Consulting on AI in organisations
AI consulting for organisations: governance, collaboration & transformation
AI is work on the operating system.Not a choice of tools.
AI changes how decisions are taken in your organisation, how people lead and how they work together, often with no project brief behind it. We shape both sides at once: the human operating system of structures, roles and rules of the game, and the machine one of agents, permissions and accountabilities.
Why this is a question of culture
Every AI rollout reaches into your operating system.
AI is not a tool sitting alongside the organisation. It reaches in exactly where behaviour is formed, and so it changes the culture, whether that is intended or not.
Where AI reaches into your organisation
Into processes, because it takes over steps in the work.
Into systems that measure and assess.
Into decision rights, because a proposal that arrives in seconds is harder to ignore than one that arrives after three days.
Into power relations, because it dissolves the information asymmetries that responsibilities rested on.
Culture is the sum of how people actually behave, and behaviour grows out of the conditions people work under. Change the conditions and you change the culture. Treat AI as a pure question of tools and you will still change your organisation, only without steering it.
The difference that shapes how we consult
With people, a condition takes effect through weighing things up. You can discuss it, negotiate it, argue for it, and it still holds when nobody is checking.
With AI, only what is technically enforced has any effect. Telling a model not to give away anything confidential is not a control.
That is why we shape the two sides by different means: the human side through participation, the machine side through architecture. If the two do not fit together, one works against the other.
Where you stand today
AI maturity: three states worth telling apart.
Between "we are experimenting" and "we work differently now" lie several years and fundamentally different states. Each raises different questions and makes different mistakes possible.
The three states
AI works alongside the organisation.
- Individuals use tools on their own initiative
- The quality of results depends on individuals
- Nobody knows which tools are running in the organisation
AI works with the organisation.
- Shared workflows, a common knowledge base, agreed review procedures
- It is laid down who checks which results
- Roles shift, and the shift is negotiated
AI works inside the organisation.
- Agents are clear points of contact, with limited rights and a named owner
- Decisions are sorted by reversibility
- The shape of the organisation follows the work as it is actually done
The key question that makes the difference
Licences, pilot projects and training participants can all be counted. Those figures say nothing about the state of the operating system. The only telling questions are these:
- Has AI shifted your leadership team's working time substantially, away from preparing information and towards judgement and the allocation of resources?
- Have the structural routines changed: the weekly rhythm, the approval chains, the steering meetings?
An AI taking the minutes in your meetings does not count. And the third state is a direction, not a target state with a date on it: no organisation has arrived there yet.
Free and without obligation
Where does your organisation stand with AI?
In a free initial consultation we clarify where you stand and what pace you can sustain. If it turns out that something else needs sorting out first, we will tell you.
What you can afford
Slow or radical: the question of pace decides it.
There are two routes through an AI transformation. Both assume that your organisation will defend the way it currently works. They differ only in how fast they deal with that. Each pace has its own precondition and its own failure mode.
Two routes, one destination
Slowly, inside the existing organisation.
- Change the conditions where the work is done
- Precondition: authority at the top that can push things through
- Failure mode: individual work gets faster, the management layer stays as it is, throughput does not rise
Radically, alongside it.
- Build a small, AI-native unit next to the core and let it prove itself
- Precondition: funding from outside the divisional budgets, backing for running two systems in parallel
- Failure mode: not failure itself, but stopping too early
Three questions that set the pace
Which resource do you have? The power to push things through inside the existing organisation, or money and cover alongside it. Anyone who has neither should start on neither route.
Who decides, and what does their answer cost them? A divisional head rarely commissions a route that ends with their own role gone. We tell the body that could change the shape of the organisation, in one sentence, what that route does not achieve.
How fast is your clock running? There is a cross-check for that, one you can answer yourself.
Is there a high-margin line in your business that two or three people with AI could rebuild in 60 to 90 days?
- The pace of an industry is set by its slowest external constraint: regulation, collective agreements, the investment cycle.
- An attacker does not need your business, only the high-margin slice of it.
- In capital-intensive businesses those are often information processes: spare-part pricing, quotations, service call-outs, claims handling. The protection around your physical core does not reach that far.
What this means for German organisations
This is where the international literature stops. In some businesses the slow route is not enough. And whether the radical one is viable here under employment law, that is, abolishing posts, building new ones in parallel and switching off the old, has not been settled conclusively on the co-determination side. The Anglo-Saxon models assume room for manoeuvre that a company with co-determination does not have.
We do not claim to have a finished answer. We put the question on the table early and involve the works council from the start. Anyone who tells you this is already solved has either not read employment law or is underestimating how fast their competition is.
Our Range of Services
The six fields we work in.
From an honest assessment of where you stand to rebuilding at the edge: six fields that together cover both the human and the machine operating system.
1 · Assessment: where do you actually stand?
- Technically: what can the AI see, what can it do, who is allowed to extend it?
- Politically: who can stop an initiative without being answerable for the outcome?
- Result: an assessment report setting out maturity level, gap and a reasoned choice of direction.
2 · AI-supported collaboration: individual use becomes a shared working environment
- Tested workflows instead of individual prompt craft.
- A shared knowledge base and project memory, with clear rules for maintaining them.
- Written standards that are updated over time, developed on real cases.
3 · Leadership and AI: who decides when the AI already has a proposal?
- Sorting decisions by reversibility: what can be reversed goes quickly, what cannot needs approval.
- Roles shift from passing information on to interpreting it, from gatekeeper to reviewer.
- Addressing the question of power openly: resistance is usually a defence of power, not a disagreement on the substance.
4 · AI governance: what is the AI allowed to do, and who sets the limit?
- Four control components from day one: test runs, logs, rollback, human review.
- Four hard limits: data leakage, identity, recoverability, accountability.
- Tied back to the EU AI Act and NIST. The works council and data protection are part of the thinking from the start.
5 · Organisational knowledge: who owns the knowledge, and where does it grow?
- Capturing knowledge before posts disappear and it walks out with the people.
- Building feedback loops so that the organisation learns from its own use of AI.
- Measurable through the correction rate: is the share that a human has to put right going down?
6 · AI-native unit: when is building anew alongside worth more than improving what already exists?
- A small unit that proves in parallel running that it is better.
- Data access justified process by process; the system of record remains the source of truth.
- The transition for the people involved planned and funded, out of a share of the savings.
Free and without obligation
Where does your organisation stand with AI?
In a free initial consultation we clarify where you stand and what pace you can sustain. If it turns out that something else needs sorting out first, we will tell you.
How we go about it
AI transformation in six steps.
Each step ends with a condition that has to be met before the next one starts. That is why these initiatives do not run into the sand halfway.
The six steps
Assessment. Diagnostics, interviews, an evaluation workshop with the managing directors, IT and the works council. Closes with: an assessment report setting out maturity level and choice of direction.
Target picture. Two to three days with the management board, planned backwards from the target state. Closes with: a target picture the leadership agrees to and stands behind.
Secure the knowledge and cut the work. Making institutional knowledge visible, breaking tasks down, removing ballast. Set agents on a bureaucracy and you get a faster bureaucracy. Closes with: a task inventory with a list of what to cut and a first process.
Settle the approvals. An approvals workshop with IT security, data protection, legal and the co-determination side. Closes with: a documented approval and a permissions framework for each agent.
Build and prove. The new process runs in parallel with the existing one, measured against criteria set in advance. Closes with: proof that allows a decision, plus the old process switched off.
Make it permanent. Putting structure around the new, setting up a steering routine, releasing further waves. Closes with: a new shape for the organisation and ongoing indicators in the leadership's work.
Two things cannot be swapped round: cutting comes before automating, and the approvals come before the build.
Einstiegsangebote
Many organisations want to find out where they stand first, without a major project. For that we have two lean ways in.
AI assessment
Diagnostics and evaluation, no build. Result: an assessment report setting out maturity level, gap and a recommendation for a sensible way in.
A good fit if you are starting a transformation programme or want to check whether a larger AI project is needed.
AI workshop for one team
One workshop day in house on real cases, with an optional follow-up day. Result: tested workflows, a draft knowledge base, written standards.
A good fit if a team already uses AI, but without coordination, and the quality of results varies.
How you can measure it
Results that stay in use.
What is produced over the course of an engagement
- an assessment report setting out maturity level, gap and choice of direction
- a target picture with the agreement of the management board
- a task inventory with a list of what to cut and a first process
- a permissions framework for each agent, with a named owner
- proof from parallel running against criteria set in advance
- written standards for AI-supported work
The honest measure: the share of results that a human has to correct should fall over time. If it stays constant, what has been built is automation with a chat window, not a system that learns. What we do not promise is savings of a given size by a given date. The figures in circulation on that rarely stand up to scrutiny.
We do not advise on something we do not run ourselves
We run our own agentic operating system in day-to-day consulting: our own workflows, a maintained knowledge base, project memories, reviewers that check the results. We have had the same friction there that we support clients through: workflows that were wide of the mark on the first run; knowledge that had to be restructured three times. What did not work is part of what we bring.
Frequently asked questions
We do not have an AI strategy yet. Is this the wrong moment?
On the contrary. An AI strategy written before the assessment describes a desired state with no route to it. First where you stand, then the target picture, then the strategy.
Do we need training for everyone first?
Rarely. The bottleneck is almost never ability, it is the missing shared working environment. We work with small groups on real tasks and let the results spread from there.
What about the works council?
It belongs in the initiative from the start. Monitoring performance and conduct is subject to co-determination, and the permissions given to agents touch on that question. An agreement that is jointly supported is more likely to be followed day to day than a policy handed down.
How long does it take?
Assessment: a few weeks. Team workshop: one day plus a follow-up. Proof for one process: about three months. A restructuring that the organisation carries runs over years.
We introduced a tool and it is barely used. What now?
The most common starting point. The cause is almost never usability, it is a missing benefit for the people involved, or an approval route that prevents use.
Do we have to commit to one provider?
Not for our sake, we work provider-neutral. For you the rule is: at the points that matter, the components must remain interchangeable, so that your ability to act does not hang on one vendor's development path.
What happens to our middle managers?
The machine takes over passing information on. What is left is more demanding: resolving ambiguity, shaping the exceptions, taking responsibility for results. We plan and fund the transition out of a share of the savings.
What do you do differently from an IT consultancy?
We address the side on which most AI initiatives fail: responsibilities, decision-making processes, power relations, accountability. The systems themselves are built, where needed, by a partner or by your own IT.
In brief
AI changes your operating system either way. The only question is whether that happens deliberately and under control.
We shape both sides at once: the human side through participation, the machine side through architecture.
The question of pace decides it: slowly inside the existing organisation or radically alongside it, each with its own precondition and its own failure mode.
We deliver artefacts rather than promises and make the effect measurable through the correction rate.
Free and without obligation
Let us find out together where you stand.
In a free initial consultation we clarify where you stand and what pace you can sustain. If it turns out that something else needs sorting out first, we will tell you.