›

31 August 2026

Client work

Google makes video ads from a storyboard

Demand Gen announcement covering Asset Studio video creation

A woman at a desk holds a printed portrait next to a laptop showing a grid of video frames, with more prints laid out in front of her

What happened

Google announced on 27 August that Multimodal Video Creation in Asset Studio is generally available. Advertisers can move from storyboarding to horizontal and vertical assets in one workflow. The feature is available across Demand Gen inventory.

The tool removes production steps, not the decisions before and after them. A weak storyboard now becomes more finished work, faster. Direction and review carry more of the value.

What gets faster

The tool makes agreed plans into multiple video formats faster. It does not choose the idea, audience or evidence, and every format still needs review.

What still needs judgement

What still needs judgement: Choosing the idea, audience and evidence before production starts. Review every format before it reaches a live campaign.

Why it matters here

Large agencies can split creative decisions, production and checking across specialists. A small team may give all three jobs to one person. Saved time only counts after the outputs are checked.

The awkward part sits in the estimate. Charging only for output rewards the tool and discounts the thinking that made the work usable. Price the work before and after generation.

Do this by Friday

The creative lead tests a small campaign with limited exposure. Allow 45 minutes to compare the generated formats with the approved storyboard.

01

Choose one low-risk campaign with an approved storyboard.

02

Generate one horizontal and one vertical version in Asset Studio.

03

Review both versions against the brief, brand and platform requirements.

04

Rewrite the estimate around creative decisions and final checks.

Try it now

Compare these generated video ads with the approved storyboard and campaign brief. List every mismatch in message, sequence, brand, evidence and format. Separate fixes required before launch from optional improvements.

Also in that Monday’s email

Worth watching: Europe wants stronger AI copyright records.

Safe to ignore: Cursor's model dispute does not affect agency work.

Run it properly

The Client Report Playbook

The full job, stage by stage, if this week’s steps open a bigger conversation.

Members get every briefing four weeks earlier.

They had this one while there was still time to act on it. In AI, a month is a long time: this week’s briefing is already with them.

google-makes-video-ads-from-a-storyboard

›

24 August 2026

Client work

ChatGPT Ads are coming to Europe

Product announcement covering 31 European markets and access routes

A phone showing a chat with a sponsored card, inside a ring of twelve lime-green stars on a black background

What happened

OpenAI said on 18 August that ChatGPT Ads would expand to 31 European markets the following week. Initial access runs through its Ads Solutions team, agency partners and technology partners. Ads Manager self-service is due later this summer.

The channel is arriving before the simplest buying route. That gives agencies time to set a client test, measurement plan and stop condition before low-friction access widens demand.

Available now

Direct access, conversion optimisation, geo-targeting and custom audiences are available now. Self-service buying through Ads Manager is not, and OpenAI has not named a date.

Not available yet

Not available yet: Self-service buying through Ads Manager. OpenAI says it will follow later this summer, without naming a date.

Why it matters here

Large agencies can pass a new channel to specialist teams. Smaller agencies need one person to define the test and explain the limits. Waiting for self-service leaves that decision to client pressure.

The risk is not missing launch week. It is spending before the agency has agreed what a useful result looks like. Early access is a reason to prepare, not proof of performance.

Do this by Friday

The media lead writes one test brief for a suitable client. Allow 45 minutes, without committing spend.

01

Choose one client whose audience already uses ChatGPT to compare options.

02

Define one decision-stage use case and the action the ad should support.

03

Set the budget ceiling, success measure and stop condition before requesting access.

04

Record what remains unknown, including format performance and self-service timing.

Try it now

Create a one-page test brief for [client] using ChatGPT Ads to support [decision-stage use case]. Include the audience, offer, budget ceiling, success measure, stop condition and questions that need evidence before launch.

Also in that Monday’s email

Worth watching: creators are entering brands' AI search plans.

Safe to ignore: new API privacy controls do not change ChatGPT.

Run it properly

The Client Report Playbook

The full job, stage by stage, if this week’s steps open a bigger conversation.

Members get every briefing four weeks earlier.

They had this one while there was still time to act on it. In AI, a month is a long time: this week’s briefing is already with them.

chatgpt-ads-are-coming-to-europe

›

17 August 2026

Compliance

Client contracts are banning AI

Original interview with Haley MacLean of Voyer Law

A folded paper note tucked into a locked metal security gate

What happened

GamesRadar+ published an interview with Haley MacLean, a video-game lawyer, on 12 August. She said no-AI wording has moved from some publishing agreements to routine boilerplate. Her view is based on the contracts she reviews.

This is one lawyer’s experience in game publishing, not a survey of agency contracts. Still, contracts set the boundary before a team starts work. Waiting until delivery leaves the decision to someone else.

What a clause covers

A clause can ban named AI uses or require written consent. Silence leaves the question open until a client asks. That is a poor time to find different expectations.

What silence means

What silence means: It leaves the question open until a client asks. That is a poor time to discover different expectations.

Why it matters here

Large agencies have counsel and procurement teams to catch contract changes. A small agency often finds the clause while work is already moving.

The risk is not every AI use. It is promising work before anyone has agreed what is allowed. This is not legal advice. Ask a lawyer to review high-value contracts.

Do this by Friday

The account lead lists each current AI use in client work. Give it one hour, then decide which uses need a client conversation.

01

List every AI use in current client work.

02

Mark the uses clients may object to.

03

Check live agreements for AI wording.

04

Ask for written boundaries before new work starts.

Try it now

Here is our client agreement and a list of AI uses in this project. Identify any clause that restricts, requires consent for, or leaves unclear those uses. Write five plain-English questions for our lawyer before we sign.

Also in that Monday’s email

Worth watching: Marketing agents promise a full week in an hour.

Safe to ignore: No verified item this week.

Run it properly

The Client Report Playbook

The full job, stage by stage, if this week’s steps open a bigger conversation.

Members get every briefing four weeks earlier.

They had this one while there was still time to act on it. In AI, a month is a long time: this week’s briefing is already with them.

client-contracts-can-ban-ai-before-you-start

›

10 August 2026

Compliance

Paid AI music is not cleared music

Munich Regional Court ruling, GEMA v Suno

A pair of black over-ear headphones resting on the corner of a wooden desk

What happened

The Munich Regional Court ruled that Suno infringed copyright by training on, storing and reproducing music controlled by GEMA, the German collecting society. Three separate acts, and the court went against the tool on all three. What it never examined was anyone's subscription. The question was where the material came from, not who had paid to be there.

That distinction is the whole story for an agency. A subscription is an agreement with the vendor about using their software. The people who own the underlying music were never part of it, and the ruling says their rights survive it.

What the plan covers

A paid plan gives you access to the tool and commercial use under the vendor’s terms. It does not prove the training music was licensed or cover a rights claim after publication.

What it does not

What it does not: Any warranty that the training data was licensed, or cover if a rights holder comes after the track. That risk sits with whoever published the work.

Why it matters here

Large agencies keep legal review between the studio and the client. A ten person agency does not. Whoever made the track also cleared it, in the same afternoon, with nothing written down.

The exposure is rarely a big brand film. It is the music bed under a social edit, made in ten minutes, still live on a client channel two years later, with nobody able to say which tool made it or on which plan.

Do this by Friday

Log the tool, the licence and the edits for every AI-made track that has gone out this year. An hour, once.

01

List every AI-made track that has shipped this year, with client, campaign and date.

02

Record the tool and the plan for each, and screenshot the terms.

03

Note what a human changed, and put a name against each track.

04

Agree the one sentence you would say to a client, and keep the record behind it.

Try it now

Here is the licence page for our AI music tool. In plain English: what does it say about commercial use, who owns the output, and what happens if a rights holder complains?

Also in that Monday’s email

Worth watching: anti-AI design is becoming a selling point.

Safe to ignore: Astra's delay changes nothing for agencies.

Run it properly

The Client Report Playbook

The full job, stage by stage, if this week’s steps open a bigger conversation.

Members get every briefing four weeks earlier.

They had this one while there was still time to act on it. In AI, a month is a long time: this week’s briefing is already with them.

paid-ai-music-is-not-cleared-music

›

3 August 2026

Client work

AI sped up the draft. Approval did not.

Survey of 333 enterprise marketing teams

A tidy office desk with a stack of papers, lit by afternoon sun through linen curtains

What happened

Knak surveyed 333 enterprise marketing teams and found that 85% of them missed at least one launch date last year. The interesting part is not the number, it is the reason: asked what held the work up, teams pointed at approvals, production and coordination. Running out of ideas came nowhere near the top.

That matters more this year than it did two years ago. AI has taken a large bite out of exactly the part that was never the problem. Drafting got faster. Nothing else in the chain moved, so the queue formed somewhere less visible.

What got faster

AI sped up first drafts, variants, resizes, subject lines and alt text. It did not shorten approvals, legal sign-off, handovers, feedback rounds or scheduling.

What did not

What did not: Approvals, legal sign off, asset handovers, feedback rounds, scheduling. Anything that needs a second person to stop what they are doing and look.

Why it matters here

Large teams absorb a duplicate approval step, because someone there is partly paid to chase it. A ten person agency has no such person. The same step lands on the account lead, who is also running two other campaigns that week.

That is why the small agency version of this looks different. It is rarely a missing tool. It is one handoff that exists because it always has, and nobody has had a quiet hour to ask whether it still needs to.

Do this by Friday

The account lead maps one campaign's approvals and removes one duplicate handoff. About an hour, once, and it is easier on a campaign that has already shipped.

01

Map one campaign's approvals end to end: every point where the work stopped and waited.

02

Find the handoff that happens twice, at draft and again at final.

03

Remove one of them and say so out loud, or it is back within a month.

04

Put the new path in the project template, in writing.

Try it now

Here is the approval path for our last campaign, step by step. Point out any step where the same thing is checked twice, and tell me which one is safest to drop.

Also in that Monday’s email

Worth watching: Meta put research and decks inside one tool.

Safe to ignore: predictive targeting did not arrive with proof.

Run it properly

The Client Report Playbook

The full job, stage by stage, if this week’s steps open a bigger conversation.

Members get every briefing four weeks earlier.

They had this one while there was still time to act on it. In AI, a month is a long time: this week’s briefing is already with them.

ai-sped-up-the-draft

›

27 July 2026

Compliance

AI disclosure moves from policy chat to client work

Guidelines on transparency obligations, AI Act Article 50

An empty wooden frame hanging on a plain wall in a warmly lit, empty room

What happened

The European Commission published its guidelines on Article 50 of the AI Act, ten days before the transparency duties started on 2 August. The guidelines set out what has to be disclosed when content is generated or altered by AI, and who carries the duty: the provider of the system, and the deployer who puts the output in front of an audience.

Deployer means you. If AI made or altered work that goes out for a European client, the obligation sits with whoever published it, not with the tool that produced it.

What needs disclosing

Disclose synthetic public-facing images, audio, video and text. Research, rewritten drafts, resizes and work that stays inside the agency do not usually require disclosure.

What does not

What does not: Assistive use that does not change the substance: research, drafts a person rewrites, resizes, alt text, and anything that never leaves the agency.

Why it matters here

Large agencies handle this in legal review. A small agency does not have that step. If disclosure lives only in a policy document, it gets remembered on the slow campaigns and forgotten on the ones that go out on a Friday.

The fix is not another policy. It is one line in the campaign brief, filled in at the start, while someone still knows what was used and why. This is not legal advice. On high-value work, check the wording with a lawyer.

Do this by Friday

Add an AI-use and disclosure check to every live campaign brief. Ten minutes per brief, and the account lead owns it.

01

List the live campaigns with European audiences publishing this quarter.

02

Add two fields to the brief: what AI was used for, and whether it needs disclosing.

03

Agree the wording once, cleared with the client, not campaign by campaign.

04

Raise it before the client does. Disclosure they discover is a problem.

Try it now

Here is our campaign brief template and a description of how we used AI on this campaign. Draft the disclosure line we would put in the brief, and the one that would sit next to the published work, in plain English.

Also in that Monday’s email

Worth watching: clients are buying less agency breadth and more AI readiness.

Worth watching: review stars need cleaner proof now.

Safe to ignore: the OpenAI ad network rumour is not a media plan.

Run it properly

The Client Report Playbook

The full job, stage by stage, if this week’s steps open a bigger conversation.

Members get every briefing four weeks earlier.

They had this one while there was still time to act on it. In AI, a month is a long time: this week’s briefing is already with them.

ai-disclosure-moves-to-client-work

›

20 July 2026

Client work

Publicis says most of its revenue is AI. Stop hiding yours

First half 2026 results

A long, empty boardroom table with black chairs beside floor-to-ceiling windows over a city

What happened

Publicis reported its first half results and told the market that most of its revenue now runs through AI-enabled work. In the same week Meta started adding a disclosure tag to ads it detects as made or edited with AI, including work that came out of third-party tools.

Between the two, the quiet phase ended. The largest holding company is selling AI as the reason to buy, and the largest ad platform is labelling it whether you mention it or not.

What clients can already see

Clients can see disclosure tags, their own tool subscriptions and software costs. They cannot see who checked the work, what was rejected or why the final version shipped.

What they still cannot see

What they still cannot see: Who checked the work, what was rejected, and why the version they got is the one that shipped. That is what the invoice is actually for.

Why it matters here

A ten person agency competing on "we use AI" is competing on the one thing a client can buy for €20 a month. The value moved to judgement and accountability. Name it out loud, or it reads as overhead.

The proposal is where that gets decided. Not a line saying you are AI-enabled, but a short section saying which parts AI handles, who reviews them, and what happens when it is wrong.

Do this by Friday

Rewrite one section of your standard proposal so it names the AI in the process. An hour, once, and it carries into every pitch after it.

01

Pick your most-used proposal, the one that goes out three times a month.

02

Name what AI does, in plain words a client can picture.

03

Name who checks it, by job title, at a named point in the process.

04

Say what the client is paying for: judgement, accountability and control.

Try it now

Here is the scope section of our standard proposal. Rewrite it so it names which steps use AI, who reviews each one, and what the client is paying us for. Under 200 words, no jargon.

Also in that Monday’s email

Worth watching: Meta will label your AI ads for you.

Worth watching: Figma puts small website fixes in designers' hands.

Worth watching: AI users are checking your work twice.

Run it properly

The Client Report Playbook

The full job, stage by stage, if this week’s steps open a bigger conversation.

Members get every briefing four weeks earlier.

They had this one while there was still time to act on it. In AI, a month is a long time: this week’s briefing is already with them.

publicis-says-most-of-its-revenue-is-ai

›

13 July 2026

Positioning

Being ahead on AI only pays if clients can follow

The AI gap between agencies and their clients is widening, not closing

A wide stone staircase rising beside a tall arched window in low sunlight

What happened

Agency bosses came back from Cannes and told Digiday the same thing: the gap between what agencies can do with AI and what their clients understand is getting wider, not narrower. None of them described that as winning. They described a translation problem that has quietly moved to the agency's side of the table.

The same week gave the gap a shape. Time and The Economist began building stripped-back versions of their sites for AI agents to read. OpenAI extended its ChatGPT ads pilot to France, Germany and Ireland. The EU pushed its high-risk AI Act duties out to December 2027. Three moves a busy client would have missed, and each one changes something in their plan.

Being ahead

Being ahead means knowing the tools and testing them on real work. Being followed means clients can explain your reasoning to their board without you there.

Being followed

Being followed: The client being able to repeat your reasoning to their own board without you in the room. That is the part that gets renewed, and it does not happen on its own.

Why it matters here

Large agencies can leave the AI conversation in a separate innovation deck, because someone else runs the account day to day. In a ten person agency, the person who understands the tools is also the person on the client call. The explaining happens there or it does not happen at all.

The risk is not that the client leaves for an agency with better tools. It is that they cannot explain internally why your work costs what it does, so it gets trimmed in a budget conversation you were never invited to.

Do this by Friday

Pick one client and send them a plain one-page note on where AI sits in their work and where it deliberately does not. Half an hour to write, and it is the note a renewal conversation gets built on.

01

Pick the client most likely to ask, not your easiest relationship.

02

Name where AI is used, in their language.

03

Name where you choose not to use it. That builds trust.

04

Give them one thing to watch, and what you would do about it.

Try it now

Here is what our agency does for this client and where we use AI in that work. Write a one-page note for their marketing director explaining where AI sits, where we choose not to use it, and why. Plain English, no jargon, under 400 words.

Also in that Monday’s email

Worth watching: publishers are building a second web for AI agents.

Worth watching: ChatGPT ads reach France, Germany and Ireland.

Safe to ignore: the EU's delay is not permission to relax.

Run it properly

The Client Report Playbook

The full job, stage by stage, if this week’s steps open a bigger conversation.

Members get every briefing four weeks earlier.

They had this one while there was still time to act on it. In AI, a month is a long time: this week’s briefing is already with them.

being-ahead-on-ai-only-pays-if-clients-can-follow

More open briefings

More open briefings

A phone showing a chat with a sponsored card, inside a ring of twelve lime-green stars on a black background

24 August 2026

·

Client work

ChatGPT Ads are coming to Europe

2 min read

A folded paper note tucked into a locked metal security gate

17 August 2026

·

Compliance

Client contracts are banning AI

1 min read

A pair of black over-ear headphones resting on the corner of a wooden desk

10 August 2026

·

Compliance

Paid AI music is not cleared music

4 min read