The work is changing faster than the job descriptions.
Agents already do a meaningful share of what public administration consists of. The states that handle this well will redesign the roles deliberately. The rest will discover the change through attrition and a scandal.
- Unit
- Task, not job
- Horizon
- 24 months
- Output
- Redesigned roles
- Scale
- National certification
Start with tasks, not job titles
The public debate about automation argues about occupations, which is the wrong unit. Almost no job disappears. Almost every job contains a set of tasks that an agent now performs faster, and another set where a human is still clearly better.
We decompose the work of a directorate into tasks, then score each one on whether an agent can do it, whether it should, and what supervision it needs.
Case triage, correspondence drafting, document review, translation, data reconciliation, and first-line enquiry handling usually score high. Judgement under political pressure, negotiation, and anything carrying statutory personal accountability usually does not.
The output is a map of where the work moves and what the remaining roles look like. It is uncomfortable reading, which is why it should be produced deliberately rather than discovered through a budget cut.
Deploy, then redesign
Agent deployment. Agents built for the specific tasks that scored well, wired into the systems staff already use, with a supervision model and an audit record for every action taken.
Role redesign. New job descriptions for the roles that change, written with your civil service commission so they survive the grading and pay negotiation rather than dying in it.
Supervisory capability. The scarce new skill is not prompting. It is knowing when an agent's output is wrong, which requires enough domain depth to catch a plausible error. That capability has to be built deliberately, and it is the one most programmes skip.
Redeployment paths. Where a role genuinely shrinks, a mapped route into the roles that are growing, with the training attached. Announced honestly and early, this is manageable. Discovered late, it is an industrial dispute.
Certification at institutional scale
Where an institution wants this across the whole organisation rather than one directorate, the work becomes an education programme with an assessment standard behind it.
We build the curriculum in tiers: a baseline for every affected role, a practitioner tier for staff who will supervise agents daily, and a specialist tier for the people who will build and evaluate them. Assessment is by demonstrated task rather than attendance, because a certificate that anyone can obtain by sitting in a room devalues the ones that mean something.
Delivery transfers to your own training function during the programme rather than after it, so the capability to certify the next cohort is yours before the first cohort graduates.
Programme components
The parts of this capability a technical evaluator will want to interrogate before a procurement decision.
Task inventory
What the organisation actually does, decomposed and scored for agent suitability, supervision need, and accountability constraints.
Agent portfolio
Agents built against the highest-value tasks, integrated into existing systems, with audit and supervision built in.
Role architecture
Redesigned job descriptions and grading, agreed with HR and staff representation before deployment rather than after.
Certification
Tiered curriculum with assessment by demonstrated task, transferred to your own training function.
Asked in most evaluations
Answers we would give in the room, written down so you can circulate them without a meeting.
Is this a redundancy programme?
It is a redesign programme, and we will not present it as anything else. In most administrations the near-term effect is that existing staff absorb backlogs and take on supervisory work rather than leave. Where a role genuinely shrinks we map the redeployment path rather than leaving it to attrition.
How do you handle union and staff association engagement?
Early and directly. Programmes that surprise staff representatives lose a year to disputes. We build the task inventory with staff involvement, because the people doing the work know which parts are mechanical better than any consultant does.
What is the first deliverable?
A task inventory and an agent portfolio proposal for one directorate, in six to eight weeks. That is enough to prove the method and to give leadership a realistic view of the scale before committing to the whole administration.
Can the certification run in our language?
Yes, and it should. Delivering an agentic-skills curriculum in a second language guarantees uneven adoption across the workforce. Curriculum and assessment materials are produced in the languages your civil service actually works in.
Adjacent capability
Each capability runs on the same collection and classification core, so evidence gathered for one is available to the others.
Sovereign AI
We design the reference architecture, stand up the platform, migrate the workloads that matter, and train the people who...
Health system transformation
National health capability rebuilt with modern AI: triage and demand management, diagnostic support, clinical documentat...
Open-source intelligence
Collection across press, broadcast, social, forums, registries, imagery, maritime and aviation data, fused against a per...
Bring us the question your last briefing could not answer.
Tell us the jurisdiction and the mandate. We will tell you within a week whether we are the right people for it.