# HAL Capability Map

> Cross-cutting condition, not a ninth domain. Map what else the work did
> before automating it. Do not presume that the existing task should be preserved.

**Workflow:** ___________________   **Role affected:** ___________________   **Date:** ___________
**Owner of this map:** ___________________   **Next test date:** ___________

## Capability sustainability

The ability of an operating model to develop and maintain the human knowledge,
experience and judgement required for meaningful oversight and accountability
as automation changes the underlying work.

## 1. Before automating a task

| Diagnostic | Answer |
| --- | --- |
| **Primary output** — What does this work visibly produce? | |
| **Capability output** — What does somebody become better at by repeatedly doing it? | |
| **Mechanism** — What creates that improvement: repetition, feedback, consequence, proximity, variation or something else? | |
| **Necessity** — Does the redesigned system still need that human capability? | |
| **Replacement** — If yes, will the experience be preserved, compressed, simulated or moved? | |
| **Evidence** — How will the organisation know the resulting person is genuinely capable? | |
| **Sustainability** — What will keep the capability credible as automation increases? | |

If necessity is “no”, record that conclusion and automate without inventing a preservation requirement.

## 2. Map the hidden learning

1. **What does repetition teach?** Which patterns, variations and exceptions become recognisable only through exposure?

   Answer:

2. **Where does judgement develop?** Where do rules stop being sufficient?

   Answer:

3. **Where does feedback happen?** How does someone discover that their original judgement was wrong, and who explains why?

   Answer:

4. **What does real responsibility add?** What does deciding on live work create that a simulation may not reproduce?

   Answer:

5. **What does proximity teach?** What is learnt through exposure to experienced practitioners, including disagreement, incidents and recovery?

   Answer:

6. **What capability exists afterwards?** What can somebody recognise, understand or decide after two years that they could not at the start?

   Answer:

## 3. Decide the response

These are not levels and they do not need an acronym. A single role may need several.

| Response | Experience to retain or create | Owner | Where it occurs | Evidence expected |
| --- | --- | --- | --- | --- |
| **Preserve** direct experience | | | | |
| **Compress** volume into meaningful variation | | | | |
| **Simulate** rare, risky or unpredictable conditions | | | | |
| **Move** development elsewhere in the workflow | | | | |

## 4. The cycle

> Map the work → Map the hidden learning → Preserve / Compress / Simulate / Move → Test the capability → Monitor capability over time

## 5. HAL capability test

Independent manual performance may be important in some workflows but is not an absolute requirement where no person could reproduce the complete system task. The requirement is sufficient independent capability to exercise the assigned role.

- [ ] Can this person perform or independently reason about the relevant underlying task without starting from the AI answer?
- [ ] Do they understand the system’s limitations and the conditions in which its output should not be trusted?
- [ ] Can they explain why they agree and why they disagree?
- [ ] Can they identify when available information is insufficient?
- [ ] Have they demonstrated appropriate escalation under uncertainty?
- [ ] Are they exposed to enough meaningful variation and feedback to maintain their judgement?
- [ ] How will new people acquire equivalent capability?
- [ ] What evidence would indicate that capability is degrading?

Test with unfamiliar scenarios, plausible but incorrect outputs, incomplete information, conflicting sources, ambiguous cases, novel exceptions and realistic time pressure. Record whether the person recognised the issue, understood why it mattered, calibrated confidence and took the appropriate next action.

**Test date:** ___________   **Result:** ___________________________________________

## 6. Evidence state

Do not average this with the eight-domain control profile.

| State | Meaning | This workflow |
| --- | --- | --- |
| **Assumed** | Inferred from title, qualification or historic experience | ☐ |
| **Defined** | The judgement required for meaningful oversight is explicit | ☐ |
| **Designed** | Necessary developmental experience has been preserved, compressed, simulated or moved | ☐ |
| **Tested** | People have demonstrated the capability against realistic uncertainty and AI error | ☐ |
| **Sustained** | Performance, continued exposure and the future pipeline are monitored and acted upon | ☐ |

Current state: ___________   Weakest assumption: _________________________________

Owner: ___________________   Next test: ___________

## 7. Competence dimensions

Keep competence distinct from capacity. Capacity asks whether the person has enough time and attention. Competence asks whether they can make the judgement.

- **Existing:** Can this person exercise the required judgement today?
- **Maintained:** Does the role provide enough continued exposure, feedback and practice?
- **Future:** Does the operating model continue to produce people able to perform the role later?

Notes:

---

Part of the HAL Governance Toolkit · theloop.legal · CC BY 4.0 · Attribution: Loop / HAL Framework by Ryan McDonough