AI readiness: what to automate first
Describe your workflows; see which ones Aktor can automate, how safely, and build them.
AI readiness (in the menu, and offered once when you first sign in) tells you which of your team's workflows Aktor can automate now, how much autonomy is safe for each, and what to put in place first. Then it drafts a solution for any of them.
Run an assessment
- Say what your team does and pick the tools you use (both optional).
- Describe three to five repetitive workflows in plain words: what starts it, the steps, which tools, where the result goes, how often. Mention any written policy or checklist.
- Run it. The page shows each workflow as it's assessed. Each is one model call, counted toward your monthly usage; a workflow your organization already assessed with exactly the same words reuses that answer, at no cost, so running it again gives the same result. To get a fresh answer, change the description. Members and up can run one; everyone in the organization can read the results.
The four questions
| Question | Ready when… |
|---|---|
| Can it reach your tools? | The systems it uses have a connector, or at least an API. A system only a person can sign in to blocks it. |
| Rules or judgement? | Fixed rules over structured data need no AI at all. Reading emails, PDFs or chats, and making calls, is where an agent helps. |
| Are the rules written down? | The policy, checklist or examples a person follows are in one place. Agents can't follow rules nobody wrote down. |
| What if it's wrong? | Mistakes can be undone. Payments, deletions and messages to customers get an approval step. |
The heatmap shows every workflow against the four questions, darker meaning more ready. Hover or tap a cell for the reason.
The recommendation
| Tier | What it means |
|---|---|
| No-AI automation | Fixed actions and scripts: fast, free to run, the same answer every time. |
| Copilot · Level 1 | You start it; the agent gathers context and drafts; you make the call. |
| Supervised · Level 2 | Runs on its own, pausing for one-click approval before the steps that matter. |
| Autonomous · Level 3 | Runs end to end; people see only the exceptions. |
| Not feasible yet | A system it needs can't be reached by software. |
The model only reads your descriptions and records facts; fixed rules turn the facts into the tier and score, so the same description always gets the same answer.
Build it
Build this solution opens the solution builder with the designer already asked to build it: the systems, the steps (agents only where judgement is needed), and the safety settings that match the tier (the workspace's autonomy, and approval rules for anything that can't be undone). Review the proposal and apply it.
Run the assessment again after you connect a tool or write down a policy: the scores move.