From Technical Builder to Business Collaboration

Designing a shared canvas where business teams map out

their process before IT builds it.

About

Opus is an enterprise platform where companies automate their business processes with AI.

KYC for Financial Insitution

Activate Workflow

What can I help you with today?

The challenge

Business users knew their processes best, but they couldn't really take part

in building the workflows.

  • The entry point was broken

    You type a prompt, press enter, and wait. 3–10 minutes, a full-screen loader, nothing to do in between.

    Natural language input

    Automatically generated workflows

    3–10 minute wait time

    Blocking loading state

    Results often inaccurate

    Required significant manual refinement

    So you waited a long time for a draft you'd most likely rewrite anyway.

  • AI helped, but only through prompts.

    Inside the Technical Canvas, AI could edit and generate parts of the workflow, but only through prompts. You describe the change, wait, check what it did, describe again.

    Assisted with editing workflows

    Accelerated repetitive tasks

    Prompt-based interaction

    Difficult to iterate naturally

    Required manual refinement

    Still demanded technical understanding

    Good for speed. Bad for iteration.

  • Creating workflows required technical expertise.

    The Technical Canvas was made for building, not for figuring out how a process should work.

    Powerful and flexible

    Optimized for IT teams

    High learning curve

    Didn't support cross-functional collaboration

    Difficult for business users

    Required technical knowledge

    A great tool - if you're an engineer. For everyone else, a closed door.

Key Insight

The real problem wasn't building workflows. It was that business intent got lost on the way to IT.

Business Owner

Express Business Intent

Lost Context

IT Teams

Technical Canvas

Workflow

By the time IT started building, most of the context was already gone.

Exploring different approaches

I used AI to prototype fast. The screens didn't need to be polished - I just needed something real enough to put in front of stakeholders, get their reactions, and try the next direction.

Exploration 1: Finding the right entry point

  • I tried giving users control over the generation itself — Lite for a quick rough draft, Full for the detailed workflow. If someone doesn't need every detail, they shouldn't have to wait 10 minutes for it.

  • I also tried killing the loading screen completely. Instead of staring at a spinner for 3–10 minutes, users see nodes appear on the canvas one by one, while the workflow is still generating. You can already look around, and you can see it's actually working.

Exploration 2: AI's role in the experience

Instead of one-shot prompts, what if the AI explained itself first? It shares its plan, walks through its reasoning, and only then builds the workflow in front of you. When you can see how it thinks, you trust what it makes.

Exploration 3: The builder was never designed for business users.

  • I added a short description to every generated node, so you can tell what a step does without opening it. Before, the only way to understand a workflow was to click into each node and read its configuration.

  • The Node Inspector was showing everything at once. I cut it down to the essentials and moved advanced configuration behind an extra click — business users see a simple panel, IT can still get to everything.

The wrong question

My first instinct was simple: make the Technical Canvas easier to learn. Lower the barrier for business users, keep the power for IT.

Improve Technical Canvas

Didn’t solve collaboration

Improve AI

Didn’t solve trust or accuracy

Talking to Product, Sales, and Delivery changed my mind. Customers didn't want an easier version of the builder, they wanted to skip the technical thinking completely and just explain how their process works. My "simpler canvas" idea solved neither side: IT loses flexibility, business users still have to think like engineers.

Reframing the problem

Instead of asking how we could simplify workflow implementation, we asked:

What if business users could define the process first, and implementation came after?

Before

After

Start with implementation

Start with business intent

Think in triggers & variables

Think in goals & processes

Business → IT handoff

Business + IT collaboration

Introducing Business Canvas

Business Canvas is where business owners define their process together, before any technical work starts.

HR Onboarding Process

I want to build an agent that extracts full name and nationality from files.

Worked for 5 sec

Good afternoon, Aidar I'm Opus. Today we're going to discover your HR Onboarding Process together.

You know how this process actually works in your team, and I'm here to help turn that into a clear, working process so the repeatable parts can eventually run on their own.

Before I draft the steps — is there anything you've already got that shows how this process works? An SOP, a process map or flowchart, training docs, even a rough diagram or notes?

1

Attach process files

Recommended

2

Describe files first

3

Something else

Type your own answer

Share it and I'll use it to build a more accurate first draft.

Ask anything...

+2

15h

4/20 Approved Steps

Process Owner

John Doe

Approve Steps

Discover the process before building it

  • Build the KYC process for a digital

    non-custodial crypto wallet.

    Worked for 5 sec

    Good evening, Aidar - I'm Opus. Today we're going to discover your KYC process for a digital non-custodial crypto wallet together.

    You know how this process actually works in your team, and I'm here to help turn that into a clear, working process so the repeatable parts can eventually run on their own.

    Before I draft the steps — is there anything you've already got that shows how this process works? An SOP, a process map or flowchart, training docs, even a rough diagram or notes? Share it and I'll use it to build a more accurate first draft.

    1

    Attach process files

    Recommended

    2

    Describe files first

    3

    Skip for now, just draft the steps

    4

    Something else

    Type your own answer

    Share it and I'll use it to build a more accurate first draft.

    Instead of generating right away, the AI interviews you. Where does the process start? Where does it end? Is there an SOP somewhere? The draft comes after the conversation, not before it.

  • Skip for now, just draft the steps.

    Worked for 12 sec

    From what I've seen, a process like this typically runs in the following order:

    Collect Customer Details

    Verify Identity Documents

    Assess Customer Risk

    Is Risk Acceptable (decision)

    Perform Enhanced Due Diligence

    Approve Wallet Access

    Reject Or Close Application

    Notify Customer Of Decision

    Wallet Access Approved

    Shall I lock this in?

    1

    Yes, lock it in

    Recommended

    2

    Let me modify the steps

    3

    Start fresh, I will describe the process

    4

    Something else

    Type your own answer

    Skip to Canvas

    From that conversation, the AI drafts the steps — business steps, not a technical workflow. "Verify identity documents," not triggers and variables. You go through the list with the AI, fix what's wrong, and lock it in. Only then does it move to the canvas.

Bring the right people into the process

One person can't describe the whole process, so each step is assigned to the person who does it in real life. They fill in the details the AI couldn't know, and sign off when the step looks right.

Finalize the business process

When every step is approved, the Business Canvas turns into the Technical Canvas. This is where the lost-context problem disappears, IT starts from the process everyone already agreed on, not from a retelling of it.

Why this is better?

Before

After

Prompt → Workflow

Conversation → Business Steps

One person

Multiple stakeholders

Technical-first

Business-first

Manual fixes

Validated before implementation

Outcome

Building a workflow no longer starts with IT. It starts as a business conversation, and the technical part comes last.

Business users don't have to think like engineers anymore.

The AI runs discovery, it interviews, drafts, and checks, instead of one-shot generating.

Every step is approved by its real owner before IT ever sees it.

The Technical Canvas is now the last step, not the first.

Home

Projects

Photos

Contact