Operating map
    for AI transformation

    Map how your company works. See which steps AI can take over. Get specific recommendations for AI tools and agents with prices, timelines, and a plan for getting there.

    Try itBook a demo
    01

    See how the work runs

    AI interviews map every step, role, system and hand-off

    02

    Try what AI can do

    Start with a skill on the assistant your team already uses, or a ready-made AI tool, priced for each step

    03

    Redesign the process

    Decide which parts of the work agents take over and see the new process on the same live map

    04

    Build or buy

    Compare building your own agent with a ready-made AI tool or agent: cost, timeline and payback side by side

    Live workflow layer for your AI agents

    The operational context your agents are missing: every step with its systems, data, decision rules and how it works alongside your team

    Interviewsbeta

    Mappingbeta

    Recommendationsbeta

    Transition plan

    Agent specs

    Monitoring

    Step 01

    Interviews

    An AI interviewer talks to the people who do the work, one question at a time: what they do, how often, how long it takes, which tools they use and where it gets stuck.

    They can attach SOPs, spreadsheets or meeting notes, and the map builds as they answer.

    Interviews run in the browser or inside your team's AI assistant through MCP.

    Step 02

    Mapping

    Every step gets its role, systems and documents. Decisions show how often each branch is taken, and handovers between teams are marked.

    Each process sits on your company’s value chain. To fix something on the map, say it in plain words.

    Step 03

    Recommendations

    For each process you get costed recommendations: which steps stay with people, which go to a ready-made AI tool and where an agent is worth building.

    Every option shows its price, time saved and payback. Switch options on and off, and the to-be map shows how the process runs with the ones you picked.

    Step 04

    Transition plan

    A rollout plan for the options you picked: what to build or buy first, in what order, and when each part goes live.

    Each agent moves through build, shadow mode, pilot and production on one timeline, with the one-off cost, the first go-live date and when the plan pays back.

    Step 05

    Agent specs

    Every agent you decide to build comes with a spec written from your process: when it starts, what it does, where a person steps in, which systems and data it uses.

    Hand the spec to your engineers or a vendor and start building.

    Step 06

    Monitoring

    Follow-up interviews at the cadence you choose show what has been adopted, what works and where the plan needs to change.

    Each round updates the map and the plan, and you see actual savings next to planned ones.

    kawaru MCP

    Run kawaru in your team’s AI tools

    Run interviews and build the map right in the chat, or ask how a process works, who does what and what AI can take over

    How kawaru MCP works →
    FAQ

    Questions
    & Answers

    You can run it in Claude or ChatGPT through kawaru MCP, or send people an invite link to answer in the browser.
    You can map the work without it. People name the systems they use at each step and the map records them. Access only matters later, when you build an agent that has to work inside those systems.
    Claude and ChatGPT.
    You can start from a template for your industry: its value chain, a library of standard processes and end-to-end flows such as Order to cash. The interview then only asks where your company does it differently.
    You can attach procedures, SOPs, spreadsheets or meeting transcripts to the interview. kawaru reads them first and only asks about what is missing or out of date. You can also import existing BPMN diagrams into the editor.
    One sitting per process, as a chat. kawaru asks one question at a time, drafts the map as you answer, and you fix anything in plain words.
    Interviews capture what systems don’t record: the handovers, judgement calls and exceptions that decide whether an agent can take a step. They also tie the processes of different teams into one connected map. Bringing in data from systems is on our roadmap, on top of that map.
    You can turn it into a plan. kawaru assesses every step, suggests ready-made AI tools from a database we check ourselves, designs the agents you would build with a ready spec, and prices the whole plan. Switch options on and off and the plan recalculates.
    Both. The main path comes first; every decision point shows how often each branch is taken, and steps that happen only sometimes are marked as variants.
    We ask the people who do the work how often it happens and how long each item takes. Every figure in the plan is either a number someone stated, a default with its source for you to review, or a labelled estimate.
    For each step, process and the whole plan: time saved, what the agent’s tokens cost, and the saving after tokens. Add the one-off build, deploying from the spec and connecting the systems, and you get the payback period. Change any input and the plan reprices.
    Four sources. A process mapping methodology that goes from company-level operations down to single steps and shows which processes to automate first. A database of AI tools for business tasks with prices, reviews and use cases we check ourselves. A database of agent skills with install counts, rechecked as tools come and go. And consulting teams who run kawaru on their projects: what they learn feeds the recommendations, and when a transition needs hands-on help, partners do it.
    Yes. Create a project for each engagement and invite the company’s people to the interviews. Your team works in the same project, and the maps, agent specs and plan export into your deliverable.
    EPC with role, system and documents on every step; BPMN is supported too. The value chain sits above the maps.

    Try kawaru now

    Get startedBook a demo