Is your company ready for AI? An honest AI readiness assessment before you build

When AI projects stall, it is rarely because the technology does not work. They stall because the company built on a foundation that could not take the weight: data scattered across ten systems, processes nobody has written down, or an organisation that never decided who owns the result. Before you build anything, it is worth asking the uncomfortable question – are we actually ready? Here is what "ready for AI" means in practice, and how an AI readiness assessment shows where you stand.

What does it mean to be ready for AI?

AI readiness measures how well prepared an organisation is to introduce AI in a way that delivers real results – not just a demo that impresses in a meeting. It depends less on which model you choose than on four layers beneath the surface.

• Data. Does the information the AI needs to work with exist, and can you get at it? Is it structured, or scattered across emails, PDFs and people's heads?
• Processes. Are the workflows clear enough to be described – and therefore automated? If nobody can draw the workflow, nobody can build an agent for it.
• Skills. Is there someone in-house who can own the tool, ask the right questions and recognise when an answer is wrong? If not, you are relying entirely on outside help – and then it matters a great deal that you choose a supplier you trust, one that takes the time to understand your business.
• Governance. Who is responsible for what the AI may and may not do? How do you handle sensitive data, wrong answers and questions of liability?

The first two layers decide what you can build. The last two decide whether it holds up once it is live. A company can be mature in one layer and immature in another – that is perfectly normal. The point of an assessment is to find out which layer is the real bottleneck before you spend money on the wrong thing.

Five signs that you are not ready yet

Some signs make it clear that the groundwork has to come before the tool. If several of them apply, that is not a red light – just a pointer to where to start.

• Nobody can answer "which problem are we solving?" AI becomes an answer in search of a question. Without a concrete problem, the project turns into a technology exercise.
• Data is scattered and unstructured. If nobody knows where the correct version of a piece of information lives, the AI will guess as badly as a new employee who never got any onboarding.
• The process only exists in people's heads. If the workflow cannot be described step by step, it cannot be automated reliably.
• Nobody owns the initiative. When "the AI project" is everyone's and no one's responsibility, it fizzles out after the demo.
• AI is expected to replace judgement, not support it. AI is at its strongest backing up a person and at its weakest as an unmanned autopilot for decisions that require accountability.

Which processes can be automated with AI?

The workflows that suit AI best have one thing in common: they are repetitive, text-based and follow logic that can be described. Below are generic examples by type of work – treat them as patterns to look for in your own business, not ready-made solutions.

• Internal knowledge search. When the answers sit in documents, policies and past cases, a RAG-based system lets employees ask questions in plain language instead of searching by hand.
• Repetitive text production. Draft replies, summaries of long documents, first versions of reports – with a person reviewing and approving.
• Classification and routing. Sorting incoming cases, emails or forms and sending them to the right person or queue.
• Data extraction and compilation. Pulling relevant details out of unstructured documents and organising them in a structured format.
• Multi-step workflows. AI agents chain several steps together – fetch, interpret, draft, hand over – with a person as the checkpoint.

It is just as important to know when AI is not the right tool. If the task requires legal or medical accountability, comes up only a handful of times a year, or can be solved with a simple rule, automation is often more expensive and more fragile than the problem it was meant to solve. An adviser worth their fee says no in those situations rather than yes to everything.

How to map roles and workflows before you build

The step that is most often skipped is also the one that matters most: mapping. Before anyone writes a single prompt, you need to know how the work is actually done today – not how a process document from 2019 describes it.

A good mapping answers three questions: who does what, which workflows recur often enough to be worth automating, and where the time really goes. It often shows that the real gain lies somewhere other than where you expected. This is also where you find out whether the data needs cleaning up first, or whether a workflow is too ad hoc to automate reliably.

This is the core of our AI consulting service: mapping roles, processes and data before anything is built, so the investment goes on what actually makes a difference. It is the same logic as a well-thought-out website strategy coming before the design – making the right decisions in the right order saves more than any tool will later.

From workshop to a working agent flow

A workshop is not the goal – it is where the map starts. A common approach is to go through the business's workflows together, identify a couple of high-impact, low-risk candidates and prioritise them against what can actually be built with the data you have.

From there, the route to a working flow follows clear steps: a limited pilot on a single workflow, testing against real cases, and only then a wider roll-out with governance and follow-up in place. What separates a demo from a workflow in production is exactly that last part – error handling, permissions, logging and a person who owns the result. It is also where most promising-looking projects fall down, if nobody planned for it from the start.

For more advanced automations and RAG systems, we work with our sister brand AI Partner, which specialises in building AI agents and automations that hold up in production – not just in a demo.

New transparency requirements: the AI Act has applied to your chatbot since August 2026

The transparency requirements in Article 50 of the EU AI Act have applied since 2 August 2026. If your website has a customer service chatbot, visitors must be told that they are talking to AI – clearly, and from the very first interaction, according to the European Commission's guidance. A line in the terms of use is not enough; the label has to be in the chat interface itself. The exception – when it is obvious to a reasonably well-informed person that they are dealing with AI – is interpreted narrowly. The rules apply as soon as your users are in the EU, wherever the company or the chatbot supplier is based.

For AI-generated content, the rules are more nuanced than they are often made out to be. Genuinely synthetic images, video and audio must carry machine-readable labelling under Article 50, and that responsibility lies primarily with the provider of the AI system. Ordinary AI-assisted text that a person fact-checks and takes editorial responsibility for is, by contrast, not normally covered by the labelling requirement. Breaches can lead to fines of up to €15 million or 3 per cent of global turnover, according to Article 99.

In practice: label the chatbot in the interface, and document which AI systems you use and who is responsible for them. That is the governance layer from the readiness assessment above – and one of the easiest requirements to meet if you build it in from the start.

Why the combination matters more than the model

There are plenty of firms that can set up an AI tool, and plenty of consultants who can talk strategy. What makes the biggest difference is the combination: the business sense to know what is worth building, and the technical grounding to build it so that it holds up in production. Strategy alone never produces a working flow; technology alone builds impressive things that nobody needs.

What does it cost – and what do you get?

The cost depends entirely on what the mapping shows. A limited pilot on one clearly defined workflow is a completely different investment from a broad roll-out with several integrations and governance in place.

We start with an assessment of your readiness and a concrete recommendation: what is worth building, in what order, and what it involves. Sometimes the answer is that you are ready to start straight away. Sometimes we recommend spending time structuring the data first. Book a call and we will look at your specific workflows and give you a concrete quote – no off-the-shelf prices for a problem we have not seen yet.

Get in touch and we will set up a no-obligation conversation about where you stand and what your next step should be.