Automation: the first item in the plan
These usually come first, because the return shows up in weeks: workflows, chatbots and voice agents on the processes that weigh most.
AI consulting · Remote, EU
Between what AI can do in a demo and what holds up inside a ten-person company there is a distance no presentation ever describes. The Check-up measures that distance on your processes. Which actions pay off, what they are worth in hours and euros, which ones to leave alone. I work remotely, and the document stays yours even if you decide to stop there.
What I deliver is not a list of things that could be done. The list of actions is ordered by return. Each one carries three numbers beside it: hours freed per week, cost of the tools, estimated impact per month. At the bottom, the items I rejected and the reason the numbers do not add up.
That tail is often worth as much as the head. A company arrives at the call with four or five ideas already in mind. A trade fair, a newsletter, a supplier who proposed something. What is usually missing is a criterion to rank them, so the flashiest one gets switched on. Knowing that two of those ideas cost more than they bring saves you the spend. It also saves the three months it takes to find out on your own.
The discovery call lasts forty-five minutes and contains no demos. The questions circle around what you decide every day, and on what data. There is always one decision you have been postponing for months, because nobody has time to answer it. I then test the hypotheses on the material you actually have on file: quotes, emails, product sheets, meeting notes, whatever is there.
That step knocks out half the ideas, and it is also why a Check-up takes four days and not two hours. A model that answers well on a textbook case can behave in a completely different way on a messy archive. That archive was written by four people over ten years, with abbreviations only they understand.
The review together takes half an hour: we read the report again, I clear up the items that stay opaque, and you choose where to start, or whether to start at all. Nobody calls back three days later to push, and the document stays yours either way.
The document is written so another supplier can read it too. It talks about processes and numbers, not about products to buy from me. If you then hand me the work, we agree the price up front and I run it myself. If you prefer someone else to do it, you hold everything you need to ask for quotes you can compare.
The items in a Check-up almost always land in one of these seven boxes. I list them with the role they usually play in the plan, because the order in which they get switched on matters as much as the choice.
These usually come first, because the return shows up in weeks: workflows, chatbots and voice agents on the processes that weigh most.
These come in when the bottleneck sits before the sale: pages that explain properly and forms that end up where they should.
Precisely because it costs, which is why I weigh it last: you build where no off-the-shelf tool covers your way of working.
Every estimate stays an opinion until the numbers are clean: server-side GA4, margins per product, price monitoring. Plenty of decisions change right here.
The channel many people ask for and that pays off over long timescales: Google plus the answers of ChatGPT, Perplexity and AI Overviews, on a calendar.
Where what is missing is output, generation changes the proportions: reels, carousels, spots and avatars, on an approved storyboard.
The item for those who have already switched several things on: orchestrated agents, a company archive you can question, one dashboard for the owner.
On a Saturday evening a prospect writes, there is nobody in the office, and by Monday morning they have already bought somewhere else. AiDirector exists to close that window: the agents keep watch over leads, deals and content even when the offices are shut.
Each agent has a measurable job and a piece of process of its own. One finds the leads, one qualifies them, one negotiates, one produces content, one puts the numbers together. Above them an orchestrator assigns the goals, and the owner gets one screen with lead flow, funnel, revenue and return per channel.
You can try the dashboard without signing up: the AiDirector demo.
Three families of work, one person to talk to. I build it, no subcontracting.
A document with the possible actions in order of return, each one with the hours freed per week, the cost of the tools and the estimated impact per month. At the end, the rejected items, with the reason the numbers do not add up.
It happens, and I say so without dancing around it. In plenty of companies the bottleneck is a badly designed process or two systems that do not talk, and no language model unblocks that. That is a useful answer too: it saves you the spend and tells you where to look.
Forty-five minutes of call, four days of waiting, half an hour of review together. Within a week you have the numbers in hand and you decide calmly, with nobody calling back to push.
It stays where it is now. For the tests I use samples agreed with you, and if a process touches sensitive information we look at tools that run on European infrastructure or inside your own network. You make that choice with the alternatives in front of you, not after signing.
Yes, and it is written exactly for that. It contains processes, numbers and selection criteria, not product names to buy from me. Anyone who knows the job can build its items; if you pick me it is because it suits you, not because you are stuck.
Yes, and it is often the item with the fastest return. Half a day on the things the team will actually use on Monday morning, with their own documents and their own cases. It is worth more than a generic eight-hour course nobody opens again.
Bring me a decision you have been postponing for months. In forty-five minutes we work out whether AI has anything to do with it, and if it does not, I say so straight away.
Book a call