Free prototype

How autonomous is your enterprise?

Score one process in about three minutes. You’ll see which of the five levels it runs at, what holds it back and how to move it up.

1 of 13
Which area is the process in?
Which process should we score?

In your own words, for example quoting from RFQs.

Who starts each task?

For this process, as it runs today.

Who does the work?

For this process, as it runs today.

Who checks the result?

For this process, as it runs today.

How are exceptions caught?

For this process, as it runs today.

How far has AI been rolled out across the company?
How connected are the systems this work touches?
Does anyone know how this work really flows from start to finish?
Could a new hire do this work from written rules?
Who would build and run AI for this work?
Which data could you export for this process?

Choose any that apply.

Last step: your work email.

Your result appears right away. We may follow up about it, and we never share your answers.

Autonomy comes down to three questions.

Level Who starts the taskWho does the workWho checks the result
1Manual PersonPersonPerson
2Assisted Most enterprises are here PersonAIPerson
3Supervised AIAIPerson
4Exceptions only AIAIPerson, for exceptions
5Autonomous AIAIAI

The five levels of autonomy.

Level 1

Manual

People do all of the work.

Starts
Person
Works
Person
Checks
Person

Where you are

Work arrives by email, phone and spreadsheet, and the people who know how to handle it pick it up. A few experienced people know where everything is and which customers need special handling. When they are away, work slows down. You grow by hiring.

What holds you here

The know-how lives in a few people’s heads, so no tool, AI or otherwise, has anything to work from.

How to reach level 2

  1. Write down the decisions your team makes most often, and who approves the exceptions.
  2. Agree which system is the source of record for each piece of information.
  3. Let people use an approved AI assistant for drafting and summarizing.

You’re ready for level 2 when

  • A colleague can handle a routine case without calling the usual expert.
  • Your team can find the current version of every record the work depends on.

What to measure

Time spent chasing internal answers.

For example: A rep builds every quote in Excel from old quotes and the price list.

Level 2

Assisted

Your people use AI, but a person still starts and checks every task.

Starts
Person
Works
AI
Checks
Person

Where you are

Your team has ChatGPT or Copilot. Someone pastes the work into a chat, edits the draft and sends it on. Each task is a little faster, but nothing happens until someone opens a chat, and someone checks every result. You still grow capacity by hiring.

What holds you here

AI sits outside the systems the work runs on, so a person has to carry the information in and carry the result back out.

How to reach level 3

  1. Pick one high-volume process with a clear definition of done.
  2. Connect an agent to the systems and records that process runs on.
  3. Let the agent start the work when it arrives, and keep a person reviewing every result.

You’re ready for level 3 when

  • Work starts on its own when it arrives, without anyone prompting.
  • A reviewer can check each result against the source records in minutes.

What to measure

Time from arrival to done, and review time per case.

For example: A rep pastes the RFQ into ChatGPT and fixes the draft quote.

Level 3

Supervised

Agents start and do the work, and a person reviews every result.

Starts
AI
Works
AI
Checks
Person

Where you are

When work arrives, an agent picks it up, does it across your systems and prepares the result. Your people review each one before it goes out. The team handles far more work than before, but every case still waits for a reviewer.

What holds you here

Review becomes the bottleneck. Every routine case gets the same attention as the hard ones.

How to reach level 4

  1. Track where reviewers change results and why.
  2. Agree which kinds of cases can go out without review, and set approval limits.
  3. Send uncertain cases to the right person with the evidence attached, and store each decision so the agent learns from it.

You’re ready for level 4 when

  • Reviewers rarely change routine results.
  • The process owner can see every unfinished or failed case and resolve it.

What to measure

How often reviewers change results, and cases per reviewer.

For example: An agent drafts every quote when the RFQ arrives, and a rep reviews each one.

Level 4

Exceptions only

Routine work goes out on its own, and people handle the exceptions.

Starts
AI
Works
AI
Checks
Person, for exceptions

Where you are

Agents complete routine cases from start to finish. Anything unusual is flagged and sent to the right person with the evidence. Your people spend their time on exceptions and on the decisions that need judgment, not on moving data between systems.

What holds you here

Decisions stop at department borders. A change in one process still needs people to update the next one.

How to reach level 5

  1. Connect the neighbouring processes, so a change follows through without someone chasing it.
  2. Feed every exception decision back into the agent’s memory.
  3. Raise approval limits as results prove out, and lower them when they do not.

You’re ready for level 5 when

  • The exception rate keeps falling without more reviewers.
  • The process owner can change limits and see the full trail of every decision.

What to measure

Exception rate, and cost per case.

For example: Routine quotes go out on their own, and a rep handles the flagged lines.

Level 5

Autonomous

Agents start, do and check the work, and your team sets the direction.

Starts
AI
Works
AI
Checks
AI

Where you are

Agents run the process end to end and check their own results against the rules and limits you set. Your team decides the goals, the limits and the strategy, and steps in when the agents ask.

What holds you here

Nothing in the process itself. What remains is governance: who sets the goals and limits, and how you know the results are right.

How to keep it there

  1. Review the results against the goals you set, not case by case.
  2. Keep the full audit trail and the ability to change limits or step in at any point.
  3. Take what you learned to the next process.

You know it’s working when

  • You can show better margins or service, after counting review effort and AI running costs.
  • You can inspect any decision, change the limits and intervene at any time.

What to measure

Margin, cost to serve and service level.

For example: An agent runs quoting and follow-ups, and the team sets pricing strategy.

How a process goes live with Bourne.

  1. 01Prototype

    Send us extracts from the systems you already use. We map how the process really runs and build a working prototype on your data, for free.

  2. 02Pilot

    For 2 to 8 weeks, a small team uses it on live work, and we compare the results against how the work is measured today.

  3. 03Production

    We connect the agents to your systems, complete security review and set permissions. Routine cases go out on their own as results come in.

  4. 04Expand

    Every decision your team makes is remembered, so the process keeps improving. The next process goes live faster than the first.

230,000+ hours of AI labor completed annually

  1. Five-minute quotes for an industrial manufacturer Industrial manufacturer
  2. Selling excess steel across world markets Steel company
  3. Resolving 68% of support tickets Software company
  4. Administering payroll and benefits Benefits administration company

300+ enterprises trust us

$45M funding

Questions

How is the level worked out?

From who starts each task, who does the work and who checks the result: the three questions behind the five levels. Without automatic exception flagging, a process stays at level 3, because routine cases and exceptions can’t be told apart. The company questions don’t change the level. They show what holds the process back.

What happens to my answers?

We save them with your work email so our team can follow up about your result. We never share them.

What if the process runs differently across teams?

Score the normal process, not a pilot or the one expert who does it differently. If teams or sites work differently, score each one separately.

Should every process reach level 5?

No. Many processes are best run at level 4 for good, with people handling the exceptions. The goal is to move each process to the level where it pays off.

Is the prototype really free?

Yes. You send extracts from the systems you already use, redacted as needed and shared under NDA, and we build a working prototype for one process. A production deployment is scoped and priced separately.

See your enterprise run itself.

Send us data extracts from the systems you already use. Bourne maps one of your operational problems from them and builds a working prototype that solves it, for free.

Run Bourne on your data for free