Fixed price after a free demonstration
One manual task less, at a fixed price
A package takes over exactly one recurring task that someone on your team does by hand today. Your spreadsheet and your processes stay as they are; your team only checks what the solution flags.
How a package works
One task
Each package takes over exactly one task that someone on your team does by hand today. No project, no strategy phase.
Your structure stays
We start from your Excel template, your mailbox or your import format. The solution adapts to them, not the other way round.
A person checks
Uncertain or contradictory values are flagged. Your team reviews the flags and decides. The solution does not post entries, make payments or approve anything.
Built securely
Information security is our line of work. Credentials are kept in a vault, data processing is covered by a contract, and we only use processing routes where your data is not used to train AI models.
The packages
Every package follows the same pattern: one input, one output, a person checks. We settle the exact scope in the initial call, based on how you work today.
Receipts and statements into your spreadsheet
Invoices, receipts or statements including attachments are read and transferred into the columns of your spreadsheet or into an import file for your business application. Each row links to the document; fields that are uncertain or whose totals do not add up are highlighted in yellow.
More on AI in bookkeeping →Pre-sorting a shared mailbox
Incoming emails to a shared address are sorted by topic, and a draft reply is ready for each one. Nothing is sent until someone on your team approves the draft.
Matching order, delivery note and invoice
Order, delivery note and invoice are checked against each other. Differences in quantity, price or line item are flagged before the invoice goes for approval.
Questions to your own documents
Manuals, price lists or contracts become searchable. Your team asks a question in plain language and gets an answer with the source, so every answer can be checked.
Meeting minutes
A recording of a meeting becomes draft minutes with decisions, tasks and open points. Recording only takes place with the consent of everyone present.
Pilot · no connection to your production systems
Preparing data from documents for your spreadsheet or business application
Is data from statements and other documents, attachments included, copied by hand into a fixed structure? On a completed case we test AI-assisted preparation as an Excel or import file and measure quality and the remaining effort. For larger cases, where a deployment first has to justify itself, this is the right way in.
Demonstration with synthetic sample data, free of charge
Once your spreadsheet layout or import format is agreed, we demonstrate the data preparation in 30 minutes. We do not need real documents for this, and we do not assume you already run an AI platform.
Pilot on a completed case
We define target fields, rules and criteria with your business team. Your existing result serves as the baseline, and we agree the review standard with the people who know the work.
Decision on next steps
Stop, test further, or a separately quoted deployment. A connection to your business application is assessed and priced separately if needed.
In the decision pilot you receive
- a draft of your spreadsheet or import file with a record of where each value was found
- a comparison at field and row level, including missing and additional entries
- an analysis of errors and the corrections they required
- the measured effort for setup, review and rework, one-off and recurring shown separately
- a decision paper. A well-founded no is also a result.
The result is a file you can check. Review rules flag anomalies, and we also check unflagged values against the baseline. No postings, no write access.
Your data
Processing only after agreed approval
The pilot needs no connection to your production systems. Only the files released for it are processed.
We agree the processing route in advance: external or, where suitable, in an approved local environment.
You receive the information for your own review before you commission us. Confidentiality and data processing are settled before we start; classification, for example under DORA, and approval remain with you.
Where the solution runs: four operating options
The solution itself is the same in all four options. What changes is where it runs and who holds the contract with the model provider. One option can grow into the next, for example from running with us to running in-house.
Run by us
Businesses without in-house IT, getting started
We operate the solution. Nothing needs to be installed on your side. We act as your data processor.
With your own AI account
Organisations with their own contract with a model provider
You choose the provider and region, and model costs go straight through your contract. We store the access key in the solution's vault.
In your infrastructure
Organisations with Microsoft 365, Azure, their own server or their own model
The solution runs in your environment, and data stays in-house or in your own tenant. Permissions, licences and updates are your responsibility.
Local AI in-house →Building your AI platform
Organisations that want their own AI environment
We plan and set up a platform of your own, on which further packages can run later. This is a separate project with its own quote.
Building and running local AI →A requirement in every option: data processing covered by contract, no training on your data, processing within the EU where you require it.
Fixed price after a free demonstration
Initial call
You describe the task: volume per month, time spent today, your template. 30 minutes, no real documents needed yet.
Free demonstration
We show the solution on synthetic sample data in your target structure.
Fixed-price quote
By then we know the scope, the input channel and the review rules, and we quote a fixed price. You choose how it is run from the four options.
Setup, testing and handover
We only process real documents once the necessary agreements and your approval are in place. Your team runs both ways in parallel for a while, then we hand over.
Why Blackfort Technology
We advise on information security and regulatory requirements and bring experience from DORA client projects. That is why we build every package the way we would want to see it in an audit ourselves: credentials in a vault, data processing covered by contract, no training on your data, a person checks before anything is taken over. Managing director Christian Gebhardt is a founding and permanent member of the AI Security expert group of the Alliance for Cyber Security and a co-author of its guide to penetration testing of language models (BSI/ACS, 2025).
Frequently asked questions
Do we need our own AI platform or an IT department?
No. In the first operating option we run the solution and nothing needs to be installed on your side. If you already have a contract with a model provider or an environment of your own, the solution can run there.
Do we have to send real documents for the demonstration?
No. The demonstration uses synthetic sample data in your target structure. We only process real documents once the necessary agreements and your approval are in place.
What does a package cost?
Every package has a fixed price. We quote it after the free demonstration, once volume, target structure and input channel are known. Ongoing operation depends on the operating option you choose and is part of the same quote.
Where is our data processed?
We agree that with you before we start: with us, through your own AI account or in your infrastructure. In every case your data must not be used to train AI models. You receive the information for your own review before you commission us.
What does the solution not do?
It does not post entries, make payments or approve anything. It prepares data, flags anomalies and presents drafts. The decisions stay with you.
Which manual task costs you the most time?
Tell us about the task. In the initial call we work out whether a package fits, and then show it on sample data in your structure.