The New Programmatic Power Couple: Humans & AI
Programmatic can handle the busywork. Humans still need to call the shots.
In programmatic, platforms can already support tasks such as bid adjustments, pacing, budget allocation, creative rotation and performance monitoring. That can remove some of the repetitive work, but it does not remove the need for active oversight.
In our latest Performance+ webinar, What’s New in Programmatic? The Path to Human-Supervised Autopilot, Dhiyay Chohan our Managing Director at FYND was joined by Rona Einat, Senior Product Manager from AppsFlyer and Alexei Moltchan, VP of product from Dataseat, to discuss what automation can already handle, what marketers still need to own, and how brands can prepare for an increasingly AI-led future. Here are the big takeaways.
Humans to take over the direction in Programmatic
Give the system a clear goal, and it can pull the tactical levers faster than a human team ever could. The tricky part is deciding what that goal should be.
Should the system optimise towards installs, registrations, purchases, retention or lifetime value? Does the selected in-app event genuinely represent a valuable customer? Is the campaign driving incremental growth, or simply claiming credit for conversions that would have happened anyway? These are still very human decisions. AI can automate execution, but marketers still need to set the direction. And when the direction is wrong, automation simply helps the campaign move in the wrong direction faster.
Your KPI is your instruction manual
It is tempting to begin every AI conversation with the same question: What can we automate? Here is a better starting point: What outcome are we trying to improve?
The KPI determines how much freedom an automated system should have.
For a relatively straightforward objective, such as installs or CPI, AI can manage a significant amount of the workflow. It can adjust bids, shift budgets, test publishers and explore new pockets of inventory. But objectives such as retention, subscriptions, revenue, LTV and incrementality require stronger signals and longer feedback loops. They also need more human judgement around what the model should learn from.
The changing role of marketers
Marketers will spend less time pushing buttons and more time deciding which buttons should exist in the first place.
Two skills become especially important:
- Prioritisation: Marketers need to understand which signals matter, where to invest, what to test and what can safely be ignored.
- Interpretation: Teams need to turn performance data into decisions, explain why something happened and decide what should happen next.
The line between AI assistance and AI decision-making will also continue to move.
High-frequency decisions that can be corrected quickly are natural candidates for automation. Think bid changes, pacing adjustments, creative rotation, performance monitoring and anomaly detection.
Decisions with larger or less reversible consequences need stronger guardrails. Changing the campaign goal, reallocating a significant budget, removing an entire creative direction or shifting the channel mix should still involve human approval.
More creative ≠ Better creative
Generative AI has made it possible to produce hundreds of creative variations quickly.
When creative production becomes almost infinite, the value sits in what each asset teaches the buying system. Which message attracts higher-quality users? Which format performs better in a particular app environment? Does a creative with a weaker click-through rate generate stronger post-install behaviour? Is an asset genuinely fatigued, or does it still have room to scale?
Every creative variation should be treated as a hypothesis, rather than another asset added to the pile.
AI agents can help close that gap by continuously monitoring creative performance, spotting fatigue and surfacing opportunities. The aim is to give teams the information they need while there is still time to act on it.
How important is Data Quality?
AI is an amplifier. Give it high-quality data and it can make insights faster, clearer and more actionable. Give it incomplete attribution, misconfigured events or a KPI that does not reflect business value, and it will optimise towards the mistake with impressive precision.
That is how you end up with a perfectly optimised inefficiency.
Data quality is therefore becoming one of the biggest competitive advantages in programmatic.
Most advertisers have access to similar platforms, algorithms and creative tools. The difference will increasingly come from the quality of the information those systems receive. First-party data plays a major role here. Retention, purchase quality, churn, revenue and LTV signals help automated systems understand what valuable growth actually looks like. But first-party data is only useful when it is clean, structured and connected to the correct business outcome.
AI cannot fix the product
AI cannot repair a weak value proposition, confusing onboarding, poor retention or unclear monetisation.
It also cannot decide how much a business should be willing to pay for growth, which markets deserve investment or how much uncertainty a team should accept in its measurement. Those decisions require commercial context. If an app is struggling to retain users, AI may help identify the problem sooner. It cannot solve the underlying product experience by itself.
What should advertisers invest in now?
- Data foundation: Before deploying more agents, integrations and AI tools, brands need clean, governed and AI-ready measurement. Every automated decision will sit on top of that layer.
- Operating model: Who owns the AI workflow? Who reviews recommendations? When should an agent be given more autonomy? Without clear ownership, AI can become another source of alerts, dashboards and noise rather than a way to simplify the work.
- Goal-setting: Marketers will need to become much better at defining what winning looks like. The clearer the objective, the easier it becomes to trust a system to work towards it.
Want to hear the full conversation? Watch the webinar, What’s New in Programmatic? The Path to Human-Supervised Autopilot, on demand.
Talk to Us
