Plenty of AI, little to show for it?

The licenses are bought, and every employee has an AI assistant. In the chat window, people summarize emails, draft quotes, write minutes and translate inquiries. Most are satisfied, and many say they save time every day.

And yet none of it shows up in the monthly figures. The time saved can be felt, but not in the result.

You are not alone in this. McKinsey’s current survey “The state of AI”, with just over 1,700 participants from 97 countries, shows that almost nine in ten companies use artificial intelligence regularly. Only 37 in 100 see any effect on their results at all. And only 6 in 100 attribute at least 5 percent of their operating profit to AI. That share has not grown since last year.

Almost 9 in 10use AI regularly

37 in 100see an effect on their results

6 in 100attribute at least 5% of their operating profit to AI

It is not the technology. The tools are open to everyone. The difference lies in the approach. And that approach suits SMBs better than many think.

Five things frontrunners do differently

What sets these companies apart is not secret knowledge. It is five habits. They do not guarantee success, but frontrunners practice them far more often than everyone else.

01They rebuild workflows instead of just introducing software

Anyone who has lived through an ERP rollout knows the pattern: new software, old workflows, little effect. With AI it repeats itself. A quote still travels from sales to engineering to costing, with Excel lists, follow-up questions and waiting time at every station. The AI writes the emails faster. That saves minutes, not days.

Frontrunners ask a different question: how would we set up this workflow if we started fresh today? Almost three quarters of them have fundamentally rebuilt workflows this way, compared with only one in four of the others. In an SMB this is often easier than in a large corporation. A workflow from quote to invoice can still be seen as a whole.

02They know in advance what the time saved is for

Almost everyone wants efficiency, especially in SMBs. Most frontrunners, however, also want AI to help them grow or build something new. That makes the difference, because time saved does not reach the result on its own. If a task takes ten minutes instead of an hour, the other 50 minutes are not automatically profit. They disappear into day-to-day business.

Impact only comes when it is clear beforehand what the time is for: more orders with the same team, faster quotes, relief for everyone who has been working extra hours for months. Where skilled people are scarce, this is the obvious answer. It also takes away employees’ worry about automating themselves out of a job. People who have that worry do not show their own time savings.

03Management takes it into its own hands

At frontrunners, top management is twice as likely to stand visibly behind AI initiatives. SMBs face a different risk than large corporations: AI runs on the side. A few employees try out AI assistants, the IT service provider sets up licenses, and day-to-day business comes first. So it never gets past trying things out.

Rebuilding a workflow, though, means shifting responsibilities and removing approvals. Only management can decide that. Nobody needs to write code for it. But someone has to decide which workflow changes and who answers for the result. The advantage in SMBs: paths are short, and the people who decide sit in the same building.

04They count in hours and euros, not in licenses

Frontrunners are twice as likely to have a fixed method for measuring the effect of their AI initiatives. Many others count what is easy to count: licenses, logins, requests to the assistant. That shows AI is being used. It does not show whether it pays off.

That takes a few simple questions up front. How long does a quote take today? How many hours go into checking invoices every week? How often is something queried, corrected or entered twice? From this comes a calculation in hours and euros. It shows early whether the effort is worth it, and it is the best basis for deciding the next step.

05They give AI responsibility once it has proven itself

Frontrunners are more than three times as likely to use AI agents at scale, meaning AI that completes whole work steps on its own. At the same time, they are far more likely to protect themselves against security gaps and unintended actions. The two belong together.

Every business knows the principle from onboarding. With new colleagues, you look over their shoulder at first. With every week in which the work is right, more responsibility follows. AI can be introduced the same way: at first a person checks every result. What proves itself keeps running on its own, with spot checks. A person always answers for the result. This builds trust that rests on results, not on promises.

What this means for SMBs

The McKinsey survey is international and not tailored to SMBs. Our work with SMBs adds four points.

Start with one workflow, not ten. Time and specialists are scarce, and day-to-day business comes first. Hardly any SMB can afford twenty pilot projects in parallel, and they achieve little anyway. One workflow with a noticeable benefit, implemented properly, is better. The question of where to start usually answers itself: where today the most hours, queries and errors arise.

Then go to the core. Once the first one or two workflows have changed and confidence in the company grows, the core value creation is next: what your customers pay for. For example the design of a machine, a formulation, or the sense of which process is right for which customer. It holds unique knowledge that no competitor can buy. That is why the effect is greatest here and the competitive advantage strongest. And this is where new business models can start, for example when decades of experience become a service that customers use directly. It is not the place to begin, because that is where the most is at stake. But it is where you should arrive.

Secure knowledge before it retires. In many businesses, the most important knowledge sits in a few heads. It is written down nowhere. Capturing it now and making it usable for AI lays the foundation for every further step. And it keeps what would otherwise leave with the next retirement.

Plan in years, deliver in months. SMBs plan for the long term, not from one quarterly report to the next. Delivery still has to be fast. The first workflow should run after three months, with a measurable effect on the result after six. Then the next one follows. This way the experience grows in your own company instead of staying with a consultancy.

Plenty of AI with little to show for it is not fate. It is a question of approach. Everyone has the tools today. The difference is made by companies that rebuild their workflows, decide beforehand what the time saved is for, and anchor the responsibility for it in management. SMBs bring more to this than many think.