There is no shortage of articles telling agencies they should use AI. There is a shortage of articles showing what agencies are actually doing with it.
Here is what we are seeing across the agencies we work with. Not theory. Not predictions. What is happening right now.
Sales and new business
This is where most agencies start, and where the results show up fastest.
- Prospect research before calls. Instead of spending 45 minutes on LinkedIn and a company website, agencies are feeding a prospect’s URL and recent news into Claude or ChatGPT and getting a structured brief in minutes. One agency told us their discovery calls improved overnight because they were walking in with better context than the prospect expected.
- Proposal drafting. The first 80% of a proposal (structure, case study formatting, standard sections) is now AI-generated. The senior team focuses on the strategic positioning and pricing, which is where proposals are actually won or lost.
- Follow-up sequences. Post-pitch follow-ups are being drafted by AI, personalised to what was discussed in the meeting. Not template emails. Genuinely tailored messages.
Delivery and production
This is where agencies are seeing the biggest time savings, but it requires more care.
- Brief structuring. Raw client inputs (rambling emails, call transcripts, scattered documents) are being processed into structured creative briefs. The brief still needs human review, but the assembly work is gone.
- First-draft content. Blog posts, social copy, email campaigns. AI writes the first draft, a human editor refines it. The agencies doing this well are spending more time on the brief and less time on the writing.
- Code generation. Development agencies are using Cursor and GitHub Copilot for boilerplate, component scaffolding, and test writing. Senior developers report spending more time on architecture and less on implementation.
- Meeting summaries. Every client call is transcribed and summarised automatically. Action items are extracted and shared within minutes of the call ending.
Operations
This is the least glamorous but most compounding layer.
- Reporting automation. Monthly client reports that used to take 3-4 hours are being generated in 20 minutes. Data pulled from analytics tools, formatted into templates, with AI-written commentary on trends.
- Resource planning. Agencies are using AI to forecast project timelines based on historical data, flagging potential bottlenecks before they happen.
- Internal knowledge bases. Process documentation, client preferences, project learnings. Instead of living in someone’s head, these are being captured and made searchable.
The pattern
The agencies getting the most from AI share three traits:
- They started with one workflow, not a strategy document. They picked a specific bottleneck and solved it.
- They measure the result. Hours saved, quality improved, team satisfaction. If they cannot measure it, they do not scale it.
- They involve the team early. Not as an announcement, but as a collaboration. The people doing the work know where the bottlenecks are.
The agencies struggling with AI also share a pattern: they bought tools before defining problems, tried to change everything at once, and never measured whether it worked.
This is part of Delivery Notes, a series on implementing AI inside your agency. Subscribe to the newsletter to get new articles weekly.
