The buyer knows. They always know, usually inside a second, and the thing they do next is close the tab or type something short and rude to see what happens.

We're not anti-AI. We build with it every week. But there's one place it shouldn't be, and it's the place almost everyone has put it: in front of a person who has just decided to buy something.

Think about what actually happened before that click. A founder has known for about three weeks that the pipeline is thinner than it should be. He's had two bad nights about it. On a Tuesday afternoon between meetings he finally does something, lands on your site, reads a bit, believes you, and reaches out. That's the highest-intent moment you will ever get from him, and it is also the most fragile. He's already braced for it to be a waste of time.

Then a bot asks him what his budget range is.

The tragedy is that the widget was installed to solve a real problem. Enquiries were sitting unanswered overnight and over weekends, and somebody sold a fix. The fix just got aimed at the wrong half of the job.

So we draw a line, and it's held up across every build we've done.

AI is allowed to acknowledge, route, and book. Reply in under sixty seconds, by name, referencing the exact thing they asked about, offer two times this week, put it in the diary, confirm it, remind them the day before. That's admin, it's the part humans do badly at 9pm on a Sunday, and buyers don't resent it because nothing is being sold to them.

AI is not allowed to qualify, handle objections, discuss price, or hold the conversation where the decision gets made. The second it starts asking about budget and timeline, it stops being a receptionist and becomes a toll booth on the road to your sales team. You've automated the one interaction where a human being would have changed the outcome.

Cold is different. A dormant database of two thousand old enquiries, nobody in a buying moment, no expectations to damage: that's exactly where AI conversation earns its keep, and it's usually the first thing we turn on because it makes money out of something the client already owns. Nobody's precious about being messaged by software when they'd forgotten your name entirely.

The distinction is intent. High intent gets a human, fast. No intent gets a machine that's polite and persistent.

Now the uncomfortable bit, which is that most of the businesses installing chat widgets don't have a speed problem they can automate their way out of. They have a rota problem. The enquiry arrives, it goes to an inbox three people can see, which means it belongs to nobody, and it sits there until someone happens to look.

The research on response time is old now and the exact multiples vary between studies, so treat the numbers you see quoted on LinkedIn with some suspicion. The direction has never been in dispute: replying inside five minutes beats replying inside an hour by a large margin, and by the next day you're mostly writing to someone who's already spoken to a competitor. Speed doesn't make you better than the other firm. It makes you first, and first is usually enough at this stage.

Do this before Friday. Fill in your own enquiry form as a buyer would, from your phone, on a Tuesday afternoon. Use a personal email address so nobody recognises it. Then do it again at 8pm. Both times, write down the clock time and don't tell your team. Now count the minutes until a human being who knows something about your business says something useful to you.

If the answer is under five minutes twice, you're in rare company and this issue isn't for you. If the second one is still sitting there on Thursday, you've just found the cheapest revenue in your business, and it costs nothing to fix except deciding who owns the first reply and by when.

One more thing worth checking while you're in there: if you do run a bot, does it tell people it's a bot? From the 2nd of August this year that stopped being a matter of taste anywhere your output reaches the EU. We went through it a fortnight ago; the archive's on the site.

If you want us to look at where your enquiries are going and what happens to them, there's a 30-minute call on the site. No deck, no pitch. We'll tell you what we'd fix first and you can go and fix it yourself if you like.

Ryan & Ali

Cost per lead is the worst headline metric in marketing, and the reason is simple enough that it should have killed it years ago: you can improve it by making your leads worse.

Widen the targeting. Drop the qualifying question from the form. Swap "book a call" for "download the guide". Do all of that and your CPL halves by Friday. Your agency has a great month. You have more names, more admin, more calls with people who were never going to buy, and less cash than you had in June. Nobody lied to you. The number genuinely went down.

The same trick works on almost every vanity metric in the stack. Reach goes up if you stop caring who sees it. Reply rate goes up if you ask for less. Pipeline value goes up if you let your salespeople guess at the numbers, which they will, generously, on a Friday afternoon. Any metric a supplier can improve without improving your bank balance will eventually be improved without improving your bank balance. That isn't cynicism about agencies; it's just what happens when you pay people for activity.

So what should be on page one instead? Three numbers. We've never needed more than three to work out whether a business is growing or just getting busier.

One: qualified opportunities per week. Not leads. An opportunity is a named person, in a business that fits, who has agreed to a conversation about buying and turned up to it. Weekly, not monthly, because monthly hides a bad fortnight and you'll only notice in the quarter after the one where it mattered. Write down the number that would be enough, then measure against it. For most founder-led businesses turning over £10k to £60k a month, the honest answer is somewhere between three and eight, and they're doing two.

Two: median time to first human contact. Median, not average. One enquiry answered in four days will drag your average into fiction while ten answered in six minutes look fine. This is the cheapest number in the business to move and the one most likely to be quietly awful, because it isn't anybody's job.

Three: cash collected per opportunity created. Not revenue. Not invoiced. Collected, in the bank, from opportunities that entered the pipeline in a given month, measured however long it takes them to close. This is the one that stops all the arguments, because it prices your leads properly. Twenty opportunities that pay you nothing for four months are worth less than eight that pay in three weeks, and no dashboard built around CPL will ever tell you that.

Run those three together and something useful happens: they can't be gamed against each other. Loosen your targeting and volume goes up but cash collected per opportunity falls, so you see it immediately. Cherry-pick only the perfect fits and cash per opportunity looks lovely while the weekly count collapses. You're forced to hold both, which is what actually running a business feels like.

The reason this matters more in 2026 than it did in 2021 is that lead volume has stopped being scarce. Anyone can generate names now, cheaply, at whatever quality you're willing to accept. The scarce thing is a conversation with someone who can pay you and has decided to. Optimising the abundant input while ignoring the scarce one is how good businesses spend two years getting busier and no richer.

Do this before Friday. Open your CRM and pull the last 90 days. Count the opportunities: real ones, people who showed up to a call. Divide by 13 for a weekly number. Then pull the money that's actually landed in the bank from that same cohort, and divide. You now have two of the three, and roughly ten minutes of work will tell you more than the last year of monthly reports.

Most founders who do this find one of two things. Either the weekly opportunity count is fine and the money isn't, which is a sales and offer problem and no amount of new traffic will fix it. Or the money per opportunity is healthy and there simply aren't enough of them, which is the good problem, and the one you can spend your way out of. They need completely different work, and the CPL slide can't tell them apart.

If you'd rather we did that pull with you and told you which of the two you've got, there's a 30-minute call on the site. Bring the CRM export. We'll tell you what we see, whether or not you ever work with us.

Ryan & Ali

Caveat first: we're a growth consultancy, not a law firm, and none of this is legal advice. Use it to work out whether you need to ask someone qualified. Where we're not certain, we say so rather than guessing.

Start with the bit people get wrong. The Act reaches past the EU's borders: it bites on businesses outside the Union where the output of an AI system is used inside it. A UK company with EU customers, EU-based staff, or EU prospects sitting in an outbound sequence can be in scope with no legal entity anywhere in Europe. The precise test is fiddly, and it's the first thing we'd send to a lawyer if you sell into Europe at all.

The timeline, and these dates we're confident about. The Act came into force on the 1st of August 2024. Bans on certain practices, plus a duty to make sure the people using AI in your business actually understand it, started on the 2nd of February 2025. Rules for general-purpose AI models, governance and penalties came in on the 2nd of August 2025. Then Sunday just gone.

Two things about that date. The transparency obligations under Article 50 apply from it. Meanwhile the high-risk rules that were also meant to arrive have been pushed back by an amending regulation, usually called the Digital Omnibus on AI, agreed in the spring and adopted over the summer: standalone high-risk systems now have until December 2027, and AI embedded in regulated products until August 2028. Treat those two as directionally right rather than gospel, because the omnibus moved fast and we'd want them confirmed against the published text before anyone relies on them.

The reassuring part, for most readers of this newsletter: if you run a small founder-led business with a chatbot, an outreach tool, a transcription service and something that drafts your copy, you're almost certainly a deployer of limited-risk systems rather than a provider of high-risk ones. The heavy machinery in the Act isn't pointed at you. Transparency is, and that's a much shorter list.

So, the checklist. Six things, and you can do most of them yourself.

1. Write down every AI system that touches a person outside your company. Website chat, phone answering, outbound sequences that generate their own replies, AI notetakers on client calls, anything that scores or sorts applicants. Most founders find between four and nine and are surprised by two of them. You cannot comply with a list you don't have.

2. Anything that holds a conversation must say it's software. Article 50 requires people be told they're interacting with an AI system, unless that's obvious from the context. In practice: a line at the top of the chat, a sentence in the voice greeting, a footer on automated replies. Plain language, not buried in your terms.

3. Mark AI-generated content aimed at the public. Synthetic images, audio and video fall under the marking and disclosure rules, as does text published to inform the public on matters of public interest. A grace period into December 2026 is reported for the machine-readable marking of systems already on the market; check that before relying on it. Ordinary marketing copy, drafted by a model and edited by a human, isn't what this provision is chasing.

4. Work out which side of the line you sit on: deployer, or provider. This one catches people. If you take a vendor's chatbot, put your own name on it and sell it on, you may have become a provider with a heavier set of duties. If you're simply using a tool a vendor sells you, you're a deployer. Anyone white-labelling AI as part of their offer should get this checked properly rather than assuming.

5. Do something real about AI literacy, and write it down. The duty to make sure staff using these tools sufficiently understand them has applied since February 2025. An hour's session plus a one-page policy on what may and may not go into a model, dated and minuted, is proportionate for a team of eight.

6. Ask your vendors for their paperwork. Every AI tool you pay for should be able to say what it does, what it does with your data, and how it meets the Act. A supplier who can't answer that in August 2026 has told you something useful.

On penalties we're deliberately not quoting figures. The regime is tiered, the top band is calculated against worldwide annual turnover, and smaller companies are treated more leniently than the headline numbers on LinkedIn suggest. Those amounts are easy to look up and easy to misquote, so look them up.

Do this before Friday. Just item one. Twenty minutes, a blank page, every AI system that speaks to a customer. Everything else gets easier once that page exists, and if you find nine of them with disclosure on none, you've found your weekend job.

If you'd like a second pair of eyes on what your stack is doing to your leads, commercially as much as legally, there's a 30-minute call on the site.

Ryan & Ali

That's the part nobody sells you. Every one of those fixes has a return, and the return depends heavily on what you fixed before it. Do them in the order they occurred to you and you'll spend nine months and get four separate small improvements that mostly cancel each other out. Do them in the right order and each one raises the ceiling on the next.

We named our delivery model after Aristotle for exactly this reason: the whole is greater than the sum of the parts, but only if you assemble the parts in sequence. It's a slightly pretentious name for a very practical rule.

The sequence below hasn't changed in two years of running it.

Money you already own comes first. Before anything new is built or bought, we go into the database: old enquiries, quotes that went quiet, closed-lost deals, customers who lapsed in 2024 and were never contacted again. You've already paid to acquire every one of those names. Most businesses have somewhere between four hundred and four thousand of them sitting in a CRM nobody has opened since the last person who cared about it left. Worked properly, with a real reason to be in touch rather than a "just checking in" blast, this is usually the fastest cash in the entire engagement, and it costs nothing in media spend.

Trust comes second, because it multiplies everything downstream. Reviews, referrals, whatever proof your market actually checks before they'll speak to you. Almost nobody asks for this systematically; they wait for it to happen. The reason it goes here rather than later is arithmetic: every lead you generate from this point forward, organic or paid or referred, converts at a better rate because of work you did once.

Then the leaks. How long a new enquiry sits before a human replies, what happens to calls nobody answers, whether the form on the site actually delivers anywhere useful. This is unglamorous and it's where most of the recoverable revenue in a small business hides, because it isn't anyone's job and therefore never gets measured. Fix it before you buy traffic, not after, or you're paying for leads twice.

Then the people who close. There's no sense pushing more volume at a team that can't convert what it already gets, and most salespeople in businesses under ten staff have never had a day of formal training or a written process. Grade three real calls before you change anything, so the improvement is provable rather than a feeling.

New demand goes last. Ads, outbound, whatever channel you were going to start with. By the time you get here, the money is going into a system where the follow-up is fast, the proof is visible, the database is warm and someone competent is picking up the phone. The same spend produces a different result, and you're buying demand from a position of profit rather than hope.

Now notice what happens if you run that list backwards, which is what usually happens. You buy traffic into a business with a 40-hour reply time, six reviews and no sales process. The leads are fine. The channel gets blamed. You change agencies, and the new one generates leads that also don't convert, and by month four you've concluded that ads don't work in your industry. They work. They were just the fifth thing you needed, done first.

There's a commercial argument buried in this too, and it's why we do reactivation before anything else. It creates value before any new machinery has to be built, and it produces the baseline data that makes everything afterwards measurable. If you don't know what was happening before, no improvement you make later can be proved, and by next spring you'll be arguing with your agency about attribution instead of counting money.

Do this before Friday. Write out the four or five things you were planning to do next quarter. Beside each one, answer two questions: does it need new money going out of the door, and does doing it first make any of the others work better? Re-sort. In our experience the thing that ends up at the top is almost never the thing that was at the top, and it's usually free.

Then do one more thing, which takes ten minutes: export every contact in your CRM who enquired and never bought, and count them. Multiply by whatever a client is worth to you. That number is the reason the sequence starts where it does.

If you want us to sort your list with you and tell you what we'd do in the first thirty days, there's a 30-minute call on the site. No deck. Bring the list.

Ryan & Ali