The cost of AI automation has 2 parts: a one-off build and a monthly cost of running the system. The one-off part depends on the number of processes and integrations; the monthly part on the volume of data processed - and for most small companies that is the smallest line of the whole thing. The first process usually goes live in 2 to 4 weeks.
These are not packages but typical scopes of work. Most companies start with the first or second and expand only once they see the saving.
The simplest entry into AI
The most common first project
Several processes and integrations
Figures are indicative and meant for budgeting. The final quote comes after process mapping, because only then is the real amount of work known.
When comparing quotes, check they cover all of these - the differences usually hide in what is not listed.
Analysis of how the task looks today, where time is lost and where errors occur. This often reveals that part of the steps are unnecessary even before any AI.
The largest share of the one-off cost. Connecting to email, spreadsheets, a CRM or accounting - each integration is a separate job.
The prototype runs against real examples from your company. This phase takes time and is not skipped, because only there do the edge cases appear.
Monthly, by volume of text processed. For most small companies this is a surprisingly small line - often less than a single business software subscription.
The server the automation runs on, the database and backups. A monthly cost that scales with volume.
AI models change, systems get updated and business processes evolve. Optional, but recommended for systems running unattended.
These are the items quotes often assume but do not write down - so the difference only shows when you need them.
Source code, server access and accounts are in your company’s name. Handover is neither charged nor conditional. If you continue with someone else tomorrow, you get everything without argument.
We define which documents the system can access, where data is stored and how long it is kept. For companies working with EU clients GDPR compliance is an obligation, not an option.
Language models change, get cheaper or get retired. We build so the model can be swapped without rewriting the rest, so an update does not cost you a new project.
The quote states a timeline by phase, not one date at the end. Each phase ends with something you see and approve, so you are never three months in without knowing where things stand.
You can do this yourself in five minutes, before calling anyone. You need three numbers:
If the numbers do not work, we will tell you ourselves. We would rather not take a project than build automation that does not pay for itself - those get abandoned after a few months anyway.
You can save by starting with one process instead of five. The saving is measurable immediately, the investment is smaller, and what is learned on the first process makes every subsequent one cheaper.
You can also save on integrations in the first phase. If the system initially writes to a spreadsheet rather than straight into accounting, that is considerably cheaper and you often will not feel the difference.
You should not save on the testing phase against your own data. A system not run through real examples from your company fails on edge cases, and trust is lost after the first wrong quote sent to a client.
Nor should you save on the system admitting when it does not know. Automation that guesses instead of admitting ignorance does more damage than good - and that is the difference a buyer cannot see in a quote, only in daily use.
If you also need a system for the data to live in, we build it as custom software, and field access as a mobile application connected to the same system.
It depends on the number of processes and integrations. A chatbot answering from your documents is the lowest band, automating one business process the middle, and a connected system with several integrations the highest. We give an indicative range on the first call, before any commitment.
There is, made up of the AI model cost by volume of text processed, hosting and optional maintenance. For most small companies the monthly model cost is lower than expected - often less than one business subscription.
If the automated task consumes 40 to 50 hours a month, payback is usually within a year. That is why we count hours first and quote second - if the numbers do not work, we say so immediately.
Because processes differ too much for a single figure to be honest. Automation that reads emails into a spreadsheet and one that integrates with an ERP are not the same job. You get an indicative range straight away and an exact figure after process mapping.
That is exactly what we recommend. One process, a measurable saving, then expansion if it makes sense. Introducing five processes at once is the surest way for none of them to stick.
You do. The code, server access and accounts are in your company’s name. We build so the AI model can be swapped without rewriting the whole solution, so you are tied neither to us nor to one model provider.
It happens in both directions - model prices have mostly fallen in recent years. That is why the system is built so the model can be replaced by another without rewriting everything. It is one of the reasons we do not tie a solution to a single provider.
The first call and a rough estimate are free. Detailed process mapping with step-by-step analysis and a saving calculation is part of the project and is included in the quote.
Describe the process that eats most of your time - you get a saving estimate and an indicative range.
Book a call
