Faster is not the same as better off: how to get from AI use to AI return
AI ROI explained: why individual time savings with AI hit a ceiling, and how redesigning processes and ways of working turns AI use into real return.
Eight in ten executives say AI makes their decisions and execution faster, but only a quarter say they are using AI to change their organisation in a fundamental way, according to Deloitte’s State of AI in the Enterprise 2026, a survey of 3,500 leaders in 25 countries. That gap explains why so many companies struggle with AI ROI. Individual time savings are real, but they have a ceiling: a faster email or summary rarely shows up in revenue or margin. The return comes when a company redesigns a process around AI and changes the way the team works, instead of letting every employee use a chatbot in their own way.
Why doesn’t AI use automatically lead to AI return?
Because the saved minutes are scattered. When each person uses AI for their own tasks, everyone gains a little time, but nothing in the process changes. The report still goes through the same three people. The quote still waits for the same approval. The saved time quietly fills up with other work.
Large-scale research shows how small that effect can be. Economists Anders Humlum and Emilie Vestergaard studied 25,000 workers at 7,000 workplaces in Denmark, in occupations where AI chatbots were widely used. They found average time savings of 3 percent and no significant effect on earnings or hours worked (BFI working paper, May 2025). A revised version from March 2026 confirms that picture two years after ChatGPT’s launch.
That is not because AI does not work. In a controlled setting, a study of 5,179 customer support agents found that an AI assistant raised the number of issues resolved per hour by 14 percent on average and 34 percent for less experienced agents (Brynjolfsson, Li and Raymond, NBER). The difference is that in that study the AI was built into the work itself, not used on the side.
What is holding back the return on AI in companies?
Four forces keep the return low, even when everyone is enthusiastic.
Use without direction. In the Netherlands, ad-hoc use of AI grew from 32 to 38 percent of organisations between 2025 and 2026, while structured use, based on policy and goals, fell from 27 to 19 percent, according to Berenschot’s AI trend survey 2026. More people use AI, fewer companies steer it.
A ceiling on individual efficiency. Writing, summarising and searching faster is useful, but each task only has so many minutes to save. Beyond a certain point, the next gain has to come from changing which tasks exist at all.
Saved time without a destination. Berenschot quotes a lesson from the Association of Netherlands Municipalities (VNG): make it concrete what you will do with the time you save. Without that decision, the time disappears into the day.
Processes that stay the same. Deloitte’s own summary puts it in one line: speed is becoming common, deeper transformation is not. Speed is what individuals produce. Transformation is what organisations have to choose.
Where does AI deliver a real return?
Where a company redesigns a process around AI, and keeps people responsible for the outcome. Two public examples show what that looks like, including the limits.
Swedish payments company Klarna put an AI assistant at the front of its customer service in 2024. In its first month, the company reported, the assistant handled two thirds of customer chats, did work equivalent to 700 full-time agents, cut resolution time from 11 minutes to under 2 and reduced repeat questions by 25 percent. That is a return no individual chatbot user could produce, because the whole service process changed. A year later, Klarna started hiring people again: its CEO told Fortune that cost had been too dominant a factor and quality had suffered. The lesson is not that redesign failed, but that a redesigned process needs people in the right places.
Closer to home, ING is redesigning its mortgage assessment with AI agents. According to ING’s own description (June 2026), AI reviews the file and proposes next steps, and an employee makes the final decision. ING has not published outcome figures yet, so we cannot say what it returns. But the design choice is the interesting part: AI is not a tool next to the work, it is a step in the process with a clear handover to a person.
For a company of 50 or 150 people, the same logic applies on a smaller scale. Think of the route from customer request to quote, from incoming invoice to payment, or from intake to planning. Each has steps that AI can take over, and steps where a person adds the judgement.
How do you measure AI ROI in a way that means something?
Measure the process, not the person. The question is not how many minutes someone saves on an email, but how long the whole route takes and what it costs, before and after.
A useful measurement looks at lead time (from request to result), capacity (how many cases the same team handles), quality (how often work comes back) and the net time, including checking and correcting AI output. That last point matters: a step that is twice as fast but needs a full check afterwards has not gained much. Agree in advance what you will do with the freed time: more clients, faster delivery, or time for work that was always postponed. Then the return becomes visible in numbers the business already tracks.
From tools to a new way of working
The step from AI use to AI return is mainly an organisational step. In our view it rests on a few sober principles.
Choose a process, not a tool. Start from a route that matters to the business and runs every week, and only then decide which AI fits.
Redesign before you automate. Ask what the process would look like if one step took minutes instead of hours. Often a step can disappear, or two roles can work differently.
Keep ownership with people. Every AI step needs someone who checks the output and is responsible for the result, as the ING example shows.
Decide where the saved time goes. Make it concrete, and measure it.
This is where outside expertise is useful: someone who has redesigned similar processes and can build the AI steps into your software. On Baionic, companies describe the process they want to change and receive proposals from AI experts. You can read more about how we help SMEs put AI to work, or look at the process questions other companies have posted.
Frequently asked questions
What is AI ROI?
AI ROI is the return a company gets from its investment in AI, measured against what it costs. That return can be lower costs, more capacity, faster delivery or higher quality. It becomes visible when you measure a whole process before and after, not when you count minutes saved per person.
Why is the return on AI often disappointing?
Usually because AI is used by individuals on the side, while the process stays the same. A Danish study of 25,000 workers found average time savings of 3 percent and no effect on earnings or hours. The saved time spreads out and fills up with other work.
How do you calculate the ROI of an AI project?
Measure the process before you start: lead time, number of cases per person, error rate and costs. Measure the same after the change, including the time spent checking AI output, and subtract the build and running costs. Decide in advance what you will do with the freed capacity, otherwise it will not show up in the result.
Is it better to give everyone an AI tool or to start with one process?
Both have value, but they deliver different things. A tool for everyone builds familiarity and small personal gains. A redesigned process delivers a return the business can measure. Many companies combine them: basic access and guidelines for all, and one process redesigned properly.
Your next step
Pick one process that runs every week and matters to your customers, and measure how long it takes today from start to finish. For more background, see how companies post an AI question on Baionic. When you know which process you want to change, post it as a question on Baionic: within one working day you receive proposals from one to three matching experts, with their approach, price and planning side by side, and you pay only after you approve the delivery.
Ready to take the next step?
Post your project for free and receive proposals within one business day.