What people don't understand about the future of AI-enabled work
Emergent / revealed ways of working will change the way you work forever…
My team, like most in the AI-enabled industry, are working on Agentic workflows for automating the mundane and slow parts of existing processes. This work takes the form of consulting with an expert, mapping out what they actually do step-by-step, then identifying what steps can be automated by an agent and what parts need human intervention. It's a fairly intuitive automation play, same that has been going on since the beginning of humanity, just at an accelerated pace.
The key concept I want to focus on here is how we're currently just copying existing processes. We are replicating the best practices of what people do today, however, I believe that new modes of working will be unlocked using AI - what I'm calling 'emergent' or 'revealed' ways of working, which people do not fully understand how it will inevitably change how we work.
Most people right now think of AI as "I do A, B, then C - and AI will do parts of A and all of C for me". Makes sense to think this way, however, AI will add an additional depth component to the equation. A better way to think about this is that AI will do ALL of "A, B and C" very quickly (in hours) to a level of fidelity, say 50% towards the finished product, then humans will use their expertise and iterate (along with AI) towards the final 100% finished product. This is why I call this new mode of working emergent, as AI gets you started with something end-to-end, then, through iteration the more optimal solution emerges.
The 50% number is just a finger in the air… eventually AI will do 60, 70, 80%. For some tasks even today it does 90%, others maybe 0%, but on average the capability will increase rapidly as newer models come out. Developing code is already at a high level. This is also known as the "bumpy" frontier of AI, as some tasks, like PHD-level mathematics, the models already well surpass the average person's abilities, but for somewhat 'simple' tasks, they fail consistently. This AI frontier capability will grow and smooth out over time.
The other thing people don't understand about future modes of working is that AI will allow people to do more things, to a higher level of quality, than they currently do. Because of the rapid ideation that can take place, AI can offer 10s if not 100s of new ideas at the start of any discovery process in the same timeframe it takes humans to come up with just a few. Having this variety of ideas early on will help to remove our own biases and aid better outcomes earlier. Humans will still be the judge and cherry-pick their favorite parts of the variations to take forward, but for sure some good ideas from the AI will be ones that hadn't originally been considered. This is also an example of where the indeterminate nature of AI is a value-add.
Here's a concrete example of how this played out in my team. We used this way of working when iterating a prototype for an interface we were developing for an agentic workflow that had many human-in-the-loop interactions. First we had a sketching session, where ideas were sketched out without the aid of AI. Next, we wrote a good, detailed prompt to give the AI context of what we were building and asked it to produce three variant ideas for interfaces. We did not feed in any of the existing human made sketches, just the context of the business problem we were trying to solve.
Using this method, and the aid of Cursor and web chat interfaces, we were able to generate 10s of realistic design variants across OpenAI, Claude and Gemini models, which we could then add to the hand-drawn sketches for consideration. We evaluated all the ideas (human and AI), then picked the parts we wanted to take forward into a prototype. This whole process took a few hours, which would have normally taken a couple of days.
What surprised me was that some of the variants that AI came up with mirrored some of the ideas the humans had before we prompted the AI. It also came up with novel approaches to solving accessibility issues we had already identified. Plus, the whole process was more collaborative and enjoyable - from a human working perspective we were able to stay on task focusing on outcomes and problem solving, rather than passing the work back and forth for endless follow up meetings.
So, I hope this post will spur you on to think about the future of AI-enabled work much differently. Think about not only what AI can automate that you're doing now, but how AI can get you started across the board to a low/mid-level of fidelity, then use human expertise (aided by AI) to iterate. Also think about what you can do with AI that wasn't possible before due to time constraints to enhance quality and better outcomes.
What do you think? What new modes of work will AI bring for your teams?