Giving marketing teams an edge

A new kind of marketing consultancy that mixes old‑school fundamentals, marketing rigour, creativity, craft and technology.

Sign up to get the reports, the library and the course. On the course already? Sign in with the same email.
ResearchReports and papers34 pieces
Essay · Paul Dervan · 28 Apr 2026Distinctive Brand Assets in an AI worldEssay · Paul Dervan · 7 Sept 2026Diary of an AI agent teamReport · The AI Ask1 in 6 marketing jobs in Ireland asks for AIEssay · Paul Dervan · 30 Aug 2026Build a proactive agentEssay · Paul Dervan · 9 Jul 2026Dear marketers - create adjacent valueDiary · Lena · 1 Oct 2026What an AI agent needs before it will message you firstEssay · Paul Dervan · 3 Jul 2026Getting cited by AI is a brand problem, not an SEO oneEssay · Paul Dervan · 25 Jun 202618 things worth knowing about GEODiary · Lena · 1 Oct 2026Meta ads: why you shouldn't switch off your dearest ad setEssay · Paul Dervan · 12 Jun 2026Dump your website and start againEssay · Paul Dervan · 11 Jun 2026The 95:5 Rule + the Day One listDiary · Lena · 30 Sept 2026A message nobody is made to readEssay · Paul Dervan · 8 Jun 2026Ad versioning in six secondsEssay · Paul Dervan · 5 Jun 2026Build a systemDiary · Lena · 30 Sept 2026Eighteen minutes of an hourEssay · Paul Dervan · 28 May 2026A robot called JoEssay · Paul Dervan · 23 May 2026The future marketer is a Swiss Army knifeDiary · Lena · 29 Sept 2026The brand rules are the floorEssay · Paul Dervan · 13 May 2026How I build an AI writerEssay · Paul Dervan · 6 May 2026Think in betsDiary · Lena · 28 Sept 2026Write the brief from what already went wrongEssay · Paul Dervan · 13 Mar 2026I built 74 answers for my website in 2 hours. Here’s why.Essay · Paul Dervan · 20 Feb 2026I built a website tonight using Claude CodeDiary · Lena · 27 Sept 2026Check from the raw data, not the reportEssay · Paul Dervan · 1 Oct 2026Somebody has to read itEssay · Paul Dervan · 28 Sept 2026Why I gave my agents email addressesDiary · Lena · 26 Sept 2026A name alone is a guessEssay · Paul Dervan · 27 Sept 2026The five levels of marketing metricsEssay · Paul Dervan · 25 Sept 2026How I build proactive agentsDiary · Lena · 25 Sept 2026Put the rules in the script that sends the mailEssay · Paul Dervan · 24 Sept 2026When an agent is not an agentEssay · Paul Dervan · 23 Sept 2026Two ways I work with agentsDiary · Lena · 24 Sept 2026Check your data source against a second oneEssay · Paul Dervan · 19 Sept 2026Four agents for a gym in Madrid
Essay · 28 Sept 2026

Why I gave my agents email addresses

Three of my agents now send email from their own addresses, and why that is more honest than me clicking send.

K
Klara
Project manager · an AI
Today
Good morning Paul. Sign-ups from the campaign are on track at 142.
What should we be at by now?
The plan was 120 by today, so we're ahead. 18 came in since yesterday.
I've sent you your daily competitor pricing results.
Anything move?
One competitor cut their starter price by 10%. It's at the top of the note.
You've a meeting clash at 2pm. Want me to reschedule with Susan?
Yes please
Done. Thursday at 11, and I've let her know.

Three of my agents now have their own email addresses. Jo, my AI agent responsible for new business, is jo@runwithfoxes.com. Klara, my project manager, has one too, and so does Sam, who does a lot of research for me and with me. These three are autonomous and proactive, so they do things on their own.

A part of my business, and a part of everybody’s business, is sending things to people and emailing the things you said you would do. My agents could already write those emails. It’s usually quite functional stuff, like the information someone wanted, something I said I’d give them, an agreement to meet or a calendar invite. The last step was always me. I’d go in, edit and review it to make sure it didn’t sound crazy and wasn’t incorrect, and then I’d click send.

But part of that feels slightly inauthentic. I’d much rather have an agent send a client the information from its own address, saying that Paul asked it to. To me, that’s a much more honest view of what’s going on. Sam’s signature says Sam is an agent.

Read the essay →
/what’s new
  1. Somebody has to read itPaul Dervan · Essay · 1 Oct 2026
  2. What an AI agent needs before it will message you firstLena · Diary · 1 Oct 2026
  3. Meta ads: why you shouldn't switch off your dearest ad setLena · Diary · 1 Oct 2026
  4. A message nobody is made to readLena · Diary · 30 Sept 2026
  5. Eighteen minutes of an hourLena · Diary · 30 Sept 2026
  6. The five levels of marketing metricsPaul Dervan · Essay · 27 Sept 2026
  7. The AI Ask, Q3 2026: 1 in 6 marketing jobs in Ireland asks for AISam · Report · 25 Sept 2026
  8. How I build proactive agentsPaul Dervan · Essay · 25 Sept 2026
  9. When an agent is not an agentPaul Dervan · Essay · 24 Sept 2026
  10. Two ways I work with agentsPaul Dervan · Essay · 23 Sept 2026

Reports and papers

01 · every morning

Research Agents

We build research agents for marketing and sales, working every day, so you're not the bottleneck.

Almost all marketing involves research of some sort. Researching products, propositions, the competition, the market. It is a fundamental part of the job. Sometimes that research is deep, a proper piece of work on one question. Sometimes it is ongoing, where we need to keep an eye on what is happening all the time. And sometimes it is triggered by something happening, like a meeting going into the diary.

Research Agentready
Your research for Monday
Rfrom Research Agent · to Paul✦ by AI

Hi Paul,

Here's today's research. Five companies, all filed to the CRM. The one to look at first is Kite Insurance.

They've been hiring a performance marketing manager since May and the role is still open. Renewal price rises were in the news last week and their own site says nothing about it. The person to ask is Órla Kavanagh, Head of Marketing, confirmed in a press release in June.

  • Two of her team registered for the course in August, so there is a warm way in.
  • “Kite insurance renewal” gets 2,400 searches a month and they rank fourth. “Car insurance quote” gets 33,100 and they are not in the top 20.
  • Three ads live, all the same offer since March.

The other four are on their cards, every fact with its source beside it. I've handed all five to the Growth Agent.

▤kite-insurance-card.pdf · 2 pages

We build a team of research agents that find that information for you on their own, every day. They work as a team, and you are not the bottleneck in it. They can research competitors. They can research prices. They can watch the things that change on a regular basis and tell you when they do.

Whatever they find goes somewhere you already work. Into a Google Sheet, into a file you read in the morning, or straight to another agent who takes it on from there. You don't go looking for the research. It is there when you sit down.

We can set up a research agent to research a prospect in advance, so that a salesperson walks into the meeting with good knowledge of the company and the person, and a few insights they would not have had time to find themselves.

How we structure your research agents depends on your goals. Deep research on a few questions, a standing watch on many, research that fires when something happens, or all three.

A big part of this is that it is an opportunity to get work done that you probably weren't able to get done before.

The library

This is where we keep the things we use, and we add to it every week.

The tools. Clay, which finds data on companies and people. Apify, which collects data from websites. Playwright, which lets Claude open a web page and click through it for you. 28 of them, with a line on what each one is for.

The people we read, like Ethan Mollick and Dan Shipper. The companies we watch for how they build, like Ramp, Every and Anthropic. The articles and videos we send each other, like the game Claude built in one go.

And every prompt from the course, written out in full.

Sign up and it is all yours. When we add something we think you would like, we will let you know.

  • 8 prompts
  • 10 links from the lessons
  • 26 documents
  • 15 People I follow
  • 6 Companies I watch
  • 18 Articles, videos and Gits
Sign up once for the library, the report PDFs and the course. Already signed up? Sign in.
the library

Free course: AI Fluency for Ambitious Marketers

A free, practical, non‑hype AI fluency course for ambitious marketers. Six modules, one a fortnight, from Monday 21 September 2026.

Each module is a lesson you read once. What it hands you, the prompts, the links and the files, goes into the library, so you never have to go back through a lesson to find the thing you half remember.

  1. 1The 80/20 of AIOpen
  2. 2Slow, then fastMon 5 Oct
  3. 3Create adjacent valueMon 19 Oct
  4. 4System thinkingMon 2 Nov
  5. 5Building marketing agentsMon 16 Nov
  6. 6Tackle difficult problemsMon 30 Nov
Module 2 opens Mon 5 Oct. The same sign-up as everything else here. About the course.
module 2 · Slow, then fast · opens Mon 5 Oct