Three questions we get asked constantly. Three answers you can check.
These are three strands of work we are in, written up so you can act on them without hiring anyone. Each piece is long, each one ends with something you can do yourself this week, and every claim in them either traces to public research we name or to a system we run.
There are three pieces because we have three things worth saying. When there is a fourth, it goes up.
THREE PIECES ENGLISH AND SPANISH EVERY CLAIM SOURCED
Start here
- You do not know what AI assistants say about your business. Read AI visibility. It contains the check, step by step, and it is free to run.
- Enquiries arrive in two languages and the reply is always late. Read bilingual enquiry handling. Five things to fix before anyone sells you software.
- Something got built last year and nobody uses it. Read why AI projects stall. Includes the four signs that yours is going the same way.
The three strands
-
Being found when the answer is written by a machine
STRAND 01 PARENT: AI VISIBILITY AUDIT
A customer used to type a question and get ten links. Increasingly they type a question and get one answer, with two or three businesses named inside it. Nobody outside those companies knows exactly how the names get chosen, and anyone selling you a position in that answer is selling something they cannot deliver. What you can do is find out where you stand, correct what is wrong about you, and give the machines something accurate to read.
What ChatGPT says about your business, and how to check it yourself
The ten questions to ask, how to ask them so your own history does not bias the answer, what to record, and what to do with each of the three results: not named, named wrongly, or named behind a competitor. About fifteen minutes a week for two weeks, and it costs nothing.
-
Working in two languages without doubling the payroll
STRAND 02 PARENT: OPERATIONS SYSTEMS
Half the businesses on this coast serve customers in two languages and staff for one. The interesting part is that translation stopped being the problem years ago. The problem is a queue: the enquiry waits for the one person who can answer it, and that person is also the person doing the actual job.
Two languages, one inbox, and the wait that costs you the job
The five points where bilingual handling breaks, the glossary hour that is worth more than any software, the mistake of answering fast in a language you cannot then continue in, and the three numbers to count by hand before you buy anything.
-
Why the thing that got built stopped being used
STRAND 03 PARENT: OUR METHOD
The published research on this is consistent and it is not flattering to anyone in our industry. Projects do not usually fail on the day they launch. They stall a few months later, quietly, and nobody writes it up. We think the reasons are recognisable enough to be caught early.
AI projects do not fail at launch. They stall in month three.
What MIT and RAND found, the five ways a working system falls out of use, the five things that prevent each one, and the four signals in month two that tell you which way yours is going.
READ
What goes on this page, and what does not
Publishing the standard costs less than being trusted, so here is ours.
Every number is attributed or it is cut. If a figure comes from public research, we name the institution. If it comes from a system we run, we say which system and in whose business it runs. If we cannot source it, it does not appear, however good it would look.
No client results, ever. Not in a post, not in an aside, not as an anonymised "one client saw". The reason is on the case studies page, and it applies here identically.
Both languages on the same day. Every piece exists in English and in Spanish, written rather than translated, and neither version ships without the other. That is why there are three rather than six.
Useful if you never speak to us. Each piece contains the method, not a summary of the method with the useful part behind a form. If you read one and go and do it yourself, that is a good outcome and we would rather you did it than nobody did.
No dates promised. We are a small studio. A publishing calendar we would miss is worth less than an empty slot we fill when we have something.
The longest things we have written are not on this blog
Two of them, and both are more use than a post if you are trying to decide something.
Our method, in full. Six phases, real timelines, what we need from you, the risks we can already name, and what happens when a paid step fails its test. It is the page a competitor would keep internal.
The Hairy Housemates case study. The content engine we sell, running in a business we own, with the real operational counts, a section on what it does not prove, and a section explaining the growth figure we refuse to publish.
is the paid version of the first strand above. The blog post tells you how to run the check yourself. The audit is what happens when we run it properly, in both languages, against three named rivals, with a plan at the end.
Four questions we have not written about yet
Named rather than hidden, because the gaps say as much as the pieces.
What an AI system costs to run once it is live. We have the numbers for our own systems and not enough engagements behind us to say anything general. Writing it now would be a guess with a chart on it.
Whether AI-assisted prospecting is legal here, in detail. The short answer is that it can be, under GDPR and LSSI-CE, with a lawful basis, a working opt out and a documented source. The long answer needs a lawyer reading it before we publish it, and it is not written yet.
Which sectors on this coast this suits. We have opinions and not enough evidence. When we have run the same diagnostic across enough businesses in one sector to see a pattern, that is a piece.
How a small team keeps a system running after the person who built it moves on. The most important operational question in this whole field, and the one we are least ready to answer in public.
If one of those is your question, a conversation answers it faster than an article. Book a call and ask it directly.
READ
Start with the one you can act on today
The visibility check needs no budget, no access to your systems and no meeting. It needs ten questions and a logged-out browser window.
If you would rather have it run properly, in both languages, against three named rivals, that is the audit. If you would rather talk first, book a call.