Frequently asked questions

It's the work of taking AI out of the conversation and into the business's daily routine. It starts with mapping: where time and money leak, what can be streamlined, and where AI can genuinely help. Then implementation: choosing the right solution for each process, connecting it to the systems you already use, and bringing it into the work step by step. And we don't leave after launch. Sometimes the answer is an off-the-shelf tool, set up properly. Sometimes it's an automation between systems. Sometimes it's a tool or an agent built specifically for you. The difference from a vendor is that we don't arrive with a product in hand, but with a question: what does the business need.

Any business. A solo professional with a few repetitive processes, a clinic with a team, a company with departments, or a large organisation. The path is fitted to the size and situation of the business; the method stays the same: mapping, implementation, security and support. If what the business needs in the end is a simple off-the-shelf tool, the mapping will say so, and we'll help you choose it well.

A short intro call, no slide decks. You tell us about the business, we ask questions, and we tell you honestly whether and how we can help. If there's a fit, the first step is always mapping: one or two sessions that end with a written plan — prioritised processes, a proposed solution for each, and a measurable target. You can take the plan and carry it out yourselves, with another vendor, or with us.

Mapping: one to two weeks. A first pilot on one process: usually two weeks to a month from approval. Expanding to more processes happens gradually, based on what the pilot proved. These timelines depend mostly on your availability and on how complex your existing systems are, not on the technology.

We start from the goals, not the technology: what the business wants to achieve in the coming year, where it loses time and money, and what sets it apart from competitors. From there we map the processes where AI can drive real change, and prioritise them by value and risk. The result is a plan with measurable milestones: what gets implemented first, what gets measured, and when to expand. That way the investment is checked against a target set in advance, not against a feeling.

Every process has a balance point: the tool has to be easy enough for the team, cheap enough relative to what it returns, and an exact fit for the task. A well-configured off-the-shelf tool beats a custom one in many cases — and the other way round in others. We follow the models and tools continuously, choose the right one for today, and plan in advance for what happens when it changes tomorrow: a swap, an update, or a move to another provider, without rebuilding everything.

No. That's the situation in most businesses, and it's part of the mapping. We usually start with a process that doesn't need perfect data, and in parallel organise what the next stage requires. Cleaning up data is usually a result of implementation, not a precondition for it.

After mapping: choosing the solution for each process, a focused pilot on one process, integration with the systems the business already runs, and full implementation in the team's daily routine. Each stage is measured before moving to the next. When the pilot works day to day, we expand. If it doesn't, we stop and understand why before investing more.

Almost never. The value of AI is measured by what it can do inside your existing systems: calendar, email, invoicing software, CRM, online store. The solution connects to them through official interfaces, without asking the team to learn new software. If the mapping shows that a particular system is the bottleneck, we'll say so and suggest an alternative. But that's the exception.

An automation connects systems by fixed rules: a lead comes in, gets logged, gets a reply. It is cheap, stable, and suited to clear, repetitive processes. An AI agent understands a request in plain language, decides what to do, and acts across several systems. It suits processes that need judgement, like managing a calendar or answering customers, and it needs working rules and approvals. A custom tool is built when nothing off the shelf fits: an internal assistant over the business's own documents, a dashboard, a catalogue generated from a spreadsheet. What fits you is decided in the mapping, by the process — not by what's fashionable.

For some tasks, yes — and when that's true, it's part of our recommendation. But ChatGPT doesn't know your systems, doesn't remember the business from one conversation to the next, and doesn't take actions in the real world: it won't book a meeting, issue an invoice or update stock. Implementation starts exactly where the chat ends: connecting to the business's data and systems, working rules, and someone accountable for keeping it running.

Each business gets its own environment, isolated from every other client. During mapping we define what stays inside the business, what goes to the cloud, and what permissions each component has. Keys and secrets are kept in a secrets manager, never in code or in a chat. Data is encrypted in transit and at rest, and every action is logged. When the data is especially sensitive, components can stay on a server inside the business.

Language models make mistakes. The design assumes that from the start instead of hoping they won't. Three safeguards: limiting the system to the business's verified information, accuracy checks on answers before they go out, and human approval for every binding action. In practice: drafting a message to a customer can be automatic; sending it needs approval. An invoice draft is created on its own; issuing it goes through you.

The limits are defined during mapping, in writing, before the system touches real data. As a rule: anything that commits the business externally — money, customers, documents, publishing — goes through human approval. Internal, repetitive actions, like filing or a reminder, can run on their own. Over time, as the system proves itself, what it's allowed to do can grow. The decision is always yours.

It happens, which is why the solution is designed so the model or tool is a replaceable component, not the heart of the system. The business's knowledge, working rules and system connections stay with you and belong to you. As part of ongoing support, we track changes at the providers and handle them before they turn into a problem.

Launch is the start of the work, not the end. Ongoing support covers continuous monitoring, backups, model and tool updates, availability for questions and issues, and improvements driven by real use. There is one point of contact accountable for keeping the system running, even as the business changes.

AI takes on the repetitive tasks: scheduling, record-keeping, follow-up, reports. The time that frees up goes back to work that needs judgement, a personal touch and selling. For that to actually happen, the team is trained on the tools that were implemented and gets clear working rules: what the system does on its own, what goes through a person, and how to flag something that doesn't look right.

No. That's exactly why there's someone alongside you. You need to know your business; we know the technology. Training is done on what was implemented for you, in the context of your day-to-day work — not a general lecture about AI.

Yes. Mapping stands on its own: it ends with a written plan you can carry out yourselves or with another vendor. Equally, you can come to us for support only, if you already run AI systems and need someone accountable for them. The four areas — strategy and mapping, build and implementation, security and risk, training and support — are sold separately and stack together. The only rule: we always start with mapping.

Every project is priced in a personal proposal, based on what the business needs.

Every implementation is measured against a target set in advance during mapping: time saved, response time to customers, share of tasks closed without a human hand, errors avoided. The numbers are checked in the pilot and reported throughout support. If the target isn't met, that's the most important data point there is, and we deal with it before expanding.

Because a panda does exactly what's needed, without fuss, and then goes for a rest. AI that works properly looks the same: quiet, steady, and doesn't demand attention. And if that made you smile, you've got the tone.