As marketing software moves from recommending actions to taking them, measurement becomes part of governance rather than a report produced at the end. That is the central argument AppsFlyer product executive Ziv Peled shared in an authorised client briefing to SEA Connect.
What happened
Peled argues that an autonomous system can only make reliable decisions when the underlying event and attribution data is trustworthy. Poor data does not become better because an AI system uses it. It can instead make a weak decision faster and repeat it at scale.
A practical control layer should distinguish human activity from machine activity, preserve the evidence behind a decision and make the result testable. Teams also need clear limits on what a system can change without human review.
Why it matters
For APAC and Southeast Asian commercial teams, the issue is especially relevant across fragmented markets, channels and privacy rules. A regional model can hide country-level differences unless measurement is designed to keep those differences visible.
What to watch next
This is an attributed expert view, not independent evidence of market-wide adoption or performance. SEA Connect excluded an unsupported percentage supplied in the pitch because no methodology, sample, period or geography was provided. Buyers should ask for those details before relying on any benchmark.
Source note
This article is an attributed viewpoint adapted from an authorised 12 August client briefing supplied to SEA Connect. It does not claim independent evidence of market-wide adoption or performance.