Below are the directions we work in, grouped by what is broken on your side rather than by how we are organised. Where we have an established practice, the work is described plainly; where we are ready to take something on but have not run that exact task before, it is marked “ready to take on” — so you know what you are buying.
Digital and marketing
This is where most requests come from, and with the most measurable pain: money leaves, enquiries do not arrive, and what happens between the ad account and the sales team is visible to nobody.
A model inside the company, not in someone else’s cloud
Documents and calls cannot go outside, and there is nobody in-house to run a model. The market offers either “set it up yourself” or a cluster costing tens of millions.
What we do. We size the setup to the real workload and operate it, and we state the cost of owning it — hardware, power, support — not just the rollout.
A working setup inside your perimeter, a rule for checking answers, and a named owner for quality. Data never leaves your network.
A role in the system does the repetitive work — not another hire
People spend their day on what repeats every day: sorting mail, reconciling documents, answering the same question, assembling a report. Hiring one more means salary, taxes and six months of ramp-up before they drown in the same routine.
What we do. We set up a role for one specific job — mail, documents, enquiries, copy, research, calls — and fit it into your process. The role sits where the data sits: inside your perimeter or on our server, your call. It takes the repetitive flow and leaves the decisions and exceptions to a person.
The role is written down as a process: what it does alone, what it shows a human, and where it must stop. It switches off in a minute, and nothing walks out of the door with it.
One AI SEO role does more in a week than a person does in three months
The retainer is paid, the report arrives, the enquiries do not. The contractor built the keyword set once when the site launched, writes pages against it one at a time, and reacts to a ranking drop by the next report.
What we do. We build the keyword set from live demand, cover every query group with a page, and keep the technical side clean. We rebuild that keyword set every month: demand moves, and a list collected at launch is stale by the first season.
The pages, the copy and the keyword set stay yours. Part ways and they keep working for you: the traffic rests on your pages, not on our retainer.
Advertising stops being a black box with a monthly invoice
The budget drains on schedule; the pipeline stays empty. What the agency changed last week, and what it did to the cost per lead, is not visible in the report — and “why did it get more expensive” goes unanswered.
What we do. We take the campaigns apart by phrase, placement, device and time, and present changes as a list: what we propose, why, and what it should do. Nothing reaches the ad account silently — you see every action before it is applied.
A record of every decision with its reasoning: any of them is undone by the same list that introduced it. Campaigns stay in your ad account; we do not move them into ours.
Pay less for visitors who were never going to buy
Traffic in the reports, silence in sales: three of fifty enquiries are real. Adding a captcha is frightening — real people leave along with the bots.
What we do. We filter automation before the click and before the form. It starts in observation mode with nothing blocked: you see who we intend to cut off, and only then it goes live.
A per-decision record: who was filtered, and on what signal.
You see the breach on the day it happens
For weeks the site serves someone else’s script or sends part of its visitors elsewhere, and the first to mention it is a visitor’s browser or a search engine handing out penalties. A downloaded theme and an open admin panel are held together by luck.
What we do. We check what the pages actually serve against a known-good copy, and watch the certificate, availability and what the server shows to strangers. You hear about a change when it happens, not in a monthly report.
A clear picture of what changed on the site and when, and a written procedure for a breach — prepared in advance rather than improvised in a panic.
Free site check
It is not always obvious what a crawler or a first-time visitor actually sees on your site: whether it is blocked from indexing, whether it serves a sitemap, what a missing page returns. Working it out yourself takes time, and calling in an auditor for this is premature.
What we do. We check the site at the address you give us: robots.txt, the sitemap, the response to a missing page, and the response headers — no login, nothing to install. Only what the site already shows to any browser.
A list of checks with a verdict on each: what is fine, what is worth fixing. Want a deeper look — the next step is the paid SSH server audit.
The site or system exists — nobody is left to develop it
It was built years ago, the developer vanished, and every change runs into “this needs rewriting from scratch”. Meanwhile form submissions never arrive, the catalogue goes stale, and the layout falls apart on a phone.
What we do. We take what is already written, get to grips with it and develop it further — without a rewrite where none is needed. We start by drawing the line: this is cheap to fix; this genuinely runs into the architecture and is cheaper to replace.
Changes in your own code, access to everything we touched, and a straight assessment of what is worth developing and what is cheaper to rebuild.
Your own model still does not mean correct answers
Once a model moves inside the trusted perimeter, its answers start being treated as “ours”. Data security quietly replaces checking whether the answer is right.
What we do. We set the boundaries: where the model decides, where it proposes to a person, and where it must refuse. We measure refusals and errors, not just speed.
A clear answer to who is accountable for a wrong output, and how it is caught before it costs money.
Books that do not rest on one person
One person knows everything. They leave and the work stops, while their replacement refuses to inherit the mess. Every quarter a director reconciles the same discrepancies by hand.
What we do. Documents are sorted by type, totals are reconciled, and any gap is shown together with what would close it. The automation never posts a document or pays on its own.
A departure stops being an emergency. The checks are written down and work without their author.
A client list is protected before the salesperson leaves
The account manager left and took the clients. Courts rarely award anything: the trade-secret regime was never put in place, and the contract had one general sentence.
What we do. We put the regime in place and separate access so it holds both on paper and technically. We do not install surveillance — it scatters a team faster than it protects a list.
A documented regime, separated access, and a clear procedure for when someone leaves.
Engineering and equipment
The other half of what we do, and where we started: electronics, firmware, someone else’s undocumented equipment, and sites that must keep running with no link to a cloud.
Your own device instead of someone else’s black box
An off-the-shelf controller imposes someone else’s logic: it knows nothing about your machine, and every change goes through an outside manufacturer and their queue.
What we do. We design the schematic, the board, the enclosure and the firmware for your task, and we own the whole chain — so nobody argues about whose fault it is at the seam between board, firmware and protocol.
Rights to the result, sources, build environment and design documentation. Another engineer can carry on without us.
The equipment is here, its author is not
The contractor delivered and left. The firmware runs, but nobody can change it: no sources were kept, there is no documentation, and the backup will not load back onto the controller.
What we do. We work out what is actually in there, recover how it behaves, and take it on as an ongoing service rather than a one-off rescue.
A plain description of how it works, a reproducible build, and someone to call.
The password, firmware and documentation nobody has
The cabinet was installed in the early 2000s. No documentation, no nameplate, no password. The supplier answers: we shipped controllers without software, ask the controller vendor.
What we do. We record how the equipment actually responds, rebuild the parameter map and the exchange order, and write the description that never existed.
Working access to your own equipment, and a document any engineer can service it by.
The controller is discontinued — the unit is not
Eleven years in service and the screen went dark. Replacing the whole cabinet costs more than the machine did, and the logic and setpoints tuned over years have nobody to carry them over.
What we do. We fit or build a replacement for the same slot and the same function, and migrate the logic with its setpoints instead of rewriting everything from scratch.
The unit behaves as before, and the description of its logic stays with you for next time.
You hear about the fault before the tenant below does
A rooftop unit froze and flooded everyone underneath. You found out from the tenant’s phone call, and the damage claim followed. Overnight cover means paying a human every month — or nobody watches at all.
What we do. We put in monitoring that tells “the link is up” apart from “the equipment is working”, and route alarms into one channel without the noise that gets it muted by day two.
The alert reaches you before the call does, and the event history exports anywhere, any time.
The site keeps running when the cloud does not
Six months in, nobody opens the dispatch system, and the site is welded to someone else’s server: lose the link, lose control.
What we do. Control stays on site; the cloud is the view and the history. A command is not lost when the link drops — the equipment collects it once it is back.
A site that runs offline, your data with export on demand, and a straight answer to “what if you disappear”.
We connect what you already run
The site is a zoo of brands and vintages. There is no single picture, and every supplier only talks about their own hardware.
What we do. We read what is already installed and bring it into one picture, without replacing working equipment for the sake of compatibility.
One screen for the whole estate and an exchange description that does not tie you to us.
Why an AI role outruns a human contractor
Look at what agencies publish as their own rates: a monthly retainer buys between four and fifteen articles, and the median output named in job ads is twelve. A quarter adds up to three or four dozen pieces. In a single day the AI role published forty on our site — a quarter’s worth in twenty-four hours. And not a sprint for the sake of a number, but one working cycle: build the keyword set, split it into query groups, cover each with a page.
Organic against paid: figures, not a deck
A site from our own group, an industrial niche, both channels running at the same time. The period is one quarter, 4 June to 2 September 2026, figures from Yandex Metrica. One site over one quarter: an illustration of the method, not a promise of the same numbers for you.
| Metric | Organic search | Paid ads |
|---|---|---|
| Bounce rate | 14% | 42% |
| Time on site | 1 min 38 sec | 52 sec |
| Form enquiries per 1,000 visits | 2.6 | 0.8 |
| Visits over the quarter | fewer than paid | more — and paid for |
We lose one row out of four, and it is here on purpose: paid brought more visits over the quarter, and on taps of the phone number the two channels are level — that gap sits within the noise. What holds is different: a visitor from organic bounces less often, reads for nearly twice as long, and leaves a form enquiry more than three times as often. The ratio is the same over a month, over a quarter, and over a stretch that overlaps neither, so it did not come from one lucky week.
How we grow marketing, step by step
The order matters more than the speed. Traffic poured into a site with a broken form and an open door for bots is a paid-for loss, so measurement and repair come first, growth second.
First we measure what is already there
first two weeksAnalytics and search consoles connected and split by channel, with goals on the form enquiry and on taps of the phone number rather than on “visited a page”. Until that exists, talking about growth is meaningless: nobody can say what grew, or from what.
Fix the leaking bucket before pouring traffic into it
first monthWe filter out bots before the click and before the form, check that form submissions actually arrive, and fix the technical side and the mobile layout. Plugging that leak is almost always cheaper than buying the same volume of traffic again.
Build organic search as an asset
months 1–3We build the keyword set from live demand, cover every query group with its own page, and put internal linking in place. Search engines take weeks, not days, to index the first pages — so the work starts at once, but the result is worth counting from month three.
Paid search supports organic instead of competing with it
in parallelWe take campaigns apart by phrase, placement and time. Queries the site already ranks for come out of paid — there is no sense in paying twice for the same click. The freed budget goes to new directions and to seasonal demand.
Count enquiries, then decide where to push
from month threeYou can see how many enquiries came from organic, how many from ads, and what each of them cost. From there the budget moves to where a lead is cheaper — not to where the report looks better.
What we do not promise
- •Rankings by a named date: we do not control the results page, and anyone guaranteeing a top spot by a date either misunderstands the mechanism or is counting on you not checking,
- •Growth without changes to the site itself: if the pages do not answer the query, adding more of them just adds useless pages,
- •That organic will replace advertising: it replaces the part of it where you pay a second time for your own ranking,
- •That any of this is free: the role costs money every month. For that money it covers a volume one person cannot get through.
What you get handed over
This is the real difference, and the only thing that genuinely protects a client. The work is not done when “it all works” — it is done when you can carry on without us:
- •Rights to the result — written into the contract, not implied
- •Sources and the build environment: not a dump of files, but what the project rebuilds from
- •Design and maintenance documentation for what you will be operating
- •Your data, exportable in an open format at any moment, without involving us
- •Pages, copy and keyword sets on your own site — the traffic stays with you, not with the agency
- •Ad campaigns in your own account, with a history of what changed, when and why
- •A description of how it works, written for the engineer who comes after us
- •The right to continue with another contractor without rewriting the system
Ask any contractor these
Ten questions that usually reveal how a project will end. Put them to us and to everyone you are comparing us with, then compare the answers rather than the promises:
How we start
- •We understand the task before quoting: we look at what is already installed and what can stay,
- •We name a price range and a timeline before work starts, together with what would change them,
- •We agree up front on the signs under which we will tell you it did not work out, and stop,
- •If the task is not ours, we say so and tell you who to take it to.