Outlook
Part of Planning connected TV advertising: a UK budget guide
Before you add AI applications to connected TV advertising
How AI applications fit connected TV advertising: what the tools automate, what they cannot fix, and how to cost and govern a six-week pilot in pounds.
What to take away
- Most teams start with the model and the vendor pitch. Start with the decision you cannot make fast enough, then ask whether AI changes it.
- AI in connected TV advertising automates bidding, audience clustering, creative versioning and reporting. It does not fix poor measurement or unclear objectives.
- Budget a pilot against a named baseline. For example, ring-fence £1,200 of an £8,000 monthly streaming budget for a six-week test with a written success measure.
- Governance comes before scale: who approves automated spend, what data feeds the model, how outputs get audited.
Where AI applications actually sit in connected TV advertising
Connected TV advertising runs across broadcaster video on demand, streaming apps and platform inventory, and splits into buying, targeting, creative and reporting. AI tools have landed in all four.
On buying, algorithms adjust bids and budgets across auctions in near real time. On targeting, models group households by inferred interests rather than declared data. Creative tools generate and rotate edits, while reporting tools flag anomalies.
The common thread is faster iteration: a planner who once reviewed weekly pacing now sees intraday shifts. The IAB UK story explains how the trade body has tracked that shift, useful when briefing stakeholders.
What the tools still cannot do
No model knows your commercial objective unless you write it down. If the brief says awareness but the dashboard optimises for clicks, the algorithm will optimise the wrong thing efficiently.
Measurement remains the weak point. Household-level attribution across streaming apps is patchy, and automated reports can present modelled numbers as confidently as observed ones. Ask suppliers to separate the two.
Creative quality is uneven. Generated variations can multiply a weak idea, so review every version before it runs.
A worked example with pound figures
A mid-sized English retailer spends £8,000 a month on streaming ads across two platforms and wants to test AI-assisted creative versioning and bid adjustment over six weeks.
Six-week pilot budget split
- £8,000monthly streaming ad spend
- £1,200ring-fenced for AI test
- £6,800kept running as control
- 15%of budget under test
It ring-fences £1,200 a month, roughly 15 per cent of budget, and keeps £6,800 running as normal for a clean comparison. The test produces eight creative variants instead of two, with bids adjusted inside a capped range.
The team then compares cost per completed view and post-campaign search lift against the control. If the variant is not clearly better on a pre-agreed measure, it does not scale.
Governance before scale
Write down who can approve automated spend changes, and set a ceiling. An algorithm that can move 20 per cent of daily budget without a human is a policy decision, not a setting.
Check the data. If household data feeds the model, your privacy notice and consent position must cover it. To understand broadcaster inventory, the Thinkbox overview of broadcaster video on demand is a sensible starting point.
Keep an audit trail of model changes, versions and approvals, so that when results swing you know whether the cause was the model, the creative or the market.
Skills and team implications
The planner's job shifts towards interrogation: what the model optimised for, on what data, over what window. Our connected TV advertising skills forecast looks at which roles are changing fastest.
Trade bodies publish planning guidance, and the ISBA knowledge hub is a practical reference for advertisers reviewing their buying arrangements.
When to revisit your approach
Revisit when a platform changes its attribution model, when you enter a new market, or when your own measurement improves. Each change moves the baseline a pilot was measured against.
Revisit again if the pilot wins: a tool that works on one objective often needs re-testing on the next. For a wider view of the market, see our connected TV advertising trends and outlook.
Common questions
What is the biggest mistake teams make with AI in connected TV advertising?
Starting with the tool rather than the decision. Without a named decision, the pilot has no success measure.
Can AI replace media planners in connected TV advertising?
No. It shifts the work towards objectives, checking outputs and handling exceptions. Judgement about brand and budget stays with people.
How much should a first AI pilot cost?
There is no standard rate. A labelled example: ring-fence 10 to 15 per cent of monthly streaming spend for six weeks, keeping the rest as a control.
