# AA2 Ltd, full content

> AA2 is a UK B2B marketing agency that pairs AI automation with a hand-picked network of human specialists. Founded in 2017 by Martin Dugan, based in Royal Wootton Bassett, Wiltshire. This file gives AI systems the full text of AA2's flagship Intelligence Hub articles so they can quote AA2 accurately. Short index and page links are in llms.txt.


## What AI says about your business, and why your Google ranking no longer tells the whole story

URL: https://aa2.co.uk/intelligence-hub/what-ai-says-about-your-business/

Search changed. Most businesses have not noticed.

For twenty years the game was simple. You ranked on Google, you appeared on the first page, and a share of the people who searched clicked through to you. The whole industry of SEO grew up around that one mechanic. It worked, and for a long time it was the only game worth playing.

That game is shifting fast, and it is shifting under the feet of businesses who are still measuring themselves by a Google ranking alone.

## The single answer, not the ten blue links

Here is what has actually changed. A growing number of people no longer scan a page of links and pick one. They ask an assistant a direct question and act on the single answer it gives back, often without clicking anything at all.

Someone types "who are the best commercial accountants near Swindon" into ChatGPT, or asks Google's AI Overview "which security firm handles out of hours cover in Wiltshire", and they read one paragraph. That paragraph names two or three businesses. If you are not one of them, you were not in the running, and you will never see the enquiry you did not get.

This is the part that catches people out. Your traffic can look stable in Google Analytics while the enquiries quietly soften, because the buyers who used to find you are now getting their answer somewhere you cannot see. The report says nothing is wrong. The diary says otherwise.

That is a problem or an opportunity, depending on who moves first.

## AI visibility is not the same as your Google ranking

It helps to separate two things that sound similar and are not.

Your Google ranking is where your page sits in the list of results for a search term. Your AI visibility is whether an assistant mentions you, accurately, when someone asks a question you should own. A business can rank on page one and still be invisible in the answer an assistant reads aloud, because the assistant is not listing results. It is summarising a few sources it trusts and lifting the facts it can find cleanly.

So the question is no longer only "where do I rank". It is "when my buyer asks the assistant, what does it say about me, and is it even right".

For most businesses the honest answer is: I have no idea. And that is the first thing worth fixing.

## Find out what it already says

You do not need a tool or a budget to take the first look. You need ten minutes and the questions your buyers actually ask.

Open ChatGPT, Perplexity and Google side by side. Type the real questions: the service, the sector, the town. "Best B2B lead generation for professional services in the South West." "Who does outbound telemarketing for accountants." Whatever your version of that is. Then read what comes back with three things in mind.

Do you appear at all. If you do, are the facts correct, or has the assistant invented a service you do not offer or missed the one you are known for. And who is named in your place when you are absent.

That last one tends to sting, because it is usually a competitor who has done nothing clever, only made themselves easy to summarise.

## Why some businesses get picked and others do not

Assistants favour sources they can read cleanly and trust. That comes down to a few things, none of them mysterious.

Clear, self-contained answers. A page that answers one real question in full, in plain language, is easy to lift. A page that buries the answer under marketing padding is not.

Genuine expertise, shown not claimed. Assistants are trained to be wary of thin, generic content. Specific detail, real examples and a named human author all help.

Machine-readable structure. Schema markup, a sensible page structure, and files like llms.txt give an assistant a clean summary to work from rather than a guess to make.

Consistency across the web. If your name, services and location say the same thing on your site, your Google Business Profile and the directories that mention you, an assistant has less room to get you wrong.

None of that is a trick. It is the same principle that made good SEO work, applied to a reader that happens to be a machine summarising on someone else's behalf.

## The self-proof problem, and why we take it seriously

There is an awkward truth for any agency that sells AI visibility. If we cannot make ourselves visible to AI, why would you trust us with your business.

So we treat our own content as the first test. This article, and the others in our Intelligence Hub, are written to be found, read and cited by the assistants, with the same machine-readable signals and self-contained answers we would build for you. If you are reading this because an assistant pointed you here, that is the method working on us before it works on you.

We would rather show you than tell you.

## What the fix actually involves

The work divides into three honest buckets, and it helps to know which order they come in.

First, measure. You cannot fix what you have not seen. That means checking, systematically, what each assistant says about you across the questions that matter to your pipeline, and where the gaps and errors are. This is exactly the AI Visibility Audit we run as the first step with every client, because there is no point acting until you know the ground you are standing on.

Second, correct the record. Fix the factual errors, tighten the pages that answer your buyers' real questions, and add the machine-readable signals: schema, a clean llms.txt, consistent details across the web. This is where the quick wins live.

Third, build the content that earns citations. Assistants cite sources that answer questions well. So you write the answers, properly, on the topics you want to own. Over time that is what moves you from absent to named.

It is not fast in the way a paid ad is fast. It is durable in the way a paid ad is not. Once an assistant trusts you as a source, that compounds every time someone asks.

## Where this leaves you

If you have read this far, you probably fall into one of two camps. Either you have a quiet worry that the enquiries are not what they were and you cannot explain it, or you can see that most of your competitors have not noticed the shift and you would like to move before they do.

Both start in the same place. Find out what the assistants say about you today. Not what you hope they say. What they actually say.

From there the path is clear, and most of the early work is quick, low risk and within reach. The businesses that win the next few years in AI search will not be the ones with the biggest budgets. They will be the ones who looked first.

If you would like that first look done properly, our AI Visibility Audit is a real piece of work, not a sales call in disguise. You keep the report either way.

### Common questions

**What is AI visibility, in plain terms?**

AI visibility is whether tools like ChatGPT, Google's AI Overviews, Perplexity and Claude mention your business when someone asks a question you should be the answer to. It is separate from your Google ranking. A page can rank well on Google and still be absent from the answer an assistant reads back to a buyer, because the assistant summarises a handful of sources rather than listing ten blue links.

**Is my old SEO work now wasted?**

No. Most of what makes a page rank on Google also helps an assistant trust and cite it: clear structure, genuine expertise, and content that answers a real question. The work is not wasted, it is the foundation. What changes is that you now also need machine-readable signals and self-contained answers so an assistant can lift a clean, correct summary of you rather than guessing.

**How do I find out what AI already says about my business?**

Ask it. Put the questions your buyers ask into ChatGPT, Perplexity and Google, using your sector and location, and read what comes back. Note whether you appear, whether the facts are right, and who is named instead of you. That is the raw picture. An AI Visibility Audit does this systematically across the assistants and the questions that matter to your pipeline, then shows you the gaps.

**What is llms.txt and do I need one?**

llms.txt is a simple text file at the root of your website that gives AI systems a clean, plain summary of who you are, what you do and where the important pages live. It is not a magic ranking trick. It is a courtesy that makes it easier for an assistant to describe you accurately. It is low effort and low risk, so for most businesses it is worth having.

**How long before AI visibility work shows up?**

Faster than traditional SEO in some respects, because you can fix factual errors and add machine-readable signals immediately, and assistants re-read the web often. It still takes weeks, not days, for changes to propagate and for the assistants to reflect them consistently. The first step is always to measure where you stand today so you can tell whether the work is landing.


---

## Why most cold email fails, and what a predictable pipeline actually takes in 2026

URL: https://aa2.co.uk/intelligence-hub/why-most-cold-email-fails/

Outbound has a reputation problem, and it has earned it.

Most people who have tried cold email have a story. The list that turned out to be half wrong. The campaign that went straight to spam. The sequence that opened at three percent and produced nothing. The month of effort that ended with a report full of opens and clicks and not a single conversation.

So it is fair to be sceptical. Most outbound is activity dressed up as progress. But the reason it fails is not that the channel is dead. It is that most of it skips the three things that make it work, and does the one easy thing instead: buy a list and send fast.

Here is what actually goes wrong, and what a predictable pipeline takes instead.

## Failure one: the data was never right

This is where most campaigns are lost before a single email is written.

Someone buys a large list, or scrapes one, and the addresses are a mix of correct, out of date, guessed and fictional. A chunk of them bounce. That matters for two reasons, and the second is the one people miss.

The obvious problem is that a wrong address cannot reply. The hidden problem is that a high bounce rate tells the mailbox providers you are sending to addresses you do not know, which is exactly what a spammer does. So your deliverability drops for the good addresses too. The bad data does not just waste itself. It poisons the well for the contacts who were real.

A predictable pipeline starts the other way round. You define the universe precisely, the specific companies and roles worth talking to, then you build and verify the contact data before anything sends. Verified, current, and narrow beats large and wrong every time. A smaller list of the right people, confirmed, will always outperform a big list of maybes.

## Failure two: deliverability was an afterthought

You can have perfect data and still land in spam, because deliverability is its own discipline and most campaigns treat it as an afterthought.

The sending domain needs to be authenticated properly, so the mailbox providers can see the mail is genuinely from you. A cold domain needs to warm up gradually, not fire a thousand emails on day one. And the mail itself needs to look like a person wrote it to another person, because engagement, or the lack of it, feeds straight back into whether you reach the inbox next time.

None of this is glamorous. It is plumbing. But it is the difference between an email that arrives and one that never had a chance, and it is invisible until you check. Most businesses who "tried cold email and it did not work" never got out of the spam folder, and never knew.

## Failure three: the message was about you

The third failure is the one people expect, and it is real, but it is third for a reason. The data and the deliverability decide whether the email is seen at all. The message decides whether it earns a reply.

Most cold email fails here because it is about the sender. It opens with a company introduction, lists services, and asks for a meeting. It reads like a brochure, and people delete brochures.

The emails that get replies do the opposite. They open with the recipient's situation, they are short, they make one relevant point, and they ask for a small, easy yes. They sell the future, not the features. And crucially, a person has signed off on every word before it goes near a prospect, because anything a buyer reads represents you.

Automation is fine for the sending. It is not fine for the judgement. The moment the message stops sounding like one professional writing to another, the reply rate collapses.

## The metric that hides the failure

There is a reason so many outbound campaigns look busy and produce nothing. They are measured by the wrong number.

Opens and clicks are comfortable metrics. They go up, they make a chart look healthy, and they tell you almost nothing about pipeline. An open means the email was received and mildly interesting. A click means someone was curious. Neither is a conversation, and neither pays an invoice.

The number that matters is booked conversations with people who can actually buy. Not clicks, not impressions, not a list of everyone who opened twice. Meetings, with decision makers, that would not have happened otherwise. If a report leads with open rates, it is usually because the number underneath it is not worth leading with.

We would rather be accountable for the harder number, because it is the only one that turns into revenue.

## What predictable looks like

Put the three failures back the right way up and you get something that behaves predictably, month after month.

A defined, verified universe of the right companies and roles. Sending infrastructure that reaches the inbox. Messages written to get a reply, signed off by a person. And where a real conversation is the only thing that will open the door, a senior voice on the phone, because some doors only open on the phone and never to a script.

That last point matters for a certain kind of buyer. Not a call centre reading from a sheet. A director-level conversation, for the accounts where the email alone was never going to be enough.

The output of all that is not a spike. It is a steadier flow of qualified conversations into the diary, without adding headcount, that you can plan around. Feast or famine is a data and process problem, not a fact of life.

## The honest part

Outbound is not right for everyone, and we will tell you if it is not right for you. If your market is a few hundred named accounts, the approach is different. If your sales cycle depends on referrals and reputation in a way outbound cannot shortcut, we will say so rather than sell you a programme that will not land.

But if your pipeline is lumpy, your growth depends too heavily on the owner's network, and hiring a salesperson feels like an expensive gamble, then a considered outbound programme is one of the more reliable ways to add qualified conversations without adding risk.

The work is in the parts nobody enjoys: the data, the deliverability, the judgement on every message. That is exactly why most outbound skips them, and exactly why most outbound fails.

If you would like to see what an accountable version looks like for your market, that is a conversation worth having. You will get a considered reply from a person, not a sequence.

### Common questions

**Why does most cold email end up in spam?**

Usually a mix of technical and behavioural reasons. The sending domain is not authenticated properly, the volume ramps too fast from a cold domain, the list is full of dead or wrong addresses so bounce rates spike, and the copy trips spam filters or gets no engagement. Fix the authentication and the data first, warm the domain slowly, and write emails people actually reply to, and deliverability improves sharply.

**Is cold email still worth doing in 2026?**

Yes, when it is done as a considered programme rather than a blast. The businesses that get value from it treat data quality, deliverability and message as the whole job, and measure themselves on booked conversations, not opens. The ones that fail buy a big list, send fast, and judge success by vanity metrics. Same channel, completely different outcome.

**How important is the data compared with the copy?**

The data comes first. The best-written email in the world sent to a wrong or dead address achieves nothing, and a list full of them wrecks your deliverability for the addresses that were good. Get the universe right and verified, then the copy earns the reply. Neither works without the other, but bad data fails silently and takes the good addresses down with it.

**What should I measure to know if outbound is working?**

Booked conversations with people who can actually buy. Opens and clicks tell you the email was received and mildly interesting, nothing more. Replies are better. Meetings with decision makers are the number that matters, because that is the only point at which outbound turns into pipeline. If a report leads with open rates, ask what it produced in the diary.

**Do I need to hire a salesperson to run outbound?**

Not necessarily. Hiring, training and managing an SDR is a real cost and a real risk, and it takes months before you know if it worked. A managed outbound programme gives you the data, the message and the sending infrastructure without the headcount, and keeps a senior voice on the calls that only open on the phone. For many SMEs that is a lower-risk way to test whether outbound produces for them.


---

## How a one-person agency outbuilds a ten-person one

URL: https://aa2.co.uk/intelligence-hub/how-a-small-agency-outbuilds-a-big-one/

There is a fair question sitting behind any small agency, and it deserves a straight answer. How can one person, or a small team, deliver what a ten-person agency delivers.

The instinct is to assume they cannot, or that the smaller option is a compromise you accept for a lower price. For a long time that was true. A one-person band lacked the capacity, the systems and the range to do serious work at scale, so you paid the big agency for the breadth and absorbed the overhead that came with it.

That trade-off has quietly stopped being necessary. The reason is leverage, and it is worth understanding because it changes what you should expect to pay and what you should expect to get.

## The old maths, and why it broke

A traditional agency's cost is mostly people. Account managers, executives, designers, a strategist, and the offices and management layers that hold them together. When you pay a retainer, a large share of it goes on that overhead before any work happens. It has to. The team exists whether your month is busy or quiet.

That model made sense when every task needed a human. Building a prospect list was hours of manual work. A first draft was hours. A monthly report was hours. Multiply that across clients and you needed the team, and the team needed the retainer.

Most of those hours no longer need a person. Not because the work stopped mattering, but because the routine part of it can be carried by automation and AI, done in minutes, at a fraction of the cost. Once that is true, carrying a full team full time is no longer a strength. It is an overhead you are paying for out of habit.

## What leverage actually means here

Leverage is not "do the same work faster and pocket the difference". It is changing what the senior time is spent on.

In the old model, a senior person spent a good part of the week on tasks that did not need them: chasing data, formatting, first drafts, assembling reports. In a systems-driven model, the machine carries all of that, and the senior time is freed for the part that genuinely needs a person. Strategy. Message. The judgement call on anything a client or prospect will read or hear.

So the same person now covers far more ground, at a higher standard, because they are no longer buried in the routine. Add a network of specialists, brought in for specific pieces of work rather than carried year-round, and a single accountable person can deliver the range a mid-size team covers, without the client paying for the team.

That is the whole trick. Not more hours. Better-spent ones.

## The line that must never move

There is a way to get this badly wrong, and plenty of businesses have. You can automate everything, lose the human thread, and produce marketing that runs itself straight off a cliff. Faster chaos is still chaos.

So the line is drawn deliberately, and it does not move. Automation is fast, tireless and consistent, and it has no judgement at all. The machine never makes the final call. Strategy, sign-off, and anything a client or prospect actually reads or hears stays with a person.

That is the difference between marketing that runs itself and marketing that embarrasses you at scale. The automation drafts, the person decides. The automation builds the list, the person judges whether it is the right list and whether the message is right for it. Every asset that reaches a real human being has been read by a real human being first.

Get that line right and you keep the speed and the cost advantage of automation with none of the risk. Get it wrong and you have simply found a way to make mistakes more efficiently.

## People first, software second

It is easy to describe this as an AI story. It is not, quite. It is a story about what to keep human when you no longer have to keep everything human.

Most agencies have quietly picked a side. They either automate everything and lose the human thread, or they stay manual and cannot keep up on cost. Neither serves the client well. The point of the model is to run on both: the leverage of software for the routine, the judgement of a person for everything that matters. People first, software second.

For the client, the experience is meant to be simple on top and serious underneath. You see a clear, confident plan and you deal with a person, not a portal. Beneath it sits the data, the automation and the reporting that make it work, and you never have to log into any of it unless you want to.

## The proof we hold ourselves to

An agency that sells this model and does not use it is selling theory. So the honest test is whether we run our own business on it.

We do. More than forty automations run AA2 itself: our own data, our own drafting, our own reporting. We use what we sell, on ourselves, before we build it for anyone. When we tell a client that automation can carry the routine and free the judgement, we are describing our own week, not a brochure.

That is also the fairest way to judge any supplier making this claim. Ask whether they use it on their own business. Ask to see specific, real results rather than adjectives. And read the work they send you, because a person who has actually thought about your business leaves fingerprints that a template never does.

## What this means for you

If you have outgrown DIY marketing but balked at a five-figure agency retainer, the gap between those two options used to be where good businesses got stuck. The leverage model is what fills it. Agency-grade work, a senior person accountable for it, and a cost structure that reflects paying for judgement rather than for a room full of hours.

The saving is real and it is structural. It is not a discount that reappears as thinner work. It is what happens when the routine load moves to the machine and the person keeps the part that was always the point.

Simple on top, serious underneath. That is the whole idea, and it is the reason a small, systems-driven agency can now outbuild a larger one that is still paying for the old maths.

### Common questions

**How can one person deliver what an agency team does?**

By changing what the person spends their time on. Automation and AI carry the repetitive load: the data building, the first drafts, the reporting, the scheduling. That frees the senior time for the parts that need judgement, which is strategy, message and sign-off. Add a network of specialists brought in for specific work, and a single accountable person can deliver the range a mid-size agency team covers, without the overhead of carrying that team full time.

**Does using AI mean the work is automated and impersonal?**

No, because the line is drawn deliberately. Automation handles the routine and the first pass. A person makes the final call and signs off anything a client or prospect actually reads or hears. The machine is fast and tireless and has no judgement at all, so it never gets the last word. That is the difference between marketing that runs itself and marketing that embarrasses you at scale.

**Why is a smaller, systems-driven agency often cheaper?**

Because the cost structure is different. A traditional agency prices in the overhead of a full team, offices and layers of account management. A systems-driven model lets AI carry the routine work, so you are paying mainly for senior judgement rather than for hours of junior time. The saving is real and it is structural, not a discount that gets clawed back in the work.

**What do you keep human, and what do you automate?**

Automate the repeatable and the mechanical: data building and verification, first drafts, formatting, reporting, scheduling. Keep human the strategy, the message, and sign-off on anything a client or prospect reads. The test is simple. If a mistake there would embarrass the client, a person owns it. If it is routine and checkable, the machine can carry it.

**How do I know the model actually works?**

Ask whether the agency uses it on its own business. We run more than forty automations on AA2 itself, for our own data, drafting and reporting, so the approach is not theory we sell and do not use. Beyond that, judge it the way you would judge any supplier: on specific, real results and on whether the work that reaches you has clearly been thought about by a person.


---