AI implementation

AI implementation for business

We do not sell chatbots or licences. We map the processes costing you the most time and replace them with a system that runs without you.

65 min process auditQuote in writingNo commitment

What implementation means here

In most companies AI ends with someone buying ChatGPT seats and nothing changing a month later. Implementation is different. We take one concrete process: enquiry handling, bookings, lead qualification, reporting. We build a system around it and wire it into the tools you already run. The result is not a tool to try out, it is a process that stops needing a person.

How an implementation runs, step by step

  1. Process audit

    65 minutes on how the work actually flows today. You leave with a map of where AI genuinely saves time, and a list of where it does not pay off. You get that second list too.

  2. Scope and quote

    You receive a document: what we build, in what order, what sits on your side and what it costs. In writing, so you can compare it against another offer.

  3. Build and testing

    We work on your data and your integrations. Before a customer sees anything, the system is tested against real cases from your business.

  4. Launch and support

    We deploy, train your team and stay on for maintenance. A system nobody in the company can operate is a cost, not an investment.

What we implement most often

Voice agent

Answers the phone 24/7 on reception and support lines. Books appointments, handles repeat questions, hands the call to a person when the case needs one.

Assistant across channels

Website, WhatsApp, Instagram. Advises, qualifies the enquirer and passes a ready contact to sales instead of leaving it in an inbox.

Process automation

Enquiries from forms and email land in your CRM described, scored and assigned to the right person. No retyping between systems.

Custom internal tools

When off-the-shelf products do not fit how you work, we build an internal tool around your actual process.

Implementations by industry

The process looks different in a clinic, a hotel and a shop. Here is what we implement most often and what it actually changes in each.

Clinics and practices

Reception answers the phone mid-appointment, or not at all, and the patient calls a competitor instead. A voice agent takes calls after hours and at peak: it books, reschedules and cancels appointments in your system, confirms by SMS and cuts no-shows. We integrate with the practice software you already run.

Hotels and apartments

Availability and price enquiries arrive around the clock, in several languages, across four channels at once. An assistant replies immediately, checks availability, takes the booking and escalates only the unusual cases. Fewer empty nights caused by an unanswered email.

E-commerce

Most pre-purchase questions are variants of the same dozen. An assistant on the site, WhatsApp and Instagram recommends products, checks order status and handles returns, so only genuine decisions reach a person.

B2B services

Form leads sit in an inbox until someone finds time. Automation replies within minutes, asks about budget and scope, scores the enquiry and creates a CRM record with the context a salesperson needs.

What you get at each stage

The stages describe what we do. This describes what stays on your drive, including if the engagement ends.

After the audit

A map of the process as it actually runs, with the places you lose time marked, plus a list of what is not worth automating and why. That document holds its value whether or not we work together, and that is not a courtesy — it describes your process, not our solution.

Before the build

Scope in writing: what gets built, in what order, what sits on your side, which access is needed and when. Price and schedule included. It should be concrete enough to compare against another firm's proposal, not just against their number.

After launch

A working system, a short guide for the team in a form somebody will actually use, and access to a panel showing what the system did. Without that last part automation is a black box nobody trusts.

Ongoing

A summary of how many cases the system handled alone, how many it passed to a person, and where it hesitated. That is the real measure of whether the investment works — not whether the technology impressed anyone in a demo.

Who is needed on your side

The most common reason a project stalls is not technology, it is one of these three people missing. Worth settling before the start rather than mid-way.

Someone who knows the process for real

Not whoever wrote the procedure, but whoever runs it daily and knows where the exceptions get made. Without them we build something that matches the documentation and not reality. We need them for two to four meetings.

Someone who can decide

An implementation raises settling questions: in this case, should the system act or ask? If every such decision waits a week for the board, the project doubles in length. One person with a mandate is enough.

Someone who grants access

An internal IT person or the firm that supports your systems. This is usually the longest item in the schedule, because access to industry software often sits with an external vendor on their own timeline. That is why we ask at the audit, not after signing.

What AI implementation costs

We do not publish ranges detached from the process, because they would carry no information. The same capability in a five-person and a two-hundred-person company is two different projects. Cost is driven mainly by the number of integrations with your systems, whether image or document processing is needed (usually two to three times more expensive), the number of languages, and whether the system only answers or also acts on your behalf. You get the quote in writing after the audit, with scope and schedule.

How long it takes

A first working implementation usually takes three to six weeks from signature. Simpler automations can be ready in two. Projects that integrate with industry systems run longer, because the pace is set by API access on the vendor's side rather than by our work.

When we advise against it

If a process is undocumented and looks different every time, AI will entrench the mess rather than fix it. Order first, automation second. We also advise against it if you want a one-off build with no support: model-based systems need correction after launch, and we would rather be accountable for the outcome than hand over a file and disappear.

What happens after launch

An implementation does not end on go-live day, and that is not a pleasantry. The first two weeks surface cases the tests never had, because reality always holds more variants than a process map. We correct behaviour as they appear. After that, support means three things: reacting when a vendor changes an API or your offer changes, reviewing the cases where the system hesitated, and adding scenarios that were out of scope at launch. Companies that drop support typically come back after six months with the same system, which stopped matching how they now work.

Frequently asked questions

Who owns the code, and what if we part ways?

You do, for what we built for you, and it goes in the contract. You get access to the solution and documentation sufficient to hand it to someone else. We do not build dependency on nobody else understanding it. Third-party tool subscriptions sit on your side separately and stay with you the same way.

How does payment work?

Normally in tranches tied to stages rather than to the calendar: part at the start, the rest on acceptance of a working implementation. The audit is paid and billed separately, because it ends in a document that holds value even if we go no further. Support is monthly and can be cancelled.

What if the implementation does not work as expected?

Scope and acceptance criteria sit in a document before the start precisely so this is not a matter of impressions. If the system does not do what we agreed, we fix it within the project. What we do not promise is a business outcome outside our control: we can be accountable for the agent booking appointments correctly, not for how many patients call.

Do you do the audit alone, without an implementation?

Yes, and more often than you might expect. Sometimes the conclusion is that tidying the process, or an off-the-shelf tool at a few hundred a month, is all that is needed. We would rather say so than sell a project that made no sense — the first costs us one project, the second costs reputation.

How long does the audit take?

The conversation is 65 minutes. You usually get the document within a few working days, and for a larger scope we need one more meeting with whoever runs the process. We do not drag it out, because an audit without a decision goes stale quickly.

Do you work remotely or on site?

Most projects run remotely and it works, because the work is conversations and access rather than presence. If a process has to be seen in person — on a factory floor or at a reception desk — we come. Usually once, at the audit stage.

Let us find out whether you have anything worth automating

65 minutes, no sales deck. You leave with a process map and a clear view of what makes sense and what does not, whether or not we work together.

Book a consultation