Services
Six disciplines that connect to each other. Each card opens with the problem it solves, not the technology: if you are not sure which one is yours, start with the one you have today.
The six disciplines
Almost no project uses just one. Pick where to start reading.
01 · Apps and web
Opening on an ordinary phone with bad signal, not only on the office fibre
Sites and applications measured on an ordinary phone with bad signal, not on the office fibre. For the browser, for the phone, or both from the same codebase.
What we build
- Sites and stores that have to show up in search engines and serve heavy images without blocking the first paint.
- Applications with users who come in every day: dashboards, client portals, internal tools.
- Phone applications built on the same codebase as the web, so a fix is made once and not three times.
What you get when it ships
- The repository in your name, with the full history, from day one.
- The technical decisions written down: what was chosen, what was discarded and why. One page per decision.
- Automated deployment: you push a change and it ships on its own, with no service downtime.
- A recorded handover session, with the system on screen.
- The LCP of the delivery measured on a mobile profile, with the measurement file alongside. The threshold we use is the Core Web Vitals one: 2.5 s.
What it does not include
- Two separate native applications, one for each store. If your case really needs them, we tell you so and we are not the ones for it.
- Copywriting, photography, video or translations.
- Content migration from a previous site.
02 · Artificial intelligence
Answering with your company information, not with what is on the internet
Assistants that answer with your company information, not with what is on the internet. And models that classify, extract and prioritise what someone does by hand today.
How an artificial intelligence system is put together
- InputA question or an event
- RetrievalYour documents and your data
- OrchestrationIt decides what is needed
- ToolsQueries, calculations, writes
- VerificationCross-checked before answering
- OutputAnswer with its source cited
What we build
- Assistants that search your documents and answer citing where each piece of data came from.
- Data extraction from invoices, delivery notes, contracts and emails, with human review in the cases the system flags as doubtful.
- Classification and prioritisation of what comes in: inquiries, incidents, orders.
What you get when it ships
- The system with its verification step built, not promised.
- The test cases it was validated with, and the ones it fails, written down.
- The log of every answer with its source, so you can audit an answer from three months ago.
- The cost per query calculation, so you know what scaling it costs before you scale it.
What it does not include
- Training a model from scratch.
- Computer vision.
- The cost of the models: it is contracted in your name and you pay it, so the day you want to leave you do not have to ask us for anything.
03 · Automation
Stop copying and pasting between WhatsApp, email and the system
The repetitive work that people spend their time on today: entering data, moving files, answering the same thing. We connect WhatsApp, email, your CRM and the internal systems you already use.
- 01TriggerWhatsApp, email, form
- 02ValidationFormat, duplicates, permissions
- 03DecisionYour business rules
- 04ActionsSystem, invoicing, alert
- 05LogAn auditable trail
What we build
- Sign-ups and orders that come in by WhatsApp, email or form and end up loaded into the system, with nobody copying and pasting.
- Automatic replies to what gets asked all the time, handing off to a person when the question is not one of the usual ones.
- Reconciliations and file cross-checks that today are done by hand once a week.
What you get when it ships
- The workflow running, with its logging stage: what happened, when and with what data. It is the stage almost nobody builds and the only one that lets you audit what happened at three in the morning.
- The business rules written outside the code, so you can change them without calling us.
- An alert when the workflow fails, one that reaches a person and not a log nobody reads.
- The manual for what to do when a third-party system goes down.
What it does not include
- Automating a process nobody has written down yet. If the process is not defined, the first thing is to define it, and that is Technical consulting.
- Licences for the tools the workflow connects.
04 · Technical consulting
Knowing what to build, in what order, and what a year of upkeep costs
Before building: what is worth doing, in what order and how much it costs to keep running per year. With the discarded options in writing and the reason for each one.
What we build
- The prioritised technical plan: what first, what next and what not now.
- The review of what you already have, with the risks ranked by what it costs when they fail.
- The cost of keeping each option running for a year, which is the figure almost never on the table when the decision gets made.
What you get when it ships
- One document per decision: context, options, the one chosen, the ones discarded and the consequences accepted. It is the format that then lives inside the repository, not a presentation.
- The diagram of the proposed system, at two levels of detail.
- The estimate with its range and with what would change it.
- A recorded closing meeting.
What it does not include
- Implementing what the plan proposes. That is a separate contract, and you can take it to another provider: the document is written to be understood without us.
05 · Infrastructure and deployment
Holding up the day traffic doubles, and without taking the service down
Where the system lives, what it costs per month and what happens the day traffic doubles. Updates ship without taking the service down.
This is not sold on its own. It comes included in everything we build. If you have a system we built, it goes straight in; if someone else built it, it has to be read first, and that is Technical consulting.
What we build
- The path of a change from the moment it is written until it is in production, automated and reversible.
- The separation between the environment where things are tested and the one your customers use, with data that is not the real data.
- Backups someone has tested restoring, not backups that merely exist.
- The alerts: what gets measured, at what threshold it goes off and who it goes off to.
What you get when it ships
- The infrastructure described in version-controlled files, not configured by hand in a console.
- The estimated monthly cost, broken down by service.
- The incident runbook: what to check first, how to roll back, who to call.
What it does not include
- Night on-call duty. We do not sell continuous operations or out-of-hours response, and saying so now is cheaper than finding it out on the day of the outage.
- Running your servers as a managed service.
- Migrations of closed enterprise systems.
06 · Security
Finding the hole before whoever is looking for it does
We look for the way into your system before someone else does, and we hand you the report with every finding, its severity and how it gets closed.
We do this on what we built ourselves, or on a system we can read in full with your written authorisation. We do not run blind reviews, and we do not run them on third-party infrastructure.
| System | Component | Severity | Status |
|---|---|---|---|
| [redacted] | Authentication | High | Closed |
| [redacted] | File upload | High | Closed |
| [redacted] | Role permissions | Medium | Closed |
| [redacted] | Dependencies | Medium | Closed with an update |
| [redacted] | HTTP headers | Low | Accepted by the client |
| [redacted] | Access log | Low | Open, with plan and date |
This is the shape of the report you will receive: each finding with its component, its severity and its status, and the affected system left unnamed. The content of a report is never published; the shape is, and it is what lets you know what you will have in hand when we finish.
What we build
- The application review: how data comes in, who can see what and what happens if someone changes an identifier in the address bar.
- The dependency review: which libraries you use, which ones have known flaws and which ones nobody maintains anymore.
- Handling of secrets and access: who holds keys, since when and what happens when that person leaves.
- The channel for someone on the outside to report a flaw to you without having to search LinkedIn for who to write to.
What you get when it ships
- The report with every finding, its severity, the affected component and the step-by-step reproduction.
- The fix for the findings that fall within scope, not just the list.
- A second pass after the fixes, to verify that they are closed.
What it does not include
- We do not have ISO 27001 or SOC 2, and we are not going to imply that we do. If your tender requires them, we are not the supplier.
- We do not sign compliance certificates: we are not a certification body. We work with the requirements that apply to you as acceptance criteria and we leave the system ready for whoever is entitled to audit it.
- We do not do social engineering or testing on people from your team.
What changes shape when you automate
We are not going to tell you how many hours you will save: that depends on your process and we have not seen it yet. What we do know is what changes shape.
| Before | After |
|---|---|
| BeforeEvery case goes through a person. | AfterEvery case goes through a rule. Exceptions go through a person. |
| BeforeThe error is found when someone notices it. | AfterThe error is found when it happens, and it raises an alert. |
| BeforeThe data is in the system, in the spreadsheet and in the email, and the three say something different. | AfterThe data is in one place, and the rest read it from there. |
| BeforeWhat happened yesterday is reconstructed by asking around. | AfterWhat happened yesterday is logged with its time and its data. |
| BeforeCapacity grows by hiring. | AfterCapacity grows as far as the system reaches. After that you still have to hire, but later. |
This is a diagram, not a measurement. The numbers for your case come from the Diagnostic, and they come with their method alongside.
The Allways model
A software project touches four areas, and the expensive part is never one of them: it is the seam between two. Here the four are in the same contract, with a single technical lead, so when something breaks nobody has to work out whose fault it is.
Product and architecture
What to build, in what order, and what it costs to sustain per year. With the discarded options in writing.
Build
Interface, server, data and integrations. One single code base and one single owner.
Operations
Where the system lives, what it costs per month, and what warns you when something breaks.
Testing and guarantees
Testing, accessibility, performance and security. Inside the budget, not as an extra.
The difference with an agency is not size: it is where the budget goes. With no sales layer and no account manager in between, it goes to architecture and code. The trade-off is written down and we do not hide it: we work on few fronts at a time. If you need five in parallel from day one, we are not the right fit.
How it starts
- You write to us or book a call. You tell us what is happening, not what technology you want.
- We tell you whether this is for us. If it is not, we tell you what kind of provider to look for. This happens more often than it sounds.
- If it makes sense, we send a proposal with the scope, the fixed price and the committed date.
- If you sign, the repository is created in your name that same day.
How much of your time it takes
A short meeting to start, one a week while it runs, and one at the end for the handover. Nothing else, unless you want more.
What is worth having at hand
Who decides, which systems you already use, and who we have to integrate with. None of this is required for the first call.
Questions
How much does it cost?
Prices are published on the pricing page, with what each package includes. Starter from 297 USD, Professional from 699 USD and Enterprise on request. Amounts do not include taxes, and that page is not a contractual offer: what binds both parties is the signed quote.
Why does it cost less than an agency?
Because there is no sales layer, no account manager and no office: the budget goes to architecture and code. The trade-off is scope: we work on few fronts at a time, in depth. If you need five fronts in parallel from day one, we are not the ones.
How long does it take?
It depends on the scope, so we do not estimate it here: the date of each stage is agreed in writing before starting and goes into the proposal as a committed date. If we run past it through our own fault, the discount set in that same proposal applies, and you read it before signing.
Is the code mine?
Yes, and the repository is created in your name on day one, not at the end. The transfer of rights takes effect as the code is written, so there is no moment when the work already done is not yet yours.
What happens if we fall out?
There is an exit plan written into the contract: you get the repository with its full history, the documentation, the schemas, the credentials and a handover note with what is pending and the known risks. The delivery deadline for all of that is in the proposal, in business days, before you sign.
Do you use artificial intelligence to write the code?
Yes, as a tool, always under human review, and the responsibility for the result is ours. It is written in the contract, and we say it here because we would rather you know before you ask.
I already have a technical team. Is this still useful?
It works for what your team does not get to, for reviewing a decision before you make it, or for starting something they maintain afterwards. What we do not do is replace your team by the week with no defined scope.
Do you work with my sector?
There are twelve sectors on the solutions page, each with the problem, what we build and what we measure. If yours is not there, the criterion is not the industry but the shape of the problem: the traits that define it are right there, and if you recognise most of them, we can probably help.
Not finding what you need?
Tell us the problem and we will tell you whether we can solve it. If we cannot, we will tell you who can.