There’s something wonderfully human about Alan Trefler’s introduction to artificial intelligence.
He arrived at Dartmouth in 1973 intending to study English, with a passion for competitive chess. Chess led him to the computer center, where researchers were experimenting with teaching computers to play the game. They understood computing. Alan understood chess. His path changed.
Listening to Alan reflect on that experience at Dartmouth’s AI@70 conference, I kept coming back to the question at the center of the event:
As AI becomes capable of more, what must remain human?
Seventy years ago, the big question was what computers could do.
Today, that list is getting rather long.
AI can interpret, create, recommend, predict, and, increasingly, act. Yet the businesses putting all this intelligence to work haven’t suddenly become simpler. They still run on people, processes, decisions, rules, systems, and plenty of exceptions.
So perhaps the more interesting question now is less about what can AI do and more about what should we ask it to do.
Intelligence still needs intent
For enterprises, AI rarely operates in isolation.
Someone still needs to decide where it should act, what information it should use, what boundaries it must respect and when a person should step in. Most importantly, someone needs to decide what outcome AI is there to achieve.
That’s not a shortcoming of AI. It’s why human judgment still matters.
There’s an interesting echo with Alan’s Dartmouth story. The researchers had computing power. Alan understood the problem they were trying to solve. Progress came from bringing the two together.
The machines have changed considerably since then. The need for human intent hasn’t.
“It’s one thing to co-work or collaborate with the AI to get some additional views and ideas, but the ultimate decision as to what your opinion is needs to be yours.”
“It’s one thing to co-work or collaborate with the AI to get some additional views and ideas, but the ultimate decision as to what your opinion is needs to be yours.”
Structure makes intelligence predictable
As AI becomes more powerful, it becomes easier to add more of it. More models. More agents. More decisions. More actions across businesses that weren’t simple to begin with.
Without structure, more AI can mean more decisions and actions without clear direction or governance.
At Pega, we know structure is what makes AI predictable. It connects intelligence to the context, processes and guardrails of the business, creating clarity around where AI acts, what guides it, where human judgment matters and what outcome it needs to deliver.
That’s the idea behind predictable AI.
Predictable AI doesn’t mean predicting every answer a model will produce. It means creating the structure, context, and guardrails that determine how AI operates within the business, where it can act, and when human judgment is required.
Structure doesn’t limit what AI can do. It gives businesses a way to use that power with purpose.
That matters even more as AI becomes increasingly autonomous. Enterprises don’t need more intelligence piled onto an already complicated technology landscape. They need a way to harness it, with the structure to act, adapt, and move forward with confidence.
“You don't build something [so] powerful and just set it loose. People need to understand that if you send off these autonomous little robots to go do things, the instructions you give them are pretty important.”
“You don't build something [so] powerful and just set it loose. People need to understand that if you send off these autonomous little robots to go do things, the instructions you give them are pretty important.”
So what must remain human?
There probably isn’t one answer. But listening to Alan, a few things stood out.
Intent. Judgment. Context. Accountability.
And perhaps the most basic question of all: What problem is actually worth solving?
AI will keep getting smarter. What looks extraordinary today will eventually look ordinary.
The harder question is what we choose to do with all that intelligence.
Seventy years ago, we were asking what computers might someday be capable of doing. Today, we have extraordinary intelligence at our disposal.
What remains human is deciding what we want to do with it.
Get that right, and we can put AI to work with greater confidence as we continue to build for what’s next.
Fittingly, I shaped this post with a little help from an AI friend. It helped sharpen the thinking. I decided what was worth saying.
