I’m a Marketer, Not a Coder.
Here’s What Changed.

Everything on this site — the diagnostic, the six tools, the game, the agent behind all of it — came from teaching myself to build with AI over the past year. Not prompting it once and hoping. Actually building, breaking things, and fixing them properly.

I’d never built anything myself before this, even with ten years in marketing already spent working around platforms, integrations, and plenty of other people’s APIs from the outside. Knowing what a tool should do and actually building one turned out to be two very different things.

Here’s the honest version of how that went.

Why I started

Kept seeing the same marketing everywhere. Same tone, same structure, same phrases: trusted partner, passionate about delivering results, helping your business grow. You could swap the logos and nothing would change.

Most of it was AI generated, and it showed. Not because the tool was bad. Nobody had given it anything specific to work from, so it defaulted to something safe and forgettable.

Figured there had to be a better way to actually use this stuff. So I started learning how to build it properly instead of just prompting it and hoping.

Starting from zero

I started with Claude, Vercel, GitHub and Supabase, and no real idea how any of them fit together, or why I’d need all four at once.

The first few weeks were mostly getting things to load at all. Not glamorous. A lot of staring at error messages I didn’t understand yet, copying them into Claude, and slowly learning what they actually meant instead of just fixing the immediate problem and moving on.

Ten years in marketing meant I already knew what good demand generation looked like. That part mattered more than any of the technical bits, and still does. Actually loved building this, unexpectedly.

Roughly what each part does: Claude writes and reasons — the actual thinking layer, drafting everything from ad copy to the diagnostic’s own scoring logic. Vercel hosts the site and every tool, live within a minute of a push, no separate deploy step. GitHub stores every version, so a bad change is one command away from undone. Supabase is the database: every diagnostic result and tool usage stored and queryable.

The argument I had with my own AI

Told it to stop guessing when it didn’t have the data. It guessed anyway, confidently. Told it again, more specifically. Guessed again, differently confident.

This went on for days, not minutes.

The bit that finally worked wasn’t asking more politely. It was writing out, field by field, exactly what it was allowed to state versus what it had to flag as unknown. Being exact beat being nice.

That’s when review counts and rankings had started appearing that simply didn’t exist. Confidently, too. Every diagnostic report gets checked by me before it goes out now, still does. Slower than letting the AI ship straight through, and the reason I trust what actually goes out under my name.

What you actually get from the diagnostic

A real score across the areas that actually predict whether pipeline converts, not a generic checklist. Priority fixes ranked by commercial impact, not just a list of everything that’s wrong. Specific to whether you’re a solicitor, a B2B business, or life sciences, because the diagnostic itself is different for each, not the same form with a different logo.

The same one I run on real client work. Five minutes, free, no pitch until you’ve seen your own number.

Building the diagnostic properly

The diagnostic platform came out of that lesson directly. Not one AI call producing a report. A structured framework, different scoring categories depending on whether you’re a solicitor, a B2B business, or a life sciences company — because those are different buying journeys, not the same funnel with different labels.

Getting the AI to say “we don’t have enough information to answer this” instead of guessing took more work than getting it to sound confident. Confident is the default. Honest took actual effort to build in.

Why the tools exist at all

The six free tools came later, once the diagnostic was solid enough that I had spare capacity to build things purely because they were useful. Content repurposing, a revenue gap calculator, subject line testing, email analysis, a buying committee mapper, a pipeline review. Different problems, same underlying approach: give the AI something specific to work from, and check what it produces before trusting it.

What each agent actually produces

The free tools are the small, visible layer. Underneath them are three separate agents, built for how each sector actually works, not one generic tool with a different label per page. Real outputs, not a task list — the kind of work that used to take a strategist days, now a checked starting point in minutes.

The Local agent produces a full scored marketing audit, a 90-day strategy playbook, and Google review responses written in your voice. The Demand agent produces a full SOSTAC situation-to-strategy analysis, competitor battlecards, and deep-dive research on named target accounts. The Life Sciences agent produces MLR-ready congress abstracts, KOL engagement letters, payer value briefs, and regulatory copy checks.

Still checked by me before anything reaches a client. Speed on the drafting, judgement on what ships.

It’s a connected plan, not a pile of tools

From positioning to pipeline, mapped stage by stage, built to work together rather than used one at a time.

Positioning comes first: ICP, value prop and messaging defined before anything else gets built. Awareness follows: SEO keyword clusters and content matched to what buyers actually search. Consideration is nurture sequences and email built stage by stage, not one generic drip. Decision is sales email sequences, case studies, and sales deck copy for the real conversation. Pipeline is lead scoring and pipeline review, forecasting revenue, not just counting activity. Retention turns win stories into the next campaign’s proof.

One buyer journey, positioning to revenue. Nothing built in isolation.

The Voice Brief, and the actual lesson underneath all of it

Somewhere in building all of this I kept running into the same problem from a different angle. AI output left to its own devices sounds like everyone else’s — not because the tool is bad, because nobody told it what makes this business different before asking it to write anything.

That’s what a Voice Brief actually fixes. Not really about AI at all, if I’m honest. The same thing a decent marketer has always done — understand the business properly before writing about it — just applied to a tool that can now execute the output in seconds instead of days.

Go deeper on how a Voice Brief actually works →

What I’d tell someone starting where I did

The AI does the heavy lifting on execution now. It does not do the thinking for you, and it will sound completely confident while being completely wrong if nobody’s checking. Good reason to build carefully, not a reason to avoid building with it at all. Check everything. Treat speed as the benefit it is, not a replacement for judgement.

Still figuring most of this out as I go.

Common questions

Did you have any development experience before starting this?

None, in terms of actually building software. Ten years in marketing meant I understood platforms, integrations, and what good output should look like from the outside, but building something myself was a genuine first. That gap between knowing what’s good and knowing how to build it is part of why the manual verification habit matters so much.

Is the AI doing the work, or are you?

Both, in different roles. The AI handles execution, drafting, and speed. Every judgement call — what to prioritise, what’s worth saying, what to verify before it ships — still comes from a person. Neither part works well without the other.

Why three separate agents instead of one?

Because a solicitor, a B2B software company, and a life sciences business genuinely buy differently. One generic tool trying to cover all three ends up shallow on each. Three agents, each built around how that specific sector actually decides, produces something closer to what a senior marketer in that sector would actually do.