Pathway card
From job groups to roles
Why does a grid of job groups — writers, marketers, project managers, developers, the reader, the AI agent — show who is involved but not who does what?
A pathway card shows where a practice starts, where it can get to, what does not change on the way, and the steps from one to the other. The path from before to after
The short answer
Because job groups say who has an interest in the source, not what each of them does with one piece of information. One job group plays several roles, and one row such as “the AI agent” hides three roles with three risks: an agent that takes values over, one that answers people and one that writes values back. The role grammar keeps the grid for the stakeholders and adds the roles where a decision depends on them.
For: Technical writers, technical project managers and documentation systems developers
Key points
- Job groups answer “who has an interest?”, roles answer “who does what?”.
- One job group plays several roles; one role can be played by several groups.
- “The AI agent” is three roles: consuming, answering, writing back.
- Keep the grid for the kickoff, the roles for change, release and AI.
Before and after
Before
Job groups as perspectives
- Four job groups, the reader and the AI agent in one grid.
- Each row says what a group contributes and needs.
- Rights and reviews follow the job group.
After
Roles where a decision depends on them
- Each decision names the role it depends on.
- The AI agent is named as consuming, answering or writing back.
- Rights and reviews follow the role.
Side by side
Each row names how far the control is as of the review date.
| Aspect | Before | After | Maturity |
|---|---|---|---|
| Unit | Job group | Role | Building block available A modelling decision |
| AI agent | One row | Three roles with three controls | Product-specific Agent platforms vary |
| Rights | By group or person | By role | Building block available Role-based access control |
| Writing back | Not distinguished | Gate and record per write | Product-specific Approval gates |
The path from before to after
Each step names who does it and what it leaves as evidence.
Keep the grid
Use the perspectives grid in the kickoff to see every stakeholder.- Who
- Technical project manager
- Evidence
- Stakeholder grid
Find the decisions
List the decisions that depend on what is done: changes, releases, agent connections.- Who
- Technical project manager
- Evidence
- Decision list
Write the role sentences
For each decision, write the role sentence with all four axes.- Who
- Technical writer with the system developers
- Evidence
- Role sentences
Split the agent
Name for each agent whether it takes over, answers or writes back, and give writing back a gate.- Who
- Documentation systems developers
- Evidence
- Agent role table
Limits and residual risk
Roles add precision only where a decision depends on them; elsewhere the grid is enough. Agent controls depend on the platform and are still developing.
The context
The simplification
The perspectives grid of the blueprint “AI-enabled information systems for technical writers” names four job groups — technical writers, technical marketers, technical project managers, documentation systems developers — plus the reader and the AI agent. It shows at a glance that an AI-enabled information system isn’t built by writers alone. For a kickoff, that is exactly the right level.
Where it becomes too coarse
A technical marketer supplies claims (writing), reviews product wording (deciding) and derives campaign texts (compiling). One row can’t say which of these a decision is about. And the row “AI agent” is three roles: an agent that takes values over for another system, one that answers people, and one that writes values back. Only the last one needs an approval gate against excessive agency — but all three are in one row.
The role grammar
The grid stays the map of stakeholders. Where a decision depends on what is done — a change, a release, the connection of an agent — the role sentence names the activity, the actor, the authority and the situation. Rights are then granted to roles, as role-based access control does, not to job groups.
Questions readers ask next
- Is the perspectives grid wrong?
- No. It answers who is involved and what each group needs. The role grammar answers a different question: who does what with one piece of information.
- Why does the AI agent split into three roles?
- Because each role carries a different risk: taking over spreads values, answering creates new text, writing back changes the source. Each needs a different control.
Review log and changes
Every context card is checked against its sources before it is published, and again whenever it changes; the date under the byline is the last review. Corrections (something was wrong) and additions (something was missing) are logged below with date and time (Berlin time). Typos, formatting and link fixes are not listed.
Reviewed
No corrections or additions since publication.