The short answer: most people start with AI as a chat window that helps with one task at a time, and it helps. The next step is repeatable workflows: the same job done the same way every week. Then AI starts carrying work through several steps across your email, calendar and files while people review it. That is where most of the time savings show up. At the last level, AI lets your business do work it couldn't do before. That is where most of the value is.
Each level below describes what changes for the business and its people, with markers you can check yourself against. This is Sidekick's practical framework; no industry body sets these levels. External research is named where it supports a claim.
The levels describe what AI can do for your firm. The three stages of AI adoption (using, trusting and redesigning) are a second lens: how your organization absorbs AI once it's rolling out. A firm can reach Level 3 in one area while most of its people are still in the first stage.
The Four Levels of AI Adoption
Each level is defined by what changes for the business and its people. Tap a level for the details.
Nobody is early anymore
If you're just starting, you aren't early. In McKinsey's State of AI survey (August 2026, 1,719 respondents), nearly nine in ten report regular use of AI in at least one business function. In Canada, 19.2% of all businesses used AI in the year to spring 2026, three times the 2024 share (Statistics Canada, June 2026).
Most of that use is still shallow. In a survey of nearly 6,000 executives in the US, UK, Germany and Australia, nine in ten reported no effect yet on employment or productivity (NBER working paper, February 2026).
The firms that went first have worked out what works: start with one repeated task, write the instructions down, then connect it to your tools. Most firms have the same tools. The gap is how deeply they use them.
Level 1: Task help
What it looks like: you open ChatGPT, Claude, Copilot or Gemini when you're stuck. You ask it to draft an email, summarize a long document, tidy a proposal paragraph or answer a question you would otherwise search for.
What it gets you: minutes saved per task and better first drafts. Every so often, a moment where it does something useful you didn't expect.
What changes for people: one person gets faster at small things. How the business runs stays the same. Each session starts from scratch, because the assistant knows nothing about your clients, your templates or how you like things done.
Level 1 is worth doing well. It is where you learn what AI is good at, where it makes mistakes, and which parts of your week are worth handing over.
Level 2: Repeatable workflows
What it looks like: AI does specific, recurring jobs the same way every time. The instructions are written down once. The assistant has the documents it needs, such as your past proposals, your price list or your client intake form. Meeting notes become a list of action items with owners. A new enquiry becomes a first-draft reply in your house style. A month of timesheets becomes a client summary.
What it gets you: hours saved per week, and more consistent quality, because the same steps run every time.
What changes for people: the setup belongs to the team, so it is no longer one person's trick. People still start each job themselves, so the value depends on someone remembering to use it.
For most firms, getting solid at this level is the most useful next step. It is where you find the workflows that matter and learn where AI adds value and where it adds risk. Leaders should get proficient here first, then bring the team along.
If you lead a team, this is where Fractional AI Leadership starts: Sidekick works with your leaders to build the first shared workflows, then brings staff along. If you want to start with your own email, documents and meetings, Solo for Execs fits here.
Level 3: Connected, multi-step work
What it looks like: AI carries work through several steps across your tools, and a person reviews it before anything important goes out. The main AI assistants can now connect to email, calendar and shared files (Anthropic help centre, October 2026). They can also run tasks on a schedule or when a new email arrives, and take actions in connected apps. They pause for your approval before sending a message or changing data (OpenAI help centre, October 2026).
In practice: a client emails asking to move a project start date. The assistant reads the email, checks the calendar, pulls the project file, drafts a reply with two new dates and updates the task list. Someone on your team reads it, adjusts one line and sends it.
What it gets you: most of the time savings. Work moves from one step to the next without someone pushing it along, and people spend their time checking and deciding.
What changes for people: roles shift from doing the work to reviewing it. That needs new habits: clear rules about what AI may send without approval, which accounts it can reach, and who checks what. Judge the work by the quality of the result as well as the time saved.
Larger firms are moving fastest here. In McKinsey's 2026 survey, 40% of respondents at organizations with over USD 1 billion in revenue say they are scaling AI agents in at least one function, against 22% at smaller ones. The work is deciding which steps to connect, what stays with a person, and how to catch mistakes.
This is where Fractional AI Leadership can fit. Sidekick works with an executive sponsor and the people who run the work to improve one useful area at a time. We map the work, build and test the change, support adoption, measure what happened, and leave the client team able to run it.
Level 4: New capabilities
What it looks like: your business does work it couldn't do before. It offers a new service, or brings in house work that used to need a specialist, an agency or a new hire. For example, an accounting firm could send every client a plain-language quarterly report with industry comparisons. A contractor could write its own bid research. An HR consultancy could build the small internal tools it used to buy as software. That last one is already common. In McKinsey's 2026 survey, 32% of respondents said their organization chose to build at least one software product or feature in house with AI coding tools instead of buying it.
What it gets you: the biggest value of getting good with AI. Doing existing work faster frees hours. New capabilities change what the business can sell and what each person can do.
What changes for people: people take on work outside their old job description. Our founder, Chase Bernier, had never designed a website or produced marketing before Sidekick. He now uses AI to build its website, marketing and research (the longer story is in our article on where freed time goes).
Level 4 doesn't need complicated technology. It needs people who are proficient at Levels 2 and 3, plus the time to try something new.
What happens to people's time
Every level frees some time. What happens to it depends on whether anyone decides in advance where it goes. Freed time can go to three places: catching up, getting ahead and doing new things.
- Level 1: a few minutes here and there. They usually disappear into the day, which is fine.
- Levels 2 and 3: real hours. Most firms put them first into catching up: the work that is behind, now done faster and to a higher standard. Then into getting ahead: the proactive work you always wanted to do, such as checking in with clients before they call, or fixing the process everyone complains about.
- Level 4 brings do new things. Freed hours go into work that used to need a specialist, an agency or a hire, or into a new service for clients. That usually means upskilling people, and sometimes reskilling or reassigning them to new roles.
Our article on who decides where freed time goes covers these three places for freed time in more detail. To check one office task before you change it, try the hours test, a 10-minute worksheet for one office task.
How to tell where you are
Use these checks. You are at the highest level where you can answer yes to every line.
- Level 1: someone in the business uses an AI assistant at least weekly for client or office work.
- Level 2: you can name three AI workflows, each has written instructions, and more than one person runs them every week.
- Level 3: AI does the steps of at least one regular piece of work across your email, calendar or files. You have written rules for what AI may do without approval. A named person checks the output.
- Level 4: you can point to a service, deliverable or piece of work you produce today that you couldn't have produced, or would have paid someone else for, before using AI.
The levels can overlap. A small team can reach Level 4 in one area, such as marketing, while most of its work sits at Level 2.
Where to go from here
From Level 1 to Level 2: pick the one task your team repeats most often each week. Pick the most repetitive one, which is rarely the most complex. Write the instructions down, give the assistant the documents it needs, and run it the same way every time.
From Level 2 to Level 3: take a workflow that already works and connect it to the email, calendar and files it depends on. Decide what AI may do on its own and what a person approves. Start with work that stays inside the business before anything goes to clients.
From Level 3 to Level 4: list the work you pay others for and the services clients ask about that you never had capacity to offer. Give a proficient person protected time to try one of them.
The most common mistake is trying to skip levels: buying an agent platform before anyone knows which workflows are worth automating. Each level builds the habits and judgment the next one needs, so the order matters.
Sources checked October 16, 2026: McKinsey, The state of AI in 2026; Statistics Canada; Yotzov and others, Firm data on AI (NBER); OpenAI and Anthropic help centres.
Where are you?
Two quick questions to find your level.
Does AI carry any of your work through several steps across your email, calendar or files, with a person checking the result?
Leading a team? Start with the free AI Opportunity Map, about 10 minutes, or see Fractional AI Leadership, where Sidekick works with your leaders to improve one area of work at a time.
