Over the past two years, we’ve heard a very strong message from the industry:
“AI can do everything.” “Just prompt it.” “Workflows are no longer needed.”
But after watching the latest Salesforce TDX keynote, I believe something important has changed.
Not loudly. Not explicitly. But fundamentally.
From AI Magic to AI Reality
The early narrative around LLMs suggested a future where:
- rules would disappear
- workflows would become obsolete
- systems would be replaced by intelligent agents
In simple terms:
We thought we were moving from “if/else” to “just prompt it.”
But what Salesforce showed at TDX tells a different story.
The Return of Determinism
Underneath the new terms like Agentforce, Headless 360, and Agent Script, there is a clear pattern:
- defining conditions
- controlling flows
- structuring decisions
- testing outcomes
- managing state transitions
In other words:
We are bringing back logic.
Not because we want to, but because we have to.
Why This Matters
LLMs are powerful, but they are not deterministic.
In enterprise environments:
- decisions must be traceable
- actions must be predictable
- systems must be auditable
You cannot run finance, HR, or operations on:
“the model usually gets it right.”
So what do we do?
We don’t replace systems with AI. We wrap AI with control.
But There Is a Bigger Question
If we now accept that:
- AI needs guardrails
- AI needs control flows
- AI needs deterministic layers
Then we should ask a harder question:
How do you manage this on top of hundreds of disconnected applications?
Because every additional app means:
- another data model
- another workflow
- another permission structure
- another set of edge cases
At that point, the problem is no longer AI.
It becomes:
orchestrating complexity.
Agents Are Not a Silver Bullet
There is a growing belief that agents will sit on top of everything and make it all work.
But in reality:
- agents don’t eliminate complexity
- they expose it
And once you introduce deterministic control (which we now know is necessary), that complexity increases even further.
What Actually Makes Sense
Instead of:
- replacing systems with AI
- or putting agents on top of chaos
A more practical model is emerging:
- humans continue to work within systems
- systems provide structured execution
- AI runs continuously in the background
- AI analyzes, predicts, and prepares actions
- humans stay in control
The Real Shift
This is the key takeaway for me:
AI is not replacing systems. It is forcing us to rethink how systems should work.
And more importantly:
AI does not reduce the need for structure. It increases it.
Final Thought
The industry started with:
“AI will eliminate rules.”
Now we are moving toward:
“AI needs rules to be useful.”
That’s not a step backward.
It’s a step toward reality.
If anything, the TDX keynote shows this clearly:
The future is not AI vs systems. It is AI operating within well-structured systems.