Korolov Book a call
Oleg Korolov

Oleg Korolov · Engineer-marketer · London

The growth hire you can’t afford yet, delivered as a system you keep.

I diagnose what’s actually broken in a seed-stage company’s growth, build the system that fixes it, and hand it over. Fixed price. No retainer.

01 · How I got here

I trained as an architect. Then a solar car got in the way.

Architecture school teaches one thing well: everything is a system. You learn it in projects, and the brief moves while you are still drawing. Architecture as a career moves a good deal slower than that, which is why I am not an architect.

Marketing found me through Cambridge University Eco Racing, where I was the designer and marketing manager. That is how I ended up drafting car livery in the Australian outback. Then came New AutoMotive in 2020, working on the policy and data behind EV adoption, back when 60 miles counted as good range.

I wanted something more commercial and faster. So now I run organic and AI search at The Electric Car Scheme, and I build the same kind of systems for founders who do not have a marketing team yet.

Oleg sitting cross-legged on grass between swag tents, in an Eco Racing cap, working at a laptop with rolls of livery vinyl beside him.
Drafting livery on the road with Cambridge University Eco Racing

02 · Who this is for

Seed-stage. No marketing hire yet. Founder still doing it.

A marketing lead is £60-80k and six months away. An agency is a retainer and a deck. This is the third option, and it usually starts with one of these.

Nobody can find you

Not on Google, not on ChatGPT, and nobody knows which pages would fix it.

Your pipeline is a spreadsheet

Nobody knows who to call next, nothing forecasts, and you can’t hand it to a hire.

Someone said “do AI search”

And nobody in the building knows what that means.

03 · Work

Four kinds of work, from AI search to custom systems

AI search · The Electric Car Scheme, in-house

13 sales meetings in a quarter from people asking ChatGPT what to buy

This is my day job. I opened AI search as an acquisition channel before the business asked for it. In Q2 2026 it produced 48 leads, 13 sales meetings and a closed B2B deal, each traced back to the CRM. Over the same year the brand went from one of several to the name the AI engines give first.

16% → 33%AI visibility in 12 months
The year · weekly
0%10%20%30%40%Nov 2025FebMayAugNov 202616%33%

Share of AI answers naming each brand, week by week. The Electric Car Scheme in red; six competitors in grey.

Now · last 30 days

20 Oct - 18 Nov 2026, 8,738 chats across ChatGPT, Gemini, Perplexity and Google AI. #1 of 11 tracked brands.

Built
A tracked set of the questions buyers put to AI engines, run daily. Pages rebuilt so the engines retrieve and cite them. Attribution from an AI referral through to the CRM, so a meeting can be traced to the answer that started it.
So what
AI search stopped being a hunch. It has a target, a weekly number and a line in the monthly report next to paid and organic, and the team measures the top of the funnel by citation share, not clicks.
Next
Three competitors have climbed to 21% in the last month. The work now is holding the lead on the questions that turn into meetings, not the ones that flatter the chart.

For you: the same setup at seed size. The 30 to 50 questions your buyers ask, the pages that answer them, and a weekly number you can put in front of investors.

Book a call

Agentic marketing engineering · The Electric Car Scheme, in-house

A monthly board report that assembles itself

Every marketing team loses days each month to pulling numbers into slides. The monthly organic review used to take four working days: two of a consultant’s and two of mine. An agent now pulls the data, computes every figure and its change, and lays out the review. I add the analysis. It is the format the CEO reads.

The agent computes and never narrates. Commentary stays a human job, and every figure on the page can be edited by hand.

4 days → 1per monthly review
SourcesFive data sourcesCRM warehouse, Search Console, analytics, AI visibility, search results
ComputeFigures and changesOnly this month and last are re-pulled
AssembleScoreboard, then detailSame layout every month
HumanAnalysis and asksWritten by me, not the agent
OutBoardroom reviewOn a schedule, on the 2nd
Also runs
A daily competitive brief. A report on how social posts show up in AI answers, which the social lead runs herself. A Monday digest of the team’s goals in Slack.

For you: the investor update, board pack or weekly numbers you rebuild by hand. If it is the same sources every time, it should build itself.

Book a call

SEO and growth diagnostics · Stasher

Three theories. One was true. One memo.

A luggage-storage marketplace whose top 100 commercial pages ranked poorly. The team had framed three possible causes and asked which to investigate first. They paid for about ten hours.

The question

The commercial pages rank poorly and the problem looks systemic. Which of these three would you investigate first, and what would you expect to find?

  1. Content. The pages are too templated or too thin to compete.
  2. Authority. There aren’t enough links to compete, whatever is on the page.
  3. Technical. Crawling, indexing or internal linking is holding the pages back.
The answer · memo, p.12-hour triage

Run content first, but not for the reason given


Authority. Not the constraint. A competitor with fewer referring domains and near-identical authority ranks on page one for the head term.

Technical. The money pages are indexable, server-rendered and clean. One day’s check inside Search Console to confirm, no more.

Content, reframed. The pages aren’t thin. They are longer than the pages beating them. They rank for a lot of keywords and rank low on all of them: a templated section the search engine indexes and then discounts as a block.


Fix at the template, not page by page. Stated up front: the one finding that would overturn this, and what would be done instead.

For you: this is what a Scope looks like. Two weeks, £1,000 fixed, and you know what is broken and what it costs to fix.

Book a call

Custom systems · Zen Educate

Every reply finds its school and its trust, with no CRM

When a trust exec forwards a sales email and a headteacher replies from a different address, the thread is lost and someone opens five tabs to work out who it was. Working from the sales spreadsheet, I designed an account-based engine on Sheets, Zapier and the DfE schools register, and built the console below.

It resolves the sender’s domain to a school and its parent trust, checks the record against the register, scores the trust, writes the lead to the tracker and routes it to the right account exec.

1 weekspreadsheet to working MVP
The console · try it Every trust, school, person, domain and ID here is a placeholder

Who sent me this email?

A reply has come in from an address nobody recognises. Press Resolve to see what the engine does with it.

The forwarded-email problem

We email the trust. An exec forwards it, and a headteacher replies from a different address. Both land on the same account.

Self-healing data

The source spreadsheet had the wrong school against some domains. The engine corrects them from the register at ingestion and flags each one.

For you: the problem with no off-the-shelf tool. If your team does the same lookup by hand every day, it can be a system.

Book a call

04 · The offer

Two ways to work with me. Three slots.

Scope2 weeks

£1,000 · fixed

In two weeks you know what’s broken, what it costs to fix, and the fixed price of the Build.

Ends in a plan you can run yourself. Most people have me build it.

50% deposit holds the slot. The fee comes off the Build. Tooling’s on me.

Build2-6 weeks

£2,500+ · fixed at the Scope

One shipped system you keep: the outbound stack, the AI search setup, or the attribution and reporting. Handover and a runbook.

You know the number before you say yes. One at a time, and it takes a slot.

I do the whole thing myself, then get out of your way.

05 · Where the work happened

Ten years of small teams and nobody to hand the work to

The Electric Car Scheme2025 to now
I own organic and AI search end to end and report to the CEO. Made AI search a channel with sales meetings attached, and built the reporting the team runs on.
Stasher2026
Worked out why the commercial pages were not ranking and wrote the plan to fix it.
Zen Educate2026
Designed and built an account engine with no CRM, starting from a sales spreadsheet.
New AutoMotive2020 to 2024
EV data and policy in a team of six. Ran a monthly bulletin reaching 85% of the UK automotive market, and worked on market trackers that were still in use nearly two years after I left.
Cambridge University Eco Racing2017 to 2019
Designer and marketing manager. Raised more than £500k through PR and launched the car at the Science Museum.

06 · Notes

What I am seeing, written down

Nov 2026AI visibility is SEO. It is just omnichannel now.

Your site still has to be machine readable. Sitemap, schema, clean on-page HTML, decent Lighthouse scores. None of that has gone away.

What has changed is that the engines check whether anyone else backs you up. A site can say anything. Do sites with real authority say it too? Is there evidence off the page, on YouTube, on social, in reviews? The models go and look, so bluffing does not work.

Prompts are also far more specific than keywords ever were. The game is no longer a keyword list. It is your ideal customer’s objections, desires and goals. If you do not understand the audience, you cannot know what to be visible for.

Some numbers from Peec AI’s Q3 2026 survey of 700 marketing decision-makers:

  • 89% of companies raised AI search spend last year. Only 24% can prove ROI, and 19% in the UK.
  • Everyone runs the same seven tactics, so none of them is an edge. What separates teams is measurement, and how fast they see what works.
  • 83% treat AI search as part of SEO, and those teams prove ROI more often: 31% against 17%.
  • ChatGPT changed in August. Sessions with several fan-out searches went from 2% to 61%, product page citations rose to 2.5 times their earlier level, and Reddit fell to about a tenth.
  • The engines differ a lot. YouTube is 7.7% of the sources in Google’s AI Overviews and 0.2% of ChatGPT’s.

So you need good tech, good prompts, good measurement and, as with all marketing now, a return you can show. I can point to sales that started with an AI answer. Can your team? That is what takes this from vibes to a commercial growth case.

It is a vendor survey with self-reported answers, so read the percentages as direction, not proof.

Sep 2026Your monthly report is an ROI problem

A manager has about twenty working days in a month. If four of them go on assembling a report, a fifth of the month is not spent on the job. At a seed-stage company that is a lot. Add a consultant billing £100 an hour for two of those days and do the maths.

Before you automate any of it, answer four questions. Where does the data come from? Are you actually measuring it? Is it reliable? And what are you reporting on, and who is it for?

Get those right and the report does more than save time. It proves which channel is paying for itself, which matters most when budgets are small and minor things move the needle.

The agentic layer is the easy part after that. It is a set of good internal automations. Start with one small piece that can be fixed, such as site analytics into a dashboard, and build out from there.

Aug 2026Ranks for a lot, ranks low

A marketplace asked me why its commercial pages were not ranking. There were three theories on the table: weak links, thin content, or a technical fault.

Links were not it. A rival with fewer referring domains and almost the same authority sat on page one for the head term. Thin content was not it either. The pages were longer than the ones beating them, and technically clean.

The tell was traffic per keyword. The site ranked for thousands of keywords and earned about 8 visits from each. A competitor earned about 26 from each, on a third as many keywords. Ranks for a lot, ranks low. That is the fingerprint of a templated section the search engine indexes and then discounts as a block.

The fix is at the template, not page by page: more that is unique to each location, less shared scaffold. And before starting, write down the one finding that would prove you wrong.

07 · Next

Book the call and tell me which one it is.

Thirty minutes, free, and you will be talking to me. I only take work where I can see the system that would fix it. If I can’t, or if you’re too early for this, I’ll say so on the call and point you somewhere better.