I was first impressed with AI’s progress when it helped me finish a product build in 3 hours that had taken me 3 weeks a few years earlier.
Of course, not everything was due to AI. Tech stacks have improved. My own ability had inched up a bit too. But it was still impressive.
Similar stories are common from many who use AI for the first time.
Such stories are being used to support doomsday forecasts that AI could replace large swaths of the workforce, turning us humans into mere energy producers for AI’s domination of jobs (thanks, Citrini).
But the reality, I believe, is a little more nuanced.
In my case, AI actually took about 3 mins to get 80% of the way. The remainder then took nearly 3 hours of intense iterations to get it right. After that, I thought, why stop - and so I built 10X my original functionality over the 3 weeks. Net-net, I still spent about 3 weeks. But the boundaries of possibility had shifted.
People who use AI deeply over time start to notice these two effects.
A. Building production-grade AI still requires serious human effort.
It requires a deep understanding of the domain: problems, needs, approaches, social constructs, and regulatory nuances, in order to tune AI outputs to expected performance levels.
AI models are probabilistic. The same query can return very varied responses based on how a question is asked, how the context is provided, the tools chosen, the temperature set and model variant used.
Shaping predictable responses consistently requires a lot of scaffolding and effort. And in domains like finance, predictability is an important requirement to building user trust and managing risks.
This will create a need for new specialist roles as AI gets embedded into multiple domains.
B. When creation becomes effortless, expectations multiply.
Suddenly, you (and others) expect that you will do multiples of what you could do before. Things you once thought were ‘too hard’, ’would take too much time’, or ‘impossible in this timeframe!’ - are all within reach.
You transition from ‘Wish I could build that...’ to ‘What if I built this, or that, or that after, or perhaps a bit more…’ - and you begin to build things that were once only in your imagination.
And imagination, as it happens, is practically unbounded.
As people reach a point where they can build almost anything they imagine, we will see a profusion of new ideas. As people push the boundaries of their imagination, we will need new roles around the table:
- PMs and designers who can reimagine their domains
- Engineers who can build and fine-tune new models and scaffolding, and
- Operating teams who can maintain and optimise these new builds.
This I believe is the opportunity: to push the boundaries of our imagination and build brave new worlds we haven’t imagined yet. This is where new roles will emerge.
