down the current hype a little.
Where won't AI replace us?
I was recently at Dave Snowden's Cynefin Base Camp, and among other things his triangle for working with AI came up in the discussion. He describes it in more detail in the article The estuary, overlaid. In it, Snowden asks where AI belongs in the system.
Salience: what AI can handle on its own
AI sees what its training data taught it. It finds the usual things reliably and easily misses the unusual ones. If you extend its data with, say, well-described work procedures, it can deliver part of the work entirely on its own. But it will only see what is explicitly written in those procedures.
Sapience: where AI is a partner
AI is a useful sparring partner when you do sense-making over data. For example, when you work out what follows from twenty interviews you have run. If you leave the conclusions to AI alone, expect some nonsense in there. In a website audit, finding the errors isn't hard. AI can do that. The hard part is setting priorities and cutting what is unnecessary. And that takes human judgement.
Sentience: where the work stays with people
Empathy, touch, chemical signals and your experience. You feel in your body that something is off. AI can simulate empathy, but it is just a machine. In our work, things like stakeholder management, workshop facilitation and research interviews remain fully ours.
Where do your work activities sit in Snowden's triangle? Usually one aspect dominates, and it is worth noticing which. Before you replace a person with AI, you need to thoroughly understand the work they do today. Then you replace the parts dominated by salience and let them do the activities that require sapience and sentience.
"If something is primarily experience-based, there's no way AI gets near it soon."
– Dave Snowden
Weekend reading
Two artists asked people in 11 countries what painting they would like to have at home. Almost everywhere, the same blue landscape came out on top. What else do we all want the same?
