AI for
Marketing
Agencies
Generative AI is going to change agency economics.
Store the client rules
Prepare the work
The account team approves
Your team works faster without losing the client context.
How are agencies using GenAI right now?
At the moment it’s being used to reduce the cost of making things: creating assets, generating insights from research, drafting reports for clients etc. Agencies are still selling time. Which means juniors’ time. But the number of hours required to make things is going to change. That will be difficult to hide. Retainers are already suffering with agencies being asked to concede margin at renewal as hourly charges become harder to defend.
What can be done about it?
Deck factory. Get an RFP / brief as input. Produce a deck using previous agency work and brand assets. Output is branded deck ready for seniors to edit, not create from scratch.
Copy generation. Feed in previous agency / client assets to create approved language. Use a simple retrieval tool. Get first drafts of copy. Assess if a finetuned model would improve results and build one if necessary.
Creative generation. Identify tools and workflows such as Midjourney for stills and Runway for video. Integrate brand assets, compliance, licensing, costings and approvals. Use to reduce number of concepts / options.
Client reporting. Extract data from ads and analytics, combine with other inputs, generate a report that can be checked and sent by the team. Improve over time with finetuning to reflect each account team’s voice. Imagine arriving into the office on Monday morning to find your client reports written overnight by an agent, waiting for you to check.
Client meetings. Agree to record meetings with clients. Agents summarise calls, log decisions and next steps and produce first draft of action items for teams to agree and distribute.
Vu has built these kinds of agents and has experience with a range of internal AI products that will be useful for your agency. We can help you implement them. We are an agency ourselves so understand what’s important.
Frequently asked questions
How will juniors’ roles change? Less making stuff, more reviewing it, learning about the client and industry. Hopefully, a path into craft and taste rather than routine first drafts.
Who is responsible if an AI breaks advertising rules? The agency, just like it is now. You can’t outsource responsibility for misleading or illegal work.
Who owns the output? It will depend on the tool. Generally speaking the agency owns the prompts and code. There are complex considerations around ownership of finetuned weights, reuse, IP in output. We will advise based on the T&Cs of ChatGPT or whichever tool you’re using.
What about the EU AI Act? Your client may have deployer duties. These need to be captured when we scope your workflows, just like other considerations (e.g. privacy impact).
How does pricing work? We work with you to implement agents across your delivery. We help you to redefine your commercial agreements with clients so that you charge for outputs and outcomes. We charge a fixed fee for the first stage, an AI audit. Each agent is scoped and costed before we build it. We then charge a fixed fee for each phase of work.
What happens in an AI audit? We look at a couple of example retainers. We understand what time is being charged for and map the delivery. We collect agency and client assets, brand guidelines, account plans, briefs etc and bring them into an internal repository with limited permissions. We prioritise an opportunity and produce a plan for your first agent, ready to discuss with your most important clients. We help you to talk to clients about outputs and outcomes and work with your commercial leads to redefine service levels, outputs, charges for reviews, costs of fixing mistakes.
How do I book an AI audit?
Send us two or three retainers you want to understand the economics of (redact anything confidential). Include a description of the process, deliverables, who’s involved and the hours spent. Include an overview of the economics, margins, challenges. Tell us about your ambitions for GenAI. Tell us anything else that might be useful.