Adopting AI in your company: the complete guide

Almost every company is trying something with AI today - according to McKinsey, 88% use it in at least one function. But using it and making money on it are two different things: an MIT study found that 95% of organisations see no measurable return from AI. The difference isn't the tool, but what happens around it - because by the BCG 10-20-70 rule about 70% of the value comes from people and processes, and only 10% from the algorithms themselves.

This guide walks through the whole adoption, in order: from deciding where to start, through cost and measuring return, to the most common causes of failure, data security and the law. We summarise each part here and link to the full post, where we break the topic down - with data and cited sources.

What is AI adoption, really?

It's a change in how you work, not buying a tool. It has five stages: building the team's awareness, mapping everyday tasks, choosing and wiring in solutions, training people, and measuring the effects. Companies that "buy AI" and bolt it onto the old flow usually see no effect; those that rebuild a concrete process around it do. Read the full post: what AI adoption is →

Where do you start?

With one process, not a tool. A good first process is repetitive, has a measurable cost and doesn't require sorting out the whole company - usually it's working with text, not analysing all your data. A small step with a "before" and "after" number gives you the proof to build the next one on. Read the full post: where to start →

What does it cost?

The cost isn't the price of a subscription, but the sum of three layers: tools, fitting to your processes, and training people. Tools are the smallest item today; most of the value - and the cost - sits in the fit. So instead of asking "what does it cost", ask "how small a step can I start with". Read the full post: how much it costs →

How do you measure the return (ROI)?

The return usually doesn't vanish - it just isn't measured. The key is to set a baseline before you start, measure business outcome rather than activity, and count over the right horizon. Fewer than one in five companies track clear KPIs for AI, and those correlate most strongly with the bottom line. Read the full post: how to measure ROI →

Why do AI rollouts fail?

Almost never because of the technology - the goal, the data and an unchanged process fail. A pilot is not a rollout: most projects collapse on the road from an impressive demo to production. Rollouts led with an experienced partner reach production twice as often as those built alone. Read the full post: why rollouts fail →

Data security and Shadow AI

Your people are most likely already using AI - the only question is whether with rules or without. Free ChatGPT trains on what you type by default; business versions don't. The answer isn't a ban (it pushes use into the shadows), but the right tool, a policy and a trained team. Read the full post: is company data safe in ChatGPT →

The AI Act and the law

The EU AI Act most likely applies to your company, but usually as a user of AI, not a provider - and that means lighter obligations. Some rules already apply (bans and the AI literacy duty), and the deadline for high-risk systems was pushed back. Your obligations depend on the use, not the tool. Read the full post: does the AI Act apply to me →

AI software - when to build your own?

When the process is your edge and no off-the-shelf tool fits without compromise. AI shortens the build, so a tailored app is cheaper than a year ago - but only if you start from a specification, not from "let's see what comes out". Read the full post: why the spec matters →

How we help

We lead companies through this whole path - we don't leave slides, we leave working processes. We work across four areas:

• AI adoption - process analysis, team training, advice on where AI adds the most value.
• Software - apps tailored to your processes, without per-seat licences or features you never use.
• Joint-venture products - we build our own tools where there's a niche and potential.
• Content and materials - photo, video, graphics, print; we reach for AI only when it genuinely speeds things up.

The goal is always the same: a company that stands on its own after the project, because the know-how and data stay with you.

Frequently asked questions

Where do I start adopting AI in a company? With one process and a measurable goal, not a tool or a big project. Pick a repetitive task, measure what it costs today, switch on AI and check the result after a few weeks.

How much does adopting AI cost? The cost is the sum of three things: tools, fitting AI to your processes, and training the team. Model subscriptions are the smallest item today - most of the value comes from wiring AI into how you work.

Is using AI in a company safe? Yes, if you use a version that doesn't train on your data and you have a policy on what may be typed into it. Most of the risk comes from Shadow AI - using tools without rules, not from the technology itself.

Key takeaways

• AI is a process, not a purchase - about 70% of the value comes from people and processes, not the algorithm.
• Start with one process and measurement - a small step with a "before" and "after" number gives you proof to scale.
• Count the return, not the tool's price - and measure business outcome, not activity.
• Security is tool + policy + training - not a ban.
• The odds improve with a partner - rollouts led with someone experienced reach production twice as often.