Fourth pillar

AI that actually works in your business

An AI pilot that gets stuck in demos delivers nothing. Newblend starts with the process: which step takes too much time, which systems are involved, and what an agent may decide on its own. Then we build it, connect it to your existing systems and train your colleagues.

Request the AI maturity scan View the approach

Choose a process

In short

Four services: strategy & advice, agents & automation, Microsoft Copilot, adoption & training.
A Newblend agent works within your own Microsoft tenant or an environment we manage for you. We only use business licences under which the supplier contractually commits not to use your data to train models. For each application we record which data is processed where.
Every application is reviewed against the GDPR and the AI Act before anything is built.
We start with one defined process with a clear owner, not with a platform choice.
We do adoption and training together with Spike Academy, specialised in training and e-learning.

Last updated: 10 November 2026

Four services

They stand on their own, but in practice often follow each other: first find where the value is, then build, then make sure it gets used.

Strategy & advice

Where AI adds value and where it creates risk. Including a GDPR and AI Act review and priorities per process.

Agents & automation

Agents that run through your processes and are connected to ERP, CRM, HR and Microsoft 365.

Microsoft Copilot

Setup, permissions model and data classification. Copilot sees everything a user is allowed to see, so cleaning up permissions comes first.

Adoption & training

Technology nobody uses is money wasted. Training and guidance together with Spike Academy.

How a Newblend agent works

Trigger, decision, integration, action, feedback

An AI agent is software that carries out a task on its own: it receives an instruction, consults systems, makes a decision within predefined rules and performs an action — without a person directing every step.

01

Trigger

An email, form or event in a system.
02

Decision

The agent checks against the rules you have defined.
03

System integration

Retrieving data from ERP, CRM, HR or Microsoft 365.
04

Action

Create a ticket, draft a quote, send an email.
05

Feedback

Confirmation to the user and logging for audit.

From analysis to production

The programme in four steps

The same approach as our IT projects: first record what is there, then give reasoned advice, then build.

STEP 01

Process analysis

We go through the processes with the people who do the work. Where is the repetitive manual work, where does it go wrong, which systems are involved.
STEP 02

Business case and review

You receive a reasoned proposal: expected time savings, required integrations, costs, and the outcome of the GDPR and AI Act review.
STEP 03

Build and pilot

We build the agent, connect it to your systems and first let it run alongside existing work. We adjust the decision rules based on real cases.
STEP 04

Production and management

Live, with an owner within your organisation, logging for audit and management under the same SLA as the rest of your IT.

Work it out yourself

What does automation deliver?

Enter your hourly rate and the time a recurring process takes now. The result is an indication, not a quote.

Hourly rate

Cost per hour per employee, including employer costs.

€ 50€ 275

Manual work in this process

1 hour40 hours

Per month

Per year

Time freed up

per year

Paid back in

Assumptions: an agent takes over 80 percent of the manual work, 46 working weeks per year, and a one-off setup of around 40 hours. What it will be in your situation, we determine after an assessment.

Have it calculated for your process

Where you stand now

The three AI maturity levels

Newblend places organisations in three levels. The AI scan determines where you stand and what the next sensible step is. The scan consists of seven questions.

Table. AI maturity levels according to Newblend.
Level	What you recognise	Next step
1. Exploring	Colleagues use separate AI tools on their own initiative. No policy, no visibility of what data leaves.	Set up an AI usage policy and a secure work environment.
2. Applying	AI is in the standard tools, for example Microsoft Copilot. Usage varies strongly per department.	Steer adoption and build the first agent on a defined process.
3. Automated	Agents run in production, connected to business systems, with logging and ownership.	Scale up to more processes and formalise governance.

Request the AI maturity scan Book a conversation

What we build agents for

Processes with a lot of repetitive manual work, fixed rules and a clear owner are most suitable. These are the requests we get most often.

Incoming post and email

Classify documents, extract data and put it in the right system. When in doubt, pass to a person.

Quotes and requests

Turn a request into a draft quote based on your price list and previous orders. The salesperson checks and sends it.

Servicedesk and tickets

Categorise requests, merge duplicate tickets and handle standard questions directly with an answer from your own documentation.

On- and offboarding

Arrange accounts, licences, permissions and equipment when people join or leave, with an audit trail for the auditor.

Reports

Compile recurring reports from multiple sources, with an explanation of what has changed.

Making knowledge findable

Search your own contracts, procedures and project files, with sources cited so the answer can be checked.

Preconditions

Secure and explainable

AI touches business data, permissions and decision-making. That is why we arrange the preconditions before anything goes into production.

Your data stays yours

A Newblend agent works within your own Microsoft tenant or an environment we manage for you. We only use business licences under which the supplier contractually commits not to use your data to train models. For each application we record which data is processed where.

GDPR and AI Act

For each application we determine the legal basis, the retention period and the risk class under the AI Act.

Clean up permissions first

Copilot and agents see what a user is allowed to see. Before rollout we put the permissions model in SharePoint and Entra ID in order. Read more about ISO 27001 and access management.

Logging and ownership

Every step of an agent is logged. Every agent has an owner within your organisation who manages the decision rules.

No users, no results

Training and education

We specialise in training and educating your employees, and take care of the adoption side of every AI programme.

An AI usage policy people understand, not just a document for the auditor.
Practical training per role, with examples from your own work.
Guidance for key users who help their colleagues further.
Measuring whether usage actually rises, and adjusting where it lags behind.

Management is part of it

An agent in production is IT: it needs monitoring, updates and a point of contact. We include agents in the same SLA agreements as the rest of your environment. See the blends and SLA.

Legislation

What the EU AI Act asks of you

The AI Act has been in force since 1 August 2024 and applies in phases. For most SMEs the role is that of deployer: you use AI, you do not build or sell it yourself. That is a lighter regime than for providers, but not optional — and it also applies to AI already built into your standard software.

It starts with knowing what you have in-house. In practice colleagues often use AI tools before there is a policy. An overview of which AI is used where, with an owner per application, is the first step towards compliance.

The phases

2 FEBRUARY 2025

Prohibited practices and AI literacy

Certain applications are not permitted, and employees who work with AI must have sufficient basic knowledge. Supervision and enforcement have applied since August 2026.
2 AUGUST 2025

General-purpose AI models

Obligations for providers of general-purpose AI models. Mainly relevant to you as a supplier requirement: ask about it when purchasing.
2 AUGUST 2026 — NOW IN EFFECT

Transparency obligations

A chatbot must identify itself as AI, AI-generated media must be machine-readably marked and anyone using emotion recognition must inform people.
2027 AND 2028

High-risk applications

Postponed via the Digital Omnibus to 2 December 2027 for standalone systems and 2 August 2028 for AI in regulated products.

What this means in practice for SMEs

Make sure your people are AI literate

Anyone working with AI must know what it can and cannot do and which data must not go into it. A short training per role plus a usage policy people understand is enough for most organisations.

Be open about AI towards customers

If you have a chatbot on your website or use AI for customer communication, that must be made known. Also check who manages the settings for it: you or your supplier.

Watch out for applications involving staff

AI for recruitment, appraisal, credit scoring or biometrics falls into the high-risk class. The deadlines have moved, but contracts you sign now run straight through them. Emotion recognition in the workplace is not permitted at all.

Record what you use

One overview of which AI is in which process, who the owner is and what data goes in and out. You need that same overview for the GDPR and for customers assessing your supply chain.

Fines

Breaching the transparency obligations carries fines of up to 15 million euros or 3 percent of global annual turnover. For prohibited practices that rises to 35 million euros or 7 percent.

What we do

Inventory which AI is used in your organisation, including tools colleagues have started using themselves.
Determine the risk class and role per application: deployer or provider.
Draw up an AI usage policy and arrange AI literacy, together with Spike Academy.
Set up technical measures: permissions, logging, data classification and, where needed, marking of AI output.
Record the outcomes in the same file as your ISO 27001 and NIS2 documentation.

We are not lawyers: we set up the technology and the records, and work with your own adviser on legal questions.

Assess your applications in the AI scan See also NIS2 and ISO 27001

Frequently asked questions about AI

What does it cost to have an AI agent built?

We usually build a first working agent for one defined process within a few weeks. The investment depends on the number of system integrations and the complexity of the decision rules. We start with a fixed price for the analysis, so the business case is clear before anything is built.

Does our data stay ours?

Yes. A Newblend agent works within your own Microsoft tenant or an environment we manage for you. We only use business licences under which the supplier contractually commits not to use your data to train models. For each application we record which data is processed where.

What do we need to watch out for under the AI Act?

The AI Act works with risk classes and phased application dates. Since 2 August 2026 the transparency obligations apply: a chatbot must identify itself as AI and AI-generated media must be recognisable. The AI literacy obligation for employees has applied since February 2025. For high-risk applications such as recruitment or credit scoring, the requirements have been postponed via the Digital Omnibus to 2 December 2027 and 2 August 2028. In the AI scan we assess which class each application falls into.

Where do you start if you are not doing anything with AI yet?

With a usage policy and a secure work environment, so colleagues do not put business data into external tools on their own initiative. Then choose one process with a lot of repetitive manual work and a clear owner. That makes the best first agent.

What if the agent makes a mistake?

Every agent works within predefined rules and logs every step. You decide which actions the agent may perform on its own and which go past a person first. When in doubt, the agent escalates instead of deciding.

Do we need Copilot to get started?

No. Copilot is a good first step if you already work in Microsoft 365, but an agent on a specific process is separate from that. Which route delivers most depends on your maturity level.

Start with one process

Take the scan or book a conversation. Together we look at which process is most suitable and what it delivers.

Request the AI scan Book an introduction