They've stopped Googling. 
They talk to ChatGPT.

They help a business decide where AI adds value and put it to work with sound judgement. My AI consultancy has two strands: getting ChatGPT, Google and the other answer engines to find and cite your business, and designing AI agents that work with your data and your rules.

You don't need an AI expert on the payroll: you need someone who tells you, with data, what's worth doing. Good AI consulting also means telling you when your case doesn't need AI at all.

It's the practice of getting a business found, mentioned and cited both on Google and in answer engines: ChatGPT Search, Perplexity, Google AI Mode, Gemini and Claude. It combines SEO (search engines), AEO (Answer Engine Optimization: direct answers) and GEO (Generative Engine Optimization: becoming a source for generated answers).

Unlike classic SEO, AI visibility is binary: either you appear in the answer or, for that user, you don't exist. That's why I work with dated, reproducible evidence, not ranking promises.

SEO optimises for appearing in search results. AEO structures content so that an engine selects it as the direct answer (answer first, question-style headings, tables, structured data). GEO works to get generative models to recognise you as an entity and cite you as a source: consistent, dated facts, stable pages, third-party mentions and open access for AI crawlers.

All three layers are measured against the same baseline: questions × engines × runs.

More detail in the guide what is GEO and how it differs from SEO and AEO.

With a reproducible baseline: a set of commercial questions agreed with the client, run several times in each engine under documented conditions (account, language, country, date). I record whether the brand is mentioned, whether its website is cited, which URL, which external sources and which competitors appear.

A single run is a snapshot, not a baseline. I only compare repeated observations before talking about a trend.

Because AI knows your website through two separate routes: trained memory (crawls such as Common Crawl, months earlier) and live retrieval (the engine reads your page as it answers). A CDN block, a restrictive robots.txt, content that only exists once JavaScript has run, or simply never having been crawled can leave you out of the first route even if Google indexes you perfectly well.

My access and inclusion audit checks robots.txt, responses to each AI user agent, presence in the Common Crawl index, rendering and structured data.

A review based on public evidence, done before our first meeting: website and profiles, basic access and mobile experience, brand searches, a sample of AI visibility using buyer questions, and a comparison with the sources the engines do cite. The outcome is three well-supported observations, explicit hypotheses and the questions we need to agree on the next step.

It's a preliminary sample, not a full audit or a promise of inclusion in search.

Three phases: a reliable snapshot (baseline, Search Console, GA4, technical crawl and evidence inventory), strategy (a canonical layer of facts and entity, citable case studies, guides that answer the buyer's questions, improvements to existing pages and structured data) and oversight of the implementation, with QA and a repeat of the baseline.

Every action has an owner, a date and a reproducible way to verify it.

To handle one specific task in your business: it's an internal AI system combining models that analyse your data, classify information, set priorities and flag risks with a conversational assistant that explains the results and suggests next steps. It uses your data and your rules, and can only do what we've agreed.

At AllHub I design this kind of agent as a service (Agent-as-a-Service); RnD Ventures AI OS shows a complete conversational system in production.

Weighing up your options? Read the comparison AI agent vs chatbot vs automation.

The ones that have you checking data by hand today: spotting a stock-out before it happens, prioritising overdue payments, sorting emails and tickets by topic and urgency, or tracking prices and availability in your market. The agent analyses, tells you what matters and why, and you decide.

If your task doesn't need an agent, I'll tell you in our first conversation.

Seven phases: briefing (we get to know your business and requirements), design (how it talks and what it can do), prototype (you try it before it's built), build (we connect it to your tools with only the permissions it needs), testing (we try to break it before anyone else does), gradual launch, and ongoing support with monthly improvements.

At every phase you know what you're getting, and nothing moves forward without your sign-off.

Your data is hosted in the European Union and processed in line with GDPR. It is never used to train any AI. The agent only accesses what it needs for its task; everything else is locked down.

Every conversation and every action is logged, so you always know what it has done and why.

That's why it's designed with clear limits and a person in charge: important decisions go through someone on your team. The agent also separates what it has observed from what it interprets, and shows the data behind every conclusion.

Before launch we test it with real cases, difficult situations and attempts to trick it, and it starts by serving a small number of users under supervision.

It depends on the task, the tools that need connecting and the volume. After the briefing you receive a fixed proposal, with phases, deliverables, timescales and price.

Get in touch through the contact form: all good AI consulting starts with understanding your case. If you're after AI visibility, include your website and the three questions you'd like AI to answer by naming you: I'll prepare an initial review before we meet. If you're after an agent, tell me which task you'd like to take off your plate.

In a 30–40 minute meeting we agree on a concrete next step.