Hours before tools
Every engagement starts by counting where the week actually goes. If we can't put a number of hours against a task, we can't tell you whether automating it is worth the money — and neither can anyone else.
About
You own the tools. We make them work. Everything else about how we operate follows from that — the pricing, the boundaries, the fact that we'll tell you when the answer is a smaller project than you asked for.
We don't sell software. We build the AI systems your team will actually use, train them to use it right, and make sure it delivers. The tools you licence directly — we make them work.ArkitechAI · operating principle
Most AI consultancies make their money in one of two places: reselling licences with a margin on top, or selling a strategy document and leaving before anything gets built. Both create the same problem for the client — advice that's quietly for sale, and a gap between the deck and the working system.
We took the other side of that. You buy your AI and automation licences directly from the vendors at their price. We charge for the labour, the expertise and the systems we build, which means when we tell you a cheaper tool will do the job, there's nothing in it for us but a shorter invoice.
The second half matters as much: we stay after launch. An automation that nobody owns drifts the moment the business around it changes — a form gains a field, a supplier changes their invoice layout, someone leaves. The retainer exists because that maintenance is real work, and the alternative is a set of clever automations quietly rotting for six months until somebody notices.
How we operate
Every engagement starts by counting where the week actually goes. If we can't put a number of hours against a task, we can't tell you whether automating it is worth the money — and neither can anyone else.
The first thing we build is deliberately something you'll notice, running in production, inside 30 days. Proof before scale. It's a better way to spend your money and a harder way for us to hide.
Everything lives in your accounts, under your ownership, documented well enough that another firm could take it on. Dependency is not a retention strategy we're interested in.
Some processes need fixing before they're worth automating, and some don't need automating at all. Saying so costs us revenue in the short run and is the only reason the recommendations are worth anything.
Who you'll be working with
ArkitechAI came out of a straightforward observation made across a lot of client work: the tools were never the bottleneck. Businesses already had the software. What they didn't have was anyone whose job it was to connect it, automate the parts that repeated, and then stay long enough to keep it all running.
The work is deliberately hands-on. You deal directly with the person building the thing — not an account manager relaying requirements to someone you'll never meet. That limits how many clients we take at once, which is a trade we're happy with.
Engagements are run from Lisbon and delivered remote-first. Discovery, training and support all work fine over video; when a build genuinely benefits from a day on site, we'll say so.
Data & confidentiality
Automation work means being handed access to the systems a business actually runs on. Here's how we handle that.
Automations, assistants and integrations are created within your own tenancies and subscriptions wherever the platform allows it. Your data doesn't take a detour through ours.
Each workflow gets scoped to the minimum it needs to do its job. Where access has to be broad during a build, it's narrowed at handover — and we tell you what we hold.
We sign a non-disclosure agreement before discovery and a data processing agreement before we touch a production system. Both before any access is granted, not after.
Where personal data is involved we design for lawful processing from the start — minimising what's collected, keeping processing inside the EU where it's required, and documenting the flow so your compliance people can read it.
Working in a regulated sector — healthcare, finance, legal? Say so on the first call. It changes how we design a workflow, and it's much cheaper to design for than to retrofit.
Let's talk
Tell us what a normal week looks like and we'll tell you which parts of it a machine should be doing. If the answer is "none of it", you'll hear that too.
Good fits for this practice