Knowledge & data
Internal firm knowledge base
How does the deadline-extension process work again? Which template do we use to chase overdue documents? Who covers for Ms Berg? In most firms, the same person answers questions like these – in between two sets of year-end accounts.
You'll get …
- 1
Firm knowledge that isn't stuck in one head
Standards, deadline processes, responsibilities and templates are written down and easy to find – even when someone is on holiday.
- 2
Look it up, don't interrupt
New colleagues find processes and answers, with the source, instead of asking the team lead in between two sets of accounts.
- 3
A system that grows with you
At first the AI only reads, then it writes minutes and suggested changes for sign-off. Client data stays in the specialist system.
01Intro
The firm knowledge base makes standards, responsibilities, templates and processes easy to find – a structured system where the team looks things up instead of asking around. And it does more than a wiki: AI searches it, summarises, files new notes and gradually takes on defined routines – from meeting minutes to the morning deadline overview.
Worth knowing: client data doesn't belong in this system – the knowledge base holds processes and standards, while DATEV and your specialist systems hold the files. I won't promise a "second brain in one go": you start with one clearly defined area of knowledge, otherwise it falls apart on upkeep. And the AI never writes anything unchecked – every change to a standard needs sign-off.
02How it works
- 01
Take stock: where does your firm's knowledge live?
In heads, email threads, Word templates from 2019, notes on the pinboard. Also: which questions interrupt whom most often, and what gets lost when your longest-serving colleague leaves?
- 02
Build the structure – small and documented
A clear filing system with simple written rules: standards, processes, responsibilities, templates. This is also where the line on client data is firmly drawn.
- 03
Fill one area first
Not everything at once: the area that hurts most comes first – often deadline processes or taking on new clients. Existing templates move in curated, not copied.
- 04
Connect the AI: read first, write later
At first the AI only answers questions from the knowledge base – with a reference to the source. Once that works reliably, writing with sign-off follows: minutes, entries, suggested changes to standards. Permissions and limits are set out in a written work instruction.
- 05
Add routines – one at a time
The morning overview of deadlines and follow-ups, the meeting minutes, the reminder list: the AI takes on defined tasks within clear limits – each routine only once the previous one is running smoothly.
03Stages
Every project starts with Tier 1, the team knowledge base – for firms, it's the right place to begin.
Starter: a knowledge base you can ask
One clearly defined area – such as deadline processes and responsibilities – is organised so you can ask it questions. New colleagues look things up instead of interrupting the team lead.
Next step: the AI joins in
Minutes, summaries, suggested changes to standards – the AI writes in the system alongside you, and every change needs sign-off.
Full build: routines run themselves
A morning deadline summary, follow-up reminders, reports – defined routines run on their own within clear limits. The knowledge base becomes a working system. To get there, it needs deep access to your data and processes. That's why Tier 3 only comes after a baseline assessment, once Tier 2 has proved itself for a while, and runs as a project I support throughout.
04Demo
The knowledge base of the fictional “Reiter Tax Firm” – on the left the order in which the firm's knowledge lives (client data deliberately stays out), on the right three everyday moments with it. Click through and watch where it lights up in the tree.
The architecture behind it has two layers: at the bottom the vault – structured folders and notes that belong to the business. Above it the AI layer with documented rules: it reads only at first, then writes with approval, and sensitive areas stay locked. Calendar and tools dock on, routines run within clear limits.
The proof of this model: this website came about this way – I work in such a system myself every day, with AI as an employee inside it.
05Technical implementation
- The knowledge layer belongs to the firm: a Markdown vault (e.g. Obsidian), Notion or a similar system – your content, no vendor lock-in.
- The AI layer is an assistant with access to the vault (e.g. Claude Code as the agentic layer), working to documented rules: which areas, which permissions, which routines.
- A hard boundary around client data: DATEV and your specialist systems remain the authoritative source; the knowledge base covers processes ("how do we do X"), never the contents of client files.
- Permissions in stages: read access first, write access only in defined areas and with sign-off, sensitive areas excluded.
- Upkeep routines are set up from the start – an owner and a review cycle for each area of knowledge, otherwise it goes stale like any wiki.
- Typical content: deadline rules, firm standards, templates, responsibilities and cover arrangements, onboarding processes, meeting minutes.
for context · This shows how something like this usually gets built – not how it has to be built for you. What actually fits is worked out in the project: around your existing setup, your team, and what stays easy to maintain.