Question 1

What does a first conversation actually involve?

Accepted Answer

Thirty minutes, no deck. You describe what is going wrong; we say which half of the business it belongs to, roughly what it would cost, and whether it is worth doing. Where the answer is that you should not spend money with us, we say that — it happens often enough that it is a stated commitment rather than a nice sentiment.

Question 2

Do you work with businesses outside the twelve industries listed?

Accepted Answer

Yes. Those twelve are where we have the most specific knowledge and can be useful fastest, not a boundary. The honest difference is that the discovery stage takes longer when we are learning your operating model from scratch, and we price that transparently rather than pretending otherwise.

Question 3

Can we start with one small thing?

Accepted Answer

That is usually the right approach. Most engagements begin with a single service or a single desk, scoped so that the first useful output arrives in weeks rather than quarters. Six-month programmes that deliver nothing until month five are how both sides lose confidence.

Question 4

Do we have to buy both halves?

Accepted Answer

No. Plenty of clients buy only a build engagement or only a desk. The two halves being under one roof is what makes the seam work well when you do want both — it is not a bundling requirement.

Question 5

Who owns the code and the accounts?

Accepted Answer

You do, from the first commit. Repositories, cloud infrastructure, ad accounts, analytics and domains all live in your accounts under your ownership. There is no runtime licence, no proprietary platform, and no version of ending the relationship that requires a rebuild.

Question 6

What happens to our data?

Accepted Answer

It stays in systems you control wherever the architecture allows, and where data cannot leave your infrastructure the system is designed to run inside it. That constraint changes the architecture, so it is established during discovery rather than discovered during build.

Question 7

What if we want to stop?

Accepted Answer

Build engagements end at a stage boundary with everything transferred and documented. Desks run on a notice period — long enough to hire or transition, short enough that it is not a lock-in. Because the accounts and the record are already yours, stopping is an access change.

Question 8

How do you handle confidentiality?

Accepted Answer

Mutual NDA before any detail is discussed, if you want one. Agents working on your desk are trained on your material only and are not shared across competing accounts in the same category — that is a policy, not a preference.

Question 9

How do you price build work?

Accepted Answer

Fixed price per stage where scope is genuinely knowable, and time-based where it is not — usually because it depends on a system we have not seen yet. Discovery is priced separately and small, so you can buy the recommendation without committing to the build it recommends.

Question 10

How do you price the desks?

Accepted Answer

Against coverage and volume rather than headcount, because headcount is our problem to solve. The automated share reduces the cost and is reported weekly, so you can see the number you are paying for.

Question 11

Do you charge a percentage of ad spend?

Accepted Answer

No. A percentage rewards us for spending more of your money, which is a poor arrangement for you. Campaign management is a flat monthly fee based on account complexity.

Question 12

Is there a minimum commitment?

Accepted Answer

Marketing engagements need three months before the data means anything, so that is the minimum, then month to month. Desks have a notice period rather than a term. Build work is committed stage by stage.

Question 13

Why are there no client names or case studies on the site?

Accepted Answer

Because most of our work sits inside someone else’s operation, and a good deal of it is under agreements that do not permit us to name them. Publishing a logo wall we cannot substantiate would contradict the commitment two paragraphs above it. What we offer instead is method transparency — process stages named with their artefacts, commitments checkable during an engagement, and discovery priced small enough that you can test whether we are any good before committing to anything substantial. Where a reference is possible, we will arrange one on request.

Question 14

Will AI replace the people on our account?

Accepted Answer

It will take the repetitive share, and that is the point — it is what makes coverage affordable. What it does not do is take the calls that need judgement, and any vendor telling you otherwise has not run a desk. The split is measured and reported so you can see where the line actually sits rather than where it was promised.

Question 15

What happens when the AI gets something wrong?

Accepted Answer

It escalates to a person, the trace is reviewed, and the case joins the evaluation set so the same failure does not recur. Systems that cannot fail visibly are the dangerous ones, so ours are built to refuse rather than to guess — which produces more escalations, which is why we staff them.

Question 16

Are you going to tell us we need more than we asked for?

Accepted Answer

The opposite happens more often. We have declined app builds, recommended templates over custom sites, and pointed businesses at paid search instead of taking an SEO retainer. Scope that should not exist is the fastest way for an engagement to go badly for both sides.

Question 17

How long before something is running?

Accepted Answer

A scoped first system is typically six to ten weeks, with the evaluation set agreed in the first two. If a proposal promises production AI in two weeks, it is describing a prototype.

Question 18

Do we need to move our data somewhere?

Accepted Answer

No. Where data cannot leave your infrastructure, the system is designed to run inside it — that constrains the model choice and the architecture, so it is decided during discovery rather than discovered later.

Question 19

Which model do you use?

Accepted Answer

Whichever passes your evaluation set at acceptable cost and latency, and it is usually more than one — a capable model where judgement is needed, a cheaper one where it is not. Because evaluation exists, swapping models later is a measurement rather than a rebuild.

Question 20

What if the AI gets something wrong?

Accepted Answer

It will. The design question is what happens next: whether it fails visibly, whether the failure is logged with enough context to explain it, and whether the case joins the evaluation set so it does not recur. All three are built in.

Question 21

Who owns the code?

Accepted Answer

You do, from the first commit, in your repository and your cloud accounts. There is no runtime licence and no dependency on us continuing.

Question 22

What is the actual difference between an agent and a chatbot?

Accepted Answer

A chatbot produces text. An agent produces outcomes — it plans, calls tools that change real state in real systems, checks the result, and reports what it did. If the output is a message rather than a completed task, it is a chatbot.

Question 23

How do you stop it doing something harmful?

Accepted Answer

Four mechanisms, layered: permissions scoped per agent so it can only reach what its task needs; schema validation so malformed calls fail at the boundary; approval gates on consequential actions; and hard step and spend ceilings. Plus a full trace, so anything that does go wrong is diagnosable.

Question 24

Can it work with the tools we already use?

Accepted Answer

Anything with an API — Zendesk, Intercom, HubSpot, Salesforce, Shopify, Stripe, Slack, your own systems. Where there is no API we build the integration layer, and where the system is old we accept its terms rather than assume modern ones.

Question 25

How much supervision does this need?

Accepted Answer

Heavy at first and light later, deliberately. Shadow mode means every action is approved by a person; autonomy widens per task type as the approval rate proves it should. Most clients are still approving one or two categories a year in, and that is a healthy outcome rather than a failure.

Question 26

What happens when it encounters something it has never seen?

Accepted Answer

It escalates. The escalation rule is explicit and written by you — not a confidence threshold the model sets for itself — and the case is logged so the gap can be closed deliberately.

Question 27

Will callers know it is not a person?

Accepted Answer

Many will, and the agent identifies itself if asked — we do not build agents that claim to be human. What callers consistently care about more is whether the call got resolved on the first attempt, which is the thing voicemail never does.

Question 28

What happens when the caller asks something unexpected?

Accepted Answer

It transfers, on a rule you wrote rather than a guess. The person who picks up receives the transcript, so the caller does not start again.

Question 29

Can it actually book into our calendar?

Accepted Answer

Yes, against live availability, and it books only what your capacity rules permit — technician skill, drive time, bay availability, room type. A booking that cannot be honoured is worse than a missed call, so those rules are modelled explicitly.

Question 30

What does it cost to run?

Accepted Answer

Per minute of conversation, and the number depends on call length and complexity. We size it against your real call volume during the audit, so the comparison against a staffed line is arithmetic rather than a promise.

Question 31

What if it goes wrong at 3am?

Accepted Answer

Calls fall back to your existing line automatically if the agent is unavailable. Failure means callers get what they get today, never a dead line.

Question 32

Do we own the code?

Accepted Answer

Yes, from the first commit, in your repository and your hosting account. There is no licence, no proprietary builder and no version of leaving that requires a rebuild.

Question 33

How long does a site take?

Accepted Answer

A focused marketing site is typically six to ten weeks. Anything with real integrations — booking, commerce, an existing back office — depends on those systems, which is why the intent mapping stage exists before a number is quoted.

Question 34

Can our team edit it without a developer?

Accepted Answer

Copy, pages and services, yes. Structural changes to how a page type works are a development task, and we are explicit during design about where that line falls so it is not a surprise later.

Question 35

What about SEO?

Accepted Answer

The technical foundation is part of the build — server-rendered HTML, semantic structure, structured data generated from content, sitemap, and speed. Content strategy and authority building are a separate discipline, and we do that too, as a separate engagement.

Question 36

What happens after launch?

Accepted Answer

Support is a choice, not a lock-in. Some clients keep us on a retainer, some take it in-house with the documentation. Both are fine, because the ownership position makes both possible.

Question 37

Do we actually need an app?

Accepted Answer

Often not, and the first stage exists to answer that honestly. The test is whether the work happens offline, needs device hardware, or is repeated many times a day by the same people. If none of those hold, a fast mobile site wins on cost, on reach and on adoption.

Question 38

React Native or native?

Accepted Answer

React Native for workflow-driven apps where one codebase across both platforms is the dominant economic fact. Native where sustained camera work, background processing or platform-specific hardware makes the bridge a real cost. We make the call during the platform decision and show the reasoning.

Question 39

What about the app stores?

Accepted Answer

We handle submission and review under your developer accounts. Review timelines belong to Apple and Google, so we build in the buffer rather than promise a date we do not control.

Question 40

How do you handle updates after launch?

Accepted Answer

Over-the-air updates for the JavaScript layer where the platform permits it, so content and logic fixes ship without a store round trip. Native changes go through review. Both paths are set up before launch.

Question 41

Who owns the app?

Accepted Answer

You do — code, store listings and developer accounts, all in your name from the start.

Question 42

Which channel should we start with?

Accepted Answer

Usually the one capturing demand that already exists — search, for most service businesses — because it is measurable fastest and needs the least creative investment to validate. Demand creation is a better second move than a first one.

Question 43

Why do you start with our response process?

Accepted Answer

Because in most businesses we audit, it is the largest available improvement. Cutting median response from four hours to fifteen minutes typically moves more revenue than any bid or creative change, and it costs less.

Question 44

Do you require a long contract?

Accepted Answer

Three months minimum, because measurement below that is noise, and month to month after. Accounts are in your name throughout, so leaving is possible at any point.

Question 45

Why does platform ROAS disagree with our revenue?

Accepted Answer

Because each platform claims conversions it influenced, and several platforms claim the same one. That is why we run holdout tests — they measure incremental revenue rather than attributed credit, and the difference is often large.

Question 46

Can you work with our internal marketing team?

Accepted Answer

Yes, and it is usually the better arrangement. They hold the brand and the product knowledge; we bring measurement discipline and execution capacity.

Question 47

What is the difference between SEO, AEO and GEO?

Accepted Answer

SEO earns a position in a results list. AEO earns the answer box above it. GEO earns a citation inside a generative assistant’s response. They share a technical foundation and diverge in content structure — a page written to rank is not automatically a page that gets quoted.

Question 48

How long before we see results?

Accepted Answer

Technical fixes can move things within weeks. Content and authority compound over six to twelve months. Anyone promising first-page results in thirty days is describing paid search or something worse.

Question 49

Do you buy links?

Accepted Answer

No. It is a policy violation with a delayed and severe cost, and the recovery is worse than the original problem. Citations come from directory accuracy, genuine industry placement and publishing something worth referencing.

Question 50

How do you know if an assistant is citing us?

Accepted Answer

We query the major assistants on your priority questions on a schedule and record what they return and who they attribute. It is sampling rather than a complete measure, and it is currently the honest version of tracking this.

Question 51

Can you work on a site you did not build?

Accepted Answer

Yes, and most engagements are exactly that. The technical audit tells us what the platform will allow — some make certain fixes impossible, and we say so early rather than bill against a wall.

Question 52

What budget do we need to start?

Accepted Answer

Enough to generate statistically meaningful data in your cost-per-click range, which varies enormously — a home services account can learn on a fraction of what a legal account needs. We size it from your actual auction data during the audit rather than quoting a universal minimum.

Question 53

Should we use Performance Max?

Accepted Answer

Sometimes, and rarely as the whole account. It will spend everything it is given with limited visibility into where. It works where the conversion signal is accurate and valuable; where tracking is weak it efficiently buys the wrong thing.

Question 54

Our leads are phone calls. Can you track those?

Accepted Answer

Yes, and in phone-led businesses it is the first thing we fix. Dynamic number insertion attributes the call to the campaign and keyword, and outcomes are recorded against it — so the platform can optimise toward booked work instead of clicks.

Question 55

Can you fix our existing account or must it be rebuilt?

Accepted Answer

Both happen. Accounts with useful conversion history are worth restructuring around; accounts whose history was recorded against broken tracking are usually worth rebuilding, because the learning is built on bad data. The audit answers it with evidence.

Question 56

How do you charge?

Accepted Answer

A flat monthly fee based on account complexity, not a percentage of spend. A percentage rewards us for spending more of your money, which is a poor arrangement for you.

Question 57

Why did our ads stop working?

Accepted Answer

Almost always creative fatigue rather than an account problem. The same audience has seen the same asset too many times. Accounts that survive this are the ones producing enough new creative to replace winners before they decline, which is a production capacity question more than a media one.

Question 58

How many creatives do we need?

Accepted Answer

More than feels reasonable. A meaningful test programme needs new concepts entering every week, not a quarterly batch — which is why we build the production system first rather than commissioning assets one at a time.

Question 59

Is the ROAS Meta reports accurate?

Accepted Answer

It is accurate about what Meta observed and it systematically claims conversions it influenced rather than caused. That is why we run geo holdouts: they measure what would have happened without the spend, and the gap is often large enough to change how much you spend.

Question 60

Do we need to appear on camera?

Accepted Answer

It helps in some categories and is not required. Product-led, motion and text-led formats all perform. We test format as a variable rather than assuming what your audience responds to.

Question 61

Can you produce the creative or do we?

Accepted Answer

Either. We often build the system and templates so an in-house team produces volume cheaply, which is the more sustainable arrangement. Where there is no in-house capacity, we produce.

Question 62

Are the agents dedicated to us?

Accepted Answer

On dedicated plans, yes — the same named people, trained on your product. Shared plans cost less and use agents covering several accounts in the same category. We recommend based on your volume rather than on margin.

Question 63

How much gets handled without a person?

Accepted Answer

Typically half to three quarters for e-commerce and lower for technical products, and we measure it during the volume study rather than quoting an industry figure. The number is reported weekly, so you can see it rather than trust it.

Question 64

What if an agent does not know the answer?

Accepted Answer

They escalate to your team on a rule you set, with the conversation and context attached. Escalation is treated as a knowledge gap and closed, not as an agent failure.

Including the ones you would rather not ask.

Working together