Agentic Mechanics 07: Why Attribution Breaks After Resolution

Attribution assumes a choice was made.

That’s the first mistake.

In agentic systems, attribution breaks the moment choice disappears.

Attribution Is a Discovery-Era Tool

Attribution exists to answer a specific question:

“What influenced this decision?”

That question only makes sense when:

• alternatives were actively evaluated

• signals competed

• timing mattered

• persuasion was possible

Attribution belongs to markets in motion.

Agents do not operate in markets once resolution stabilises.

What Agents Actually Do Instead

When an agent resolves a task class, it does not:

• re-evaluate inputs

• weigh channels

• assign influence

It reuses a prior resolution.

No fresh decision occurs.

No influence is processed.

No causal chain is recomputed.

So there is nothing to attribute.

Why Attribution Models Go Dark

After resolution:

• clicks don’t matter

• impressions don’t matter

• touchpoints don’t matter

Not because they failed —

but because they are no longer consulted.

Attribution systems keep running.

They just stop describing reality.

The Silent Shift From Cause to Condition

Before resolution:

• actions cause outcomes

• inputs influence decisions

• optimisation changes behaviour

After resolution:

• conditions determine reuse

• stability beats persuasion

• change requires environmental shock

Attribution tracks causes.

Agents operate on conditions.

That mismatch is fatal.

Why “Last Touch” and “Multi-Touch” Both Fail

Last-touch assumes recency matters.

Multi-touch assumes accumulation matters.

Agents care about neither.

They care about:

• trust already earned

• coherence already verified

• risk already minimised

When an agent reuses a default, the entire interaction is conditional — not causal.

Nothing “led” to the decision.

The decision was already there.

Why Procurement Feels Invisible

This is why teams say:

“We don’t know why we keep getting chosen.”

“We can’t tell what’s driving this.”

“Nothing seems to move the needle anymore.”

That’s not confusion.

That’s default lock.

The system is no longer deciding.

It is executing.

The Attribution Paradox

The better you become as a default,

the worse attribution appears.

Because:

• fewer signals are consulted

• fewer comparisons are made

• fewer explanations are generated

Success looks like opacity.

Failure still looks measurable.

Why This Terrifies Governance Teams

Governance relies on:

• explainability

• traceability

• causal justification

But after resolution:

• decisions are path-dependent

• explanations are post-hoc

• behaviour is inherited, not reasoned

The system can tell you what it did.

It often cannot tell you why — because “why” no longer exists.

Attribution Dies Before Anyone Notices

There is no clean breakpoint.

Attribution doesn’t fail loudly.

It just slowly decouples from outcomes.

Dashboards still update.

Reports still generate.

Insights still circulate.

They just stop being relevant.

Why This Is Misdiagnosed as a Data Problem

Most organisations respond by:

• adding more tracking

• stitching more sources

• building better models

But the issue isn’t missing data.

It’s that the decision phase has ended.

You’re instrumenting a process that no longer runs.

The New Question Attribution Cannot Answer

Attribution asks:

“What caused this choice?”

Agents operate on:

“What would increase risk if changed?”

Those are fundamentally different logics.

One looks backward.

The other looks sideways.

What Replaces Attribution

Not better attribution.

Resolution diagnostics.

Instead of asking:

• What influenced the decision?

You must ask:

• When did the system stop reconsidering?

• What conditions must change to reopen evaluation?

• Which dependencies are now implicit?

• Where is reuse assumed rather than chosen?

This is not marketing analytics.

It’s system forensics.

Why This Changes Strategy Forever

If attribution is broken:

• optimisation loops collapse

• campaign thinking degrades

• channel strategies flatten

The competitive surface moves upstream:

into coherence, ownership, and default formation.

The Seventh Quiet Test

Agentic Mechanics 07 asks:

“If attribution disappeared tomorrow, would your position change?”

If yes, you were still competing.

If no, you were already infrastructure.

The Final Line

Attribution doesn’t fail because AI is complex.

It fails because the system has stopped choosing.

And once a system stops choosing,

there is nothing left to explain —

only something left to reuse.