40%+
of agentic AI projects will be cancelled by the end of 2027, due to escalating costs, unclear value or inadequate risk controls.
Source: Gartner, 2025
OpenAI Accredited Partner
Service · AI MANAGED SERVICES
AI Managed Services: AI backup, always on
AI changes every month. New models, new tools, new prices, new rules. We take on the ongoing management of your AI adoption and deployment: from monthly AI updates for your people to cost control, data governance and the agentic harness around your agents. Your people build. We make sure it stays manageable, scalable and reliable.
- Technology-agnostic
- Claude, ChatGPT, Copilot, Gemini and open-weight
- One point of contact
- Monthly rhythm
Why now
Go-live is only the beginning
Most organisations are past their first experiments. People work with copilots and the first agents are running. Then the real work starts: keeping tools configured, watching costs, knowing where data goes, keeping up with models and letting agents grow safely. That work never ends, because the technology never stands still for a month.
1 in 5
companies has a mature model for governing autonomous AI agents.
Source: Deloitte, State of AI in the Enterprise 2026
21%
of organisations keep an up-to-date register of the agents they run.
Source: Cloud Security Alliance, 2026
The model
You build, we safeguard
With you: the builder mentality
Your people see the opportunities in their own work. They come up with, design and build their own skills, routines and agents. We train them for that, and we keep doing so.
With us: the foundation underneath
We make sure what you build is manageable, scalable and reliable. With the agentic harness around it: register, passports, guardrails, evaluations, monitoring, costs and compliance.
What we manage
Six disciplines, one point of contact
01
Monthly AI update
What is happening in AI and what does it mean for you? Every month a tailored update for each audience. The leadership team gets the strategic picture, the business gets new applications and the technical teams get models, frameworks and architecture.
- Update sessions on site or online
- Webinars and short explainers
- Internal AI podcast
- Briefing on changes in your own tools
02
AI tool setup and adoption
We manage the accounts and workspaces of your AI tools and configure them properly: permissions, connectors, skills, projects and shared knowledge. We track usage and see where adoption lags.
- Workspaces in Claude, ChatGPT, Copilot and Gemini
- Usage monitoring and adoption dashboard
- Licence management per role
- Advice on alternatives as the market shifts
03
AI costs and token optimisation
AI spending grows fast and is hard to predict. We make inference costs visible and adjust course, so costs grow in line with value.
- Cost reporting per team, tool and agent
- Licence or API: which is better value
- Model routing, caching and budgets with alerts
- Token-aware prompts and context design
04
Data governance, GDPR and compliance
Where does your data go, who can access it and under what terms? We keep the overview and align with the GDPR, your security policy and the EU AI Act.
- Data flows and processors mapped
- AI policy and usage agreements kept current
- Risk classes under the EU AI Act
- Alerts when supplier terms change
05
Agentic deployment and the harness
Agents in production need a harness: frameworks, control and evaluation. We manage the environment your agents run in and keep them sharp.
- Agentic register with owner, systems and status
- Agent passport for every agent
- Guardrails, human in the loop and audit trail
- Evals, monitoring and context engineering
06
Process redesign and roadmap
Every quarter we look ahead. Which processes are ready for the next step, where does a decision model or agent deliver the most, and what does a process look like when you redesign it around AI?
- Quarterly review with the leadership team
- Impact versus effort per department
- From workflow to autonomy, step by step
- Use of decision models
The right brains
The right brain for every task
New models appear every month, from the US, China and Europe. We track which brain works best for what and test new models on your own tasks before they go into production. Choosing a model is a trade-off between accuracy, cost and speed, and we make it with evals.
Sprinter
Fast and affordable. For classifying, summarising and routing at high volume.
Workhorse
The workhorse for most day-to-day work and most agentic workflows.
Einstein
Deep reasoning for complex analysis, planning and decisions where it really counts.
Decision models: specialised models for recurring decisions, trained on your own data and outcomes.
View decision modelsFrom workflow to autonomy
Autonomy grows in steps
Agents are evolving from fixed workflows towards more and more independence, and that evolution is still well under way. We only let autonomy increase once control is in place.
- 1
Assistant
AI helps an employee with a task.
- 2
Workflow
AI carries out fixed steps in a process.
- 3
Agent with a human at the helm
The agent proposes, a human approves.
- 4
Supervised agent
The agent acts on its own, people check by sampling and afterwards.
- 5
Autonomous within limits
Agents direct sub-agents within defined boundaries.
How it works
A steady rhythm
Start
Baseline and inventory: which tools, agents, costs and data flows are there? The result is a first agentic register and a list of priorities.
Every month
AI update for each audience, reporting on usage and costs, a model check and adjustments to tools and agents.
Every quarter
Strategic review with the leadership team: what has it delivered, what is the next step and which processes do we redesign.
Ongoing
One point of contact at The Automation Group and support with questions and incidents.
Who it is for
For organisations past the first step
For mid-market and enterprise organisations that already use AI and now want to organise it properly. We work with everyone who has a role in it.
Board and leadership team
direction, urgency and results
CIO, CTO and IT
architecture, operations and security
CFO and finance
control over AI costs
Privacy, legal and compliance
GDPR and EU AI Act
HR and L&D
skills and adoption
Delivered by our agentic consultants with a 10x mindset, under the architecture of CTO Stefan van der Leeden.
Frequently asked questions
Questions we often hear
Which AI tools do our people actually use, and where does our data go?+
We start with an inventory of all tools, licences and data flows, including usage outside IT's view. We then keep that overview up to date every month.
Why are our AI costs rising faster than the value?+
Often because of duplicate licences, the wrong model for the task or agents sending more context than needed. We make costs visible per team and per agent and adjust with model choice, caching and budgets.
How do we get pilots into production?+
Pilots usually stall on ownership, integration and the question of whether they are safe enough. With a register, agent passports, evals and clear agreements on human oversight, we make the move to production responsibly.
Which agents are running, who owns them and what are they allowed to do?+
That is in the agentic register: for each agent the owner, systems, models, permissions and latest update. Every agent also gets an agent passport with its risk class under the EU AI Act.
How do we keep up with new models without breaking anything?+
We first test new models on your own tasks and compare them on accuracy, cost and speed. Only when a model demonstrably performs better do we put it to work.
How do you handle the GDPR and the EU AI Act?+
We map which data goes to which supplier, under what terms and where it is processed, and we classify applications under the EU AI Act. That way you know where you stand and what still needs to be done.
Won't this make us dependent on you?+
The goal is for your organisation to become stronger itself. Your people learn to build, everything we set up is documented and the register and passports belong to you.
Which tools and platforms do you work with?+
We are technology-agnostic. We have a lot of experience with the Claude stack and work just as well with ChatGPT, Microsoft Copilot and Foundry, Google Gemini, AWS, Google Cloud and open-weight models.
What does the commercial arrangement look like?+
An ongoing engagement with a fixed monthly fee, matched to the size of your organisation and the number of tools and agents. You pay licences, tokens and hosting directly to the supplier.
Continue reading
Discover more about our approach and solutions.
AI backup, always on
Curious what AI Managed Services could mean for your organisation? We start with an introduction and a short inventory of where you stand today.