How I work

AI is my sidekick.
The judgement call is mine.

The tools bring the speed. I bring experience and thinking, staying with it from start to finish.

Most of strategy used to be labour intensive. Transcribing interviews. Pulling research together. Checking a brief line by line. AI now does a lot of the repetitive tasks, and does it fast. What it can't do is sit in the room, read the body language, or know when an answer is technically right and strategically wrong. That part is still owned by me.

The loop behind everything

Two steps for AI. Three for me.

Every process below runs on the same loop. Anything wrong, thin or generic goes back round.

Set the question

ME

I frame the problem and choose the framework.

Gather and draft

AI

Research, transcripts, first drafts and pattern-finding.

Sense-check

ME

I check what comes back against what I know. Wrong, thin or generic goes back.

Refine

AI

AI reworks it with my steer.

Decide and sign off

ME

Final judgement. Nothing goes out without it.

In practice

Seven ways the loop shows up in my week.

Pick one to see what changed, what AI does, and what I still do.

Rapid research

Get up to speed on a brand fast: company, category, consumer and culture.

Before

Days of desk research

Now

A sourced first view in a few hours

What AI does

Runs structured prompts through Perplexity and ChatGPT, one set per C, each asking for answers with sources.

What I still do

Checks the sources, throws out the generic, and spots the tension worth chasing.

The 4Cs give me the facts. The insight still comes from me.

Framework adapted from Julian Cole, Planning Dirty.

Stress-testing the approach

Test the project approach against marketing theory before anyone sees it, and make sure it holds together.

Before

Gaps found when the work came back

Now

Fixed before any work starts

What AI does

Stress-tests my long-form brief against foundational marketing theory. Names the theory the brief is really relying on, flags contradictions between objectives, KPIs and creative, and checks for a real insight.

What I still do

Writes the brief, judges whether the diagnosis fits this client, budget and moment, and owns the final version.

A sidekick I think out loud with. Problems solved before they become problems.

Built on foundational marketing theory, from the classic funnel to Ehrenberg and mental availability.

Qualitative research

Run interviews and find out what people really think.

Before

Hours of typing per interview

Now

Key themes turned around rapidly

What AI does

Records and transcribes each interview, writes a synopsis, then pulls out the recurring themes across all of them.

What I still do

Is in the room. Body language, tone, the pause before an answer. Adds it to the themes, challenges what doesn't ring true, and decides the insight.

The same depth of research, without the days of transcription.

Customer journey mapping

Map how people discover, choose and experience something, and where it breaks.

Before

Present it, then wait for the holes

Now

Holes fixed before the client sees it

What AI does

Builds the map stage by stage with me, from the client's documents and my research.

What I still do

Carries out the research, runs the workshop, shapes the guiding principles, and decides where the real opportunities are.

A journey map that holds up first time, and becomes the backbone of the comms plan.

Example: a national outdoor tourism body.

Comms strategy and channels

Decide what to say, to whom, where and when.

Before

One message for everyone, channels on gut

Now

One message per audience and stage, channels verified

What AI does

Builds tailored messages for each audience by stage of the purchase cycle, and tests my channel choices against validated research.

What I still do

Sets the segments, judges the messages, and forms the first view on channels from experience.

Every message does one job, for one person, at one moment, in the right place.

Examples: a national outdoor tourism body, and a B2B launch for a specialist logistics service.

Orchestrating agents

Run whole pieces of work through a team of AI agents, with me in charge.

Before

One person doing every task

Now

A small production line I manage

What AI does

Agents on Buzz can handle research, copy and design, handing work to each other like a small team.

What I still do

Sets the direction, checks the work at every stage, and keeps the brand, its metrics and its identity at the centre. No agent drifts off brand.

It's like being the conductor of an orchestra.

Thinking out loud

Get ideas, prompts and first drafts out of my head quickly.

Before

Typing at 30 words a minute

Now

Talking at 200 words a minute

What AI does

Wispr Flow turns what I say into text, straight into prompts and drafts.

What I still do

Does the thinking, out loud. Talking a problem through is how I find the angle.

I rarely type now. The ideas come out at the speed I think them.

Orchestrating agents, in motion

The conductor, not the orchestra.

My agent system, simplified. The brand sits at the centre. Agents make the work. Nothing reaches a channel without passing a red gate, where I approve it or send it back. Results flow home and the plan adjusts.

Key

  • Agent work.
    Drafts, posts, emails and pages heading out.
  • I decide.
    Every piece stops at a gate. Approved, or sent back.
  • Results.
    Performance comes home so the plan adjusts.

Live Run log

    Illustrative. Channels, agents and log entries are examples of how the system works.

    In practice. Behind this is a working team of twelve agents in Buzz: a project manager, a strategist, a research lead, writers, designers, a developer and a compliance reviewer. Each one has a role and written instructions, like a job description. I set the brief and approve everything they make.

    Go deeper

    See it on a real project.

    More time thinking. Less time typing.

    That's the whole point.