
A lot of businesses don't really know where they are with AI yet. They think they're using it. But how do you evaluate where you are on your AI growth journey, and what it would take to go further?
At Codi Studio, we use four levels to work that out. They're our own way of evaluating AI in a business:
- Level 1, the assistant: people ask AI questions and it responds.
- Level 2, automation: systems are connected, AI has context, and routine tasks are set up.
- Level 3, agents: AI works towards a goal and measures whether it has hit it.
- Level 4, the orchestrated business: multiple agents work together towards a shared goal, like a department.
In this guide, we'll look at where most businesses are today, what each level looks like inside an organisation, what it takes to advance, and where to aim by the end of 2027.
Where most businesses are with AI today
When I speak to businesses, a lot of them are at level 1 or level 2. Very few are truly at level 3, and I'm yet to see a level 4. There are businesses out there running multi-agent systems, but not among the £10M to £25M businesses I work with.
There's a misconception in how people describe their use of AI. They tell me they're using it, and then I sit down with people across different departments and see that they're using it as a replacement for Google. What they're not doing is embedding it into their business workflows, or building agents that take on work on their behalf.
It doesn't change much as you go up the organisation. If anything, the CEOs I speak to are sometimes less advanced than their teams. They're very busy people, and often less immersed in the technology day to day.
I think the gap comes from not understanding what AI can really do. When I show these levels in a talk, a lot of people immediately say "I'm at level 1". Until that moment, automations and agents seemed so technical and overwhelming that they hadn't realised how much more is possible with the technology that's here today. Most people think that by being at level 1, they've completed AI. The reality is that there's so much more to it.
The four levels of AI in business
The way to evaluate your level is to look at what's in place. How much people use AI tells you very little.
Level 1: using AI as an assistant
At level 1, what we're seeing in organisations is that people are still using AI reactively. Whether it's ChatGPT, Claude or Copilot, they're asking it questions, much like they would Google. Can you edit this report? Can you draft this email? Can you tweak this social post? AI responds, but it's still very dependent on the person asking. Nothing is being done for them.
You're at level 1 if:
- People chat to AI to answer questions or polish a piece of writing.
- Nothing is connected. AI can't see your drive, your inbox or your calendar.
- Every conversation starts from scratch, with no context about you or your business.
Level 2: AI automation
At level 2, what we're seeing is that businesses are connecting their systems to AI: their Google Drive or OneDrive, their email and their calendar. They're building memory files and templates that AI can use, like brand guidelines, a template slide deck or a template proposal.
And they've set up routine tasks that run on a trigger. That could be a time, like AI drafting their emails every day at 9am. Or it could be something happening, like a meeting finishing: every time I come off a call where a client and I discuss a new proposal, AI starts drafting that proposal from the meeting notes.
You're at level 2 if:
- Your tools are connected to AI: drive, email, calendar, meeting notes.
- AI has context files about you and the business, like your tone of voice, goals and templates.
- Routine tasks are set up to run on a trigger, without anyone asking each time.
Level 3: AI agents
At level 3, what we're seeing is that businesses have a goal they want AI to hit, and AI has a way of measuring whether it has in fact hit that goal. They've connected their systems and given AI context, and they're building skills: training AI on the things they do day to day, so it can complete a whole workflow end to end.
The example I always give is an agent that helps a business rank for its top five keywords. It's connected to a tool like Semrush or Ahrefs, so it can see how you're ranking every week. It recommends the blog to write, the page to update and the script for a YouTube video, and gives you the drafts. Then, every week, it checks again: how has our ranking on Google moved for those five keywords?
You're at level 3 if:
- You have agents with a specific goal, like booking meetings or improving keyword rankings.
- They can measure their own progress against that goal, and run in a loop.
- They use several skills to do the whole workflow, not just one step.
Level 4: the orchestrated business
Level 4 is where it gets really interesting. What we're seeing in the most advanced businesses is multiple agents working towards goals together, similar to a department. In a marketing team, that might be a group of agents working towards a pipeline or revenue goal, with agents checking other agents' work, orchestrated by a person who approves the work.
You're at level 4 if:
- All your technology is connected, and AI has the memory, templates and context it needs.
- Whole departments have defined goals, particularly growth goals, with clear rules on what information each agent can access.
- A person approves the work and owns the outcome.
Doing more with AI isn't the goal
Even in businesses that have moved into level 2 or level 3, I'm yet to see many relate it back to the P&L. I've seen a lot of marketing teams increase their output: more content, more campaigns, more landing pages. But I'm yet to see growth metrics from a lot of what they're building.
This isn't just a marketing problem. It's a trap of AI at every level. We can focus too quickly on more output. Sometimes that's a good thing, but unless it's driving growth, what's the point in doing more? Being busier because of AI is not necessarily a good thing.
Advancing a level isn't about using AI more, either. It shouldn't be about token usage. It should be about the output, and whether that output grows the business. What we need to be doing is automating to enable growth.
Becoming an AI-native organisation starts with mapping your business
When we do this with a client, we bring together leaders from across the organisation. We map out each department and everything that department does: every job of work.
Most CEOs and leadership teams have never been through this exercise. They've never written down every workflow across the whole business. Seeing it as a whole, then zooming out to think about where AI can be used, changes the conversation.
As we dig deeper into particular workflows, we often find a business isn't quite ready for AI yet. There are three things AI needs that are usually missing:
- Good data quality. Good in is good out, and a lot of the time data quality is still really poor.
- Documented processes and templates. There's no template proposal or slide deck, the brand guidelines haven't been updated, or the process for how a salesperson generates a lead through outbound has never been written down.
- The right technology. For an agent to work, it needs to be able to check whether it has met its goal. My keyword agent needs a tool like Semrush or Ahrefs, plus access to Google Search Console to check whether the business has in fact ranked on Google. Lots of businesses already have a large tech stack, so often it's about fitting what you have to AI. But if you think AI can just do lead generation for you, the reality is that you have to invest in the technology that makes it work.
A lot of AI transformation comes from amplifying existing efforts. It's much better if you've already seen value from a workflow, and then you figure out how to use AI around it.
What usually surprises leadership teams is that it isn't so daunting. Until you start working through it, AI can feel overwhelming. Once the workflows are mapped out, the aha moment happens. People get really excited: "Oh my gosh, I can use AI here, or we could do this, or we could do that."
How to advance AI in your organisation: from level 1 to level 2
The main step from level 1 to level 2 is connecting your systems: your Google Drive or OneDrive, your calendar and your email.
Then start creating some memory files, also called context files, such as:
- your tone of voice;
- a document about you;
- your brand guidelines;
- your business goals;
- your own personal goals.
These help AI learn how it can help you. They give context to what it's trying to do, so it performs better for you.
What should stay human as you adopt AI
What should stay human is up to each business. Think deeply about what makes you unique. As AI is rolled out across every organisation, technology stops being a competitive advantage, because everyone will have it.
So when everyone is automating everything, what is your competitive moat? Deciding what stays human, and what keeps you unique, is a decision for the CEO.
Your AI roadmap: where to aim by the end of 2027
You might not be a fully orchestrated business by the end of 2027, and that's fine. By then, a business should have:
- mapped out the entire business;
- a clear AI roadmap;
- made progress on key priorities, like data quality and governance;
- pilots running across different departments, focused on workflows that support growth;
- clear measurement, so you know where you are today and what you want to achieve.
That's when you start to see returns on AI, and it stops being a cost centre.
How to evaluate your AI level in under a minute
Ask yourself, and then ask your team:
- Where do you think you are?
- Where do you think your team is?
- What have you built, and how are you using AI?
- Do you have agents doing work for you?
What to do if your business is at level 1
There's a huge opportunity in 2027 to make AI a growth lever for your business. It starts with mapping out your entire business and identifying the areas to automate, while holding back on what makes you unique.
Done well, AI can accelerate profit margins and commercial growth while reducing costs. That comes from moving through these levels, and especially from reaching level 4: redesigning your organisation around AI, for the AI era.
At Codi Studio, we start by mapping the business. If you'd like to find out where your business really sits and what to do next, book a 30-minute call.
FAQs
How do you evaluate AI in a business?
Look at what's in place rather than how much people use AI. At Codi Studio, we use four levels: the assistant (people ask AI questions), automation (connected systems, context files and triggered tasks), agents (AI working towards a goal it can measure) and the orchestrated business (multiple agents working together towards department goals, with a person approving the work).
What's the difference between AI automation and an AI agent?
An automation runs a routine task when something triggers it, like drafting your emails at 9am. An agent is given a goal, like ranking for five keywords, and works towards it in a loop, measuring its own progress and using several skills along the way.
What does a business need before it can use AI agents?
Good data quality, documented processes and templates, connected systems, and the technology an agent needs to check whether it has met its goal. Most businesses also need to map out their departments and workflows first.
How do you become an AI-native organisation?
Start by mapping every department and workflow in the business. Then fix the foundations AI needs, decide what should stay human, and point AI at the workflows that drive growth, measuring each one.
Where should my business aim to be with AI by the end of 2027?
Mapped out, with a clear AI roadmap, progress on data quality and governance, and growth-focused pilots running across departments, all clearly measured. A fully orchestrated business can come later.
