Skip to content
Nolli
Request a demo
Nolli Learn

Guide · Equip

What does AI-native actually look like?

A plain definition of AI-native, the five stages from AI-curious to AI-native, what an AI-native workday looks like, and ten signs your company is getting there.

The short answer

An AI-native company designs its work around AI. Agents handle the routine steps, people focus on judgment, review and improving the system, and what works is captured as shared skills, SOPs and evals that get better over time. Buying AI tools is not the same thing: many companies with AI licenses still run the same processes, with a chatbot on the side.

By Nadi Haque, Founder, NolliUpdated 9 min read

Key takeaways

  1. AI-native is about how the work is designed, not which tools you have bought.
  2. Companies move through five stages, from manual work to AI-native. Each stage has clear signs.
  3. In an AI-native company, people spend their time on judgment and oversight, and share what works so it compounds.

What does AI-native mean?

“AI-native” gets used loosely, so it helps to be precise. Bain & Company describes an AI-native enterprise as one that has moved beyond scattered AI tools and rebuilt its workflows, data and agents to change how the business competes. IBM puts it more simply: designed around AI from the ground up, not bolted on later.

For a company that already exists, the useful test isn’t whether AI is somewhere in the stack. It is whether the work itself has been redesigned. For each important workflow, do you know which steps an agent does, which run as plain code and which a person decides? And can you show it works better than before?

That is different from being AI-enabled, which means adding AI tools to work that otherwise runs the same way. Many companies are here. Bain reports that roughly 80% of CEOs are unhappy with the pace of their AI programs, and that most companies are managing a portfolio of pilots rather than changing how they work.

The idea isn’t new. In 1990, Michael Hammer warned in the Harvard Business Review that companies gain little when they use technology to speed up old processes, and urged them to redesign the work instead. His title said it in three words: “Don’t automate, obliterate.”

The five stages from AI-curious to AI-native

Companies tend to move through five stages, and different teams in the same company are often at different ones.

  1. Stage 0: Manual

    Work runs on people, email and spreadsheets. AI comes up in conversation, not in the work.

    Sign: no workflow uses AI    Work redesigned around AI
  2. Stage 1: Chat on the side

    People use ChatGPT, Claude or Copilot on their own. The process is unchanged.

    Sign: licenses, but no workflow looks different    Work redesigned around AI
  3. Stage 2: Assisted

    AI is built into everyday tools, and some teams share prompts and skills.

    Sign: a few teams save hours a week    Work redesigned around AI
  4. Stage 3: Agentic

    Agents run specific steps of real workflows, with people reviewing their work.

    Sign: an agent owns a step in production, with an approval point    Work redesigned around AI
  5. Stage 4: AI-native

    Work is designed around agents. People own judgment, review and improvement, and skills, SOPs and evals compound.

    Sign: every core workflow mapped step by step, and measured    Work redesigned around AI
Figure 1. Five stages of AI adoption. The bar shows how much of the work has been redesigned around AI, not how many tools are in use.

What does an AI-native workday look like?

Take one recurring job: a revenue operations lead preparing the weekly revenue review. In this example it takes two and a half hours every Monday morning, mostly spent gathering updates from 14 email threads, reconciling figures and rebuilding the deck.

Today: the lead2 h 30 m

Gather updates45 minReconcile figures40 minRebuild the deck40 minDecide & write25 min

AI-native: an agent, before 09:30no one’s time

Gathers updates, reconciles figures, builds the deck, flags 2 conflicts

AI-native: the lead40 min

Review15 minDecide & write25 minFreed for forecasting and coaching110 min
  • Moving information
  • Judgment
  • Done by an agent
  • Time freed
Figure 2. One Monday morning, before and after. An illustrative example. The agent ran in shadow mode for four weeks before it went live.

Three things changed. The routine work moved to an agent. The judgment stayed with the person and got more of their attention. And the time saved went to work that needs a person: forecasting, coaching and customers.

The agent didn’t appear overnight. The team first mapped how the review really ran, sorted each step into the four buckets, and ran the agent in shadow mode, comparing its output with theirs for four Mondays before giving it the job.

What changes for people?

Andrej Karpathy described the shift this way: as models do more of the legwork, more of our work will

“rise up the abstractions into oversight and understanding”

— Andrej Karpathy, on X, October 2026

In practice, people in an AI-native company:

  • Decide and review instead of gathering and re-keying.
  • Own the definition of correct for their work: the checks, the exceptions and the evals.
  • Improve the system, turning what they know into skills and SOPs that others can reuse.
  • Share what works, and can see which skills and agents actually perform.

Some firms already run this way. Workflows.io, a 20-person services company, keeps its entire operating knowledge in one shared repository: SOPs and playbooks, a file for each client, call transcripts, 79 reusable skills and 26 agents, as Fivos Aresti describes it. Every person, and every agent, works from the same context.

People are the hard part, and usually not because they resist. Bain finds the gap is clarity: employees know what is changing, but not how their own day-to-day work should change.

~80%

of CEOs are unhappy with the pace of their AI programs. Most companies are running a portfolio of pilots rather than redesigning how they work.

Source: Bain & Company, 2026

<40%

of employees say the scope and reasons for an AI-driven change are clear to them. Showing each person how their own work changes is part of the job.

Source: Bain & Company, 2026

Ten signs your company is becoming AI-native

Use these as a quick self-check, team by team.

  • For each core workflow, you know which steps an agent, plain code or a person handles.
  • Agents run real steps in production, with a person approving the risky ones.
  • You measure AI against how the work was done before, not by the number of pilots.
  • Teams share skills, and you can see which ones work.
  • Your evals capture what “correct” means for your work, exceptions included.
  • You choose models per task, based on cost and quality on your own cases.
  • Know-how is written down and current, not locked in a few people’s heads.
  • Agents can find what they need: owners, systems and rules, in a digital org.
  • You know which AI tools people actually use, and there is little shadow AI.
  • People spend more time on judgment than on moving information around.
Figure 3. Ten signs of an AI-native company. Teams in the same company usually score differently. Start where the score is lowest and the work matters most.

How do you become AI-native?

One workflow at a time, in four moves:

  1. See how the work really runs, from the people, the systems and the docs. See what forward-deployed engineers do.
  2. Decide where AI belongs by sorting every step into four buckets, and price each change by cost per task.
  3. Prove it on your own work: benchmark people, agents and both together, and run new agents in shadow mode first.
  4. Equip every person with what works: shared skills, SOPs and agents, with clear governance.

Start where AI changes the economics most, not where the demo looks best. Bain makes the same point: the companies that lead concentrate on a few areas and rebuild them from the ground up, instead of spreading pilots everywhere.

Questions people ask

What is the difference between AI-native and AI-enabled?

An AI-enabled company adds AI tools to the way it already works. An AI-native company redesigns the work itself, deciding step by step what agents, plain code and people should do, and measures the result.

Does becoming AI-native mean cutting jobs?

It doesn’t have to. Companies that do this well tend to move people from moving information around to higher-value work: judgment, customers and improving the system. How you use the time you free up is a choice.

How long does it take to become AI-native?

It happens one workflow at a time. A single workflow can be mapped, redesigned and proven in a few months; a whole company takes longer. Start with the few workflows where AI changes the economics most.

Can non-technical teams be AI-native?

Yes. Skills and agents are increasingly set up in plain language. What every team needs is someone who owns how “correct” is checked, and clear rules for what agents may do.

What is an example of an AI-native company?

Workflows.io, a 20-person services firm, says it runs the whole company from one shared repository: SOPs, client files, 79 skills and 26 agents that every person and agent works from.

Sources

  1. “What Is an AI-Native Enterprise?” Bain & Company, August 2026.
  2. Cole Stryker. “What Is AI Native?” IBM Think, February 3, 2026.
  3. Andrej Karpathy. “Post on understanding the outputs of language models.” X, October 2, 2026.
  4. Fivos Aresti. “How We Run Our Entire Company on Claude Code and GitHub.” Workflows.io, May 19, 2026.
  5. Daniel Kornum. “If AI is so great, why isn’t it working?” Varick Agents.
  6. Michael Hammer. “Reengineering Work: Don’t Automate, Obliterate.” Harvard Business Review, July–August 1990.