Foundation first, tooling second

Most AI projects do not fail on technology. They fail because company knowledge is scattered, outdated or held only in people's heads, and because nobody can say who is allowed to see what. The AI Knowledge Audit settles both before you buy tools. The result is a document your management can act on.

23 September 20266 min
2–3 consulting days3 weeks to readoutUseful even without follow-up work
Illustration: three data sources pass through a shared approval into one document.

Why this has a deadline

According to the German Federal Statistical Office, around 13.3 million people in employment will reach statutory retirement age within the next 15 years. In many companies the most valuable knowledge is not documented but tied to a few long-serving employees. Unlike most business risks, this one has a date attached.

At the same time, research by McKinsey and IDC puts the share of working time spent searching for information at roughly one fifth of the week. Assistive systems are meant to solve exactly that. They can only find what is stored discoverably and permissioned cleanly.

That is why rollouts are often delayed by access rights rather than technology: an assistant shows every employee precisely what they could already open, making years of overly broad sharing visible all at once.

Weekly AI live calls are now embedded across the site.

Every Saturday at 11:00 Europe/Berlin, the format gives a compact mix of market filtering, practical cases, questions, and clear next steps.

Saturday, September 26, 2026 at 11:00 · Europe/Berlin1x per weekLive Q&A

Next session: Saturday, September 26, 2026 at 11:00 · Europe/Berlin. The series then continues on a weekly rhythm.

Three questions the audit answers

These are the questions rarely answered online, because they require knowledge of your operation rather than of a product.

Who holds which knowledge?

Which business-critical knowledge exists only in people's heads, with time horizon until departure and the damage its loss would cause.

Who may see what?

Current and intended state of access rights, including the permissions that today reach further than anyone assumes.

What stays current?

Which knowledge is stable enough to be written down, and which must always come from the system of record.

Three weeks, four steps

Two to three consulting days spread across three weeks from kickoff to readout.

1

Kickoff

Objectives, scope, contacts and data protection framework. Half a day.

2

Interviews

Four to six conversations of 45 minutes: management, two or three knowledge holders and one operational successor.

3

Inventory

Existing documentation, storage locations, systems and permissions. Where does knowledge exist only in someone's head?

4

Analysis and readout

Evaluation, a written document and a 90-minute presentation to management.

The readout document

15 to 25 pages, written for a decision rather than as a technical report.

  • Knowledge risk map: who holds which critical knowledge, with time horizon and impact assessment
  • The five processes with the best ratio of knowledge risk to how well they can be secured
  • Rights matrix: who may receive which information, and where today's storage contradicts that rule
  • Separation by volatility: what gets written down and what must come live from the system of record
  • Implementation roadmap with sequence, effort range and prerequisites
  • Make-or-buy assessment per process: in-house, standard software or guided introduction

Two sequences

The difference between a system still used after six months and one nobody opens again.

Tooling first

  • Licences are bought before the sources are clarified
  • The assistant answers from outdated documents
  • Access rights surface only in live operation
  • Nobody owns maintenance, and trust erodes

Foundation first

  • Sources, rights and ownership are settled before purchase
  • Every answer has a named system of record
  • Permissions are part of acceptance, not an afterthought
  • Maintenance is agreed before the system goes live

Frequently asked

Do we need an AI strategy first?

No. The audit is the basis for that decision. The result remains useful even if you introduce no AI at all, because the knowledge risks and rights questions exist regardless.

Are you selling us your own software?

No. Where a tool is needed we use existing solutions and pass their cost through unchanged. Our work is the order beforehand and the maintenance afterwards.

What happens to our documents?

The inventory needs structure and storage locations, not confidential originals. The data protection framework is agreed at kickoff and processing follows it.

What does the audit cost?

It is a fixed-price engagement, depending on size and the number of areas included. We state the exact price in the first conversation, together with scope and acceptance criteria.

Describe an area, receive an assessment

Describe one area anonymously: which questions keep coming back, who answers them today, and what happens when that person is away? Please do not send customer data or credentials.

WhatsApp AI assistant

Start potential analysis

If you want to prioritize a real process, a few clear inputs are enough for a strong first assessment.

WhatsApp AI assistant