AI delivers results only when processes are rebuilt

Eight in ten say AI has made them more productive. Only just over one in three companies see it in the bottom line (McKinsey, 2026). According to Kira Unger Söderlind, CEO of Miramis Technologies, the difference lies in identifying problems where the solution fundamentally changes the economics, and where implementation means real changes to ways of working and organisational structures.

Start where the economics change

Companies should start with AI where it changes the economics of a process, not where it saves an individual an hour.

Many treat AI as a separate “tech project”: they appoint an AI working group and a budget, evaluate business cases, run pilots, and buy copilots. Workflows, organisation, and economics are left untouched. But the effect is never bigger than the problem you choose to address. Should AI save an hour per person per day, or change what a team spends its time on, how fast decisions are made, and what a process costs to run?

Leaders need to identify and understand where AI and technology can fundamentally change the conditions of the company. The effect is greatest when the way of working itself changes.

“Installing a copilot is not the same as driving through real change. On the contrary, it can create a false sense of security, because the question feels ticked off,” says Kira Unger Söderlind.

What AI high performers do differently

AI high performers, the six percent who attribute at least five percent of operating profit to AI, have more often rebuilt their workflows than layered AI on top of existing processes. Nearly three quarters have done so, up from 55 percent the year before (McKinsey).

The gain moves from the individual to the process. As long as AI is a tool you choose to use, the effect depends on who is sitting at the keyboard. When AI is built into every step of the flow, that no longer matters, and the effect becomes systematic.

It also requires the right foundation. An agent is only as good as the information it can reach. The company’s institutional knowledge, such as contract history, contract templates, and delegation of authority, must therefore be available to it.

The more agents, the more important the rules

44 percent of all companies are now scaling AI across the business, up from 38 percent a year ago. The more you scale, the more important data, rules, and control become.

Agents can give the impression of doing the right thing. That lowers the threshold for going around established processes (especially painful ones) and, for example, writing or reviewing contracts on their own. The work does not get better, only faster, and the risk of errors increases.

Results come from combining the speed and flexibility of agents with a predefined layer of company-specific rules and data. That layer decides what the agent may do, what it may not do, and when a human should decide. This creates a controlled path to execution, where efficiency gains are inevitable.

Contract approval: do not automate a broken process

Contract management is a clear example. Manual processes and poorly suited tools cost time, lengthen the sales cycle, and create risk in the form of missed commitments and renewals. Yet it is an area where many AI initiatives miss the mark.

“AI on top of inadequate contract infrastructure often just produces more text. It does not solve the three problems we see at virtually every company. Contract drafts cannot be produced in a counterparty-ready format without a lawyer being involved. Coordinating everyone who has to approve deviations is enormously time-consuming. And renegotiations and commitments are missed, because nobody owns the contract after signing,” says Kira Unger Söderlind.

Take contract review and delegation of authority, where contracts are sent to the lawyer or a more senior manager. Reviewing faster is automating the problem. The goal should instead be that the business side can approve on its own, with control. Dynamic decision support lets an unchanged, pre-approved template through without approval, while a changed limitation of liability is always flagged and sometimes stopped.

A contract lifecycle management platform like Miramis solves the three problems directly. It also makes it possible for AI services such as Claude, Copilot, or GPT to find, understand, and process the contracts.

What needs to change in 2027?

What needs to change in 2027 is how companies judge AI: by which workflows look different, not by how many people use it.

“Don’t ask how many people use AI. Ask which workflows need to look different from today. That AI will be part of it is obvious. Then the results will show,” says Kira Unger Söderlind.

About Miramis

Miramis Technologies has built an AI-native contract management platform. The platform brings the whole contract lifecycle into one place: contracts are created from templates, negotiated against the company’s own playbook, sent for approval, signed digitally, and monitored and followed up after signing. Customers include Nordea, Loomis, and Permobil.