Generative AI for business: whether to invest, where it pays and what has to change

Most have started. In spring 2025, 38 per cent of Finnish enterprises used AI technologies, 14 percentage points more than a year earlier, and among enterprises with at least 100 employees the share was 68 per cent (Statistics Finland, 2025 (opens in a new tab)). Across the EU the share was 20 per cent, with Finland second only to Denmark; 55 per cent of large enterprises and 30 per cent of medium-sized ones used AI (Eurostat, 2025 (opens in a new tab)). Almost half of Finnish SMEs invested in software, information systems or AI in the past year (Suomen Yrittäjät, 2026 (opens in a new tab)).

The return has not kept pace with the use. In McKinsey's 2026 global survey, nearly nine in ten respondents said their organisations use AI regularly in at least one business function, yet only 37 per cent attribute any effect on operating profit (EBIT) to it. About 6 per cent attribute 5 per cent or more of their EBIT to AI, and nearly three in four of that group have redesigned the work around it, against about one in four of the others (McKinsey, 2026 (opens in a new tab)).

For a leadership team, whether to use generative AI is mostly settled. The open questions are where it will pay, how much to put in, and what has to change in the company to collect the return.

Two questions decide it. The first is exposure, meaning how much of your business generative AI can change: a company whose work is mostly documents, customer conversations, analysis and code has more to gain than one whose value sits in physical production. The second is how ready the company is to capture the gain, which means data it can use, systems that can connect, people who can judge the output, and rules for what is allowed.

High exposure and high readiness point to acting now. High exposure with low readiness calls for preparation: data, skills and rules first, then spending on scale. Where exposure is low, a few targeted uses are usually enough, and waiting for the tools to mature costs little. Exhibit 1 sets out the decisions that follow, in the order a leadership team meets them.

It pays best where work is frequent, mostly text and measurable, and where an error can be caught before it costs much: drafting and checking documents, answering routine questions from customers and suppliers, summarising long material, writing first versions of code, and preparing sales and service conversations. It pays least where volumes are low, where the process itself is unclear, or where nobody is qualified to check the output.

The test works function by function, because the uses that pay in finance are not the ones that pay in sales. For sales and marketing, our earlier articles on generative AI in growth, marketing and sales and on personalising go-to-market messages at scale go into the detail.

Less on the model than on three things the business controls: how many people use it in their daily work, how much work it removes, and what the change costs in training, process redesign and management time. Business cases often count the licences and the hours saved, and leave out adoption and the cost of change, yet those two decide whether the saved hours turn into money.

A sound case states its assumptions, shows the payback at low and at high adoption, and names who will release the saved time and what that time will be used for.

The work itself. In the McKinsey survey, the companies with the clearest profit effect were the ones that redesigned the work around AI (McKinsey, 2026 (opens in a new tab)). In a mid-sized company that comes down to a few concrete changes:

• an owner for each use, with a target and a date;
• processes rewritten so that AI prepares the first draft and a person checks it and decides;
• clear rules on which data may be used, in which tools and for what;
• training that covers the new way of working, not only the tool.

EDON's view is that generative AI raises output most in the hands of people who can judge what good looks like. Experienced staff check an answer quickly and spot the one that is wrong. Keep them in the loop, and plan how newer staff will still build the judgement that makes the checking possible.

EDON is a strategy and transformation firm based in Finland, and generative AI is part of its own delivery. The work follows the company's decisions in the order of Exhibit 1: the position to take, the readiness check, the uses worth funding and their business case, and then delivery alongside your people, with the toolkits and training that let your team run and extend the work after EDON leaves.

Early in the work, EDON puts a high-value, usable deliverable in your hands, something you can act on and keep, before you commit further. We call it the Head Start.

Each engagement is scoped around the decisions in front of the owners and the board: what has to be decided, by when, and what you will hold at the end. Scope and deliverables are agreed in writing before the work starts, and we work alongside your IT team, your system suppliers and your legal adviser.