The first step is the easiest
Has an AI assistant helped your team answer emails faster, summarise meetings, or put together a first draft? Then you've already taken the first, and easiest, step on a five-step ladder.
We've written before about why the payoff often doesn't show up — usually because data, access, or training weren't in place. But when the groundwork is actually done, the payoff is real and measurable. The question few businesses ask themselves is what comes after that first payoff has been captured.
The numbers when it works
In a recent internal pilot across 2,500 licensed users, 262 employees responded to a survey about how they actually use their AI assistant day to day. The results are worth noting, precisely because they confirm the payoff is real once the conditions are right:
- 54% of assigned licences were in active use
- 2.2 hours saved per week, on average
- 83% felt more productive
- 80% experienced better quality in their own work (against an industry benchmark of 68%)
- 84% completed tasks faster (against a benchmark of 73%)
- 70% experienced less mental effort on routine tasks
- 31% were already using AI agent features — not just chat
This isn't unique to one vendor. We see similar patterns regardless of which AI assistant sits underneath, as long as the groundwork is done. But notice that last number: 31% are already using agent features — AI doing something beyond answering questions. That number points to something bigger than saving time on email.
But productivity isn't a strategy
Most businesses we meet measure AI success with exactly the numbers above: adoption, time saved, perceived productivity. That's not wrong — it's just incomplete. It's one step out of five.
1. Personal AI. The employee asks a question, gets an answer, drafts something. The value is real, but confined to one person at a time. Most Norwegian SMBs are here today.
2. Knowledge AI. The AI retrieves and synthesises information across the business's own sources — documents, email, line-of-business systems — instead of employees searching manually through folders and threads.
3. Transactional AI. The AI prepares concrete actions — a draft reply, a compiled report, a fully structured deliverable — that a human approves before it's sent or saved.
4. Autonomous agents. The AI carries out composite tasks over time, across multiple tools, with minimal hand-holding — but still with a human setting the boundaries and approving what actually matters.
5. Digital worker. The AI owns an outcome over time, not just a task. This is the level fewest businesses should be aiming for yet, and it's not where we recommend Norwegian SMBs start.
We want to be clear on one point: climbing the ladder isn't about removing humans from decisions. The higher you go, the more it matters to know who approves what, and which decisions must always be made by a person. That's not a limitation we apologise for — it's how responsible AI use is supposed to work in a business, no matter how advanced the tool gets.
Four questions leadership should ask
Before deciding how high to climb, put these questions to your leadership team:
Vision. Will the next leap in AI capability be a tailwind or a headwind for you? Have you thought through what happens when the tool becomes substantially more capable a year from now — because it will?
Competitiveness. Will the business be more relevant in 1–3 years because of your AI strategy — or despite not having one?
Production function. Are you leading on productivity and outcomes across both human and AI-driven capacity — or just measuring how many people logged in?
Trust. How much are you willing to trust AI as business-critical workflows become AI-assisted? Where's the line, and who in the business sets it?
Few businesses have answered these in writing. That's often the difference between a business that climbs deliberately and one that stays stuck on step one without knowing why.
Where most businesses actually are — and why that's fine
The vast majority of Norwegian SMBs we talk to are on step one, some on step two. That's not a problem in itself. The problem is staying there without having decided to.
The goal isn't to push everyone to level four or five as fast as possible. It's to know which step you're on, which step actually creates the most value for your specific business, and what needs to be in place — data, access, policy — before you take the next one.
What comes next: why we point to Claude
When we recommend a tool to help clients move from step one to step three or four, we usually land on Claude from Anthropic. Not because it's the only option out there, but because it's built for exactly this climb:
- Claude Cowork is a desktop agent that works directly with local files and folders. It handles step-two and step-three work — pulling together information across documents, preparing a deliverable — without the employee having to copy data back and forth by hand.
- Claude in Microsoft 365 plugs directly into Word, Excel, PowerPoint, and Outlook — the same place you already work, not a separate window off to the side.
- Claude Managed Agents let you assign an agent a bounded, composite task over time — with logging and human approval built in from the start, not bolted on afterwards.
We've written more about how Claude enters the workflow without becoming a detached chatbot and about Claude Cowork specifically, if you want to go deeper.
This is also exactly what our own structure ladder is built to answer. Our entry-level products get you a safe start on steps one and two — with RBAC and a light-touch DPIA already in place. Structure & Scaling takes you toward step three: process automation and deeper integrations. Transformation is for those genuinely ready for step four — proactive AI agents and tailored analysis, still with a human owning the decision.
An AI Ready assessment is where to start. It tells you which step you're actually on today — not which step you think you're on.
The point that keeps repeating
The numbers at the top aren't the question. They show the payoff is real once the groundwork is in place — regardless of which AI assistant sits underneath. The question is whether the business has a deliberate plan for what happens after that first payoff is captured, or whether you'll stay on step one while your competitors keep climbing.
AI starts not with technology. It starts with structure.
Get in touch for a free AI Ready assessment →
Read also: Why Microsoft Copilot Isn't Working – and What You Need to Have in Place · Claude in the Browser Is Not an AI Strategy