No result found
This report needs a scored result to generate.
This is a static sample and does not depend on a scored result.
Back to the course page
PRAXORALAB Illustrative example · fictional company, for format reference only
AI-Literacy Baseline
Northfield Mutual Insurance · the Claims Processing team
Overall Literacy Score Developing
A mixed profile: the team already runs the assistant daily and checks its numbers, but two named gaps, task judgement and prompt discipline, decide whether the rollout stays consistent or turns into fifteen different habits.
| Dimension | Score | Status | |
|---|---|---|---|
| Tool fluency | 2 / 3 | Developing | |
| Task judgement | 1 / 3 | Blocker | |
| Escalation awareness | 2 / 3 | Developing | |
| Prompt discipline | 1 / 3 | Blocker | |
| Verification habit | 2 / 3 | Developing |
Priority: Task judgement
Adjusters can't yet reliably sort which of their own claims are safe to let the assistant draft unsupervised versus which need a human first; today that call is made by habit, not by a rule anyone could write down and hand to a new hire.
An eight-out-of-fifteen score is the profile I see most often on a first pass, and the shape of it matters more than the total. Northfield Mutual's Claims Processing team isn't behind; tool fluency and verification habit both land at 2, which tells me the team has already absorbed the assistant into daily work and hasn't stopped checking the numbers it produces. That's a genuinely good starting position, and it's exactly the kind of profile that lulls a leadership team into thinking the literacy question is basically answered. It isn't, because the two dimensions sitting at 1, task judgement and prompt discipline, aren't decorative gaps sitting quietly beside the good scores. They're structurally connected to each other, and they're the two dimensions that determine whether this rollout produces a consistent claims experience or a randomised one.
Task judgement scoring below prompt discipline is the more consequential read, not the more visible one. When an adjuster can't articulate a rule for which claim types are safe to hand the assistant unsupervised, that judgement doesn't disappear, it gets replaced by fifteen individual habits, one per adjuster, none of them written down. I've watched this exact pattern across claims, underwriting and customer-service teams in more than twenty markets: the team feels ready because everyone individually feels confident, and confidence at the individual level is precisely what prompt discipline scoring a 1 exposes as false comfort. Two adjusters pulling two different draft coverage determinations from an identical file isn't a hypothetical risk here, it's the direct, provable consequence of a team where judgement was never made explicit and prompting was never standardised.
What I'd flag hardest for Northfield Mutual specifically, given this is an insurer, is that escalation awareness scoring a 2 is doing less protective work than the number suggests. The team correctly escalates the loud risks, fatalities, six-figure exposure, but a contested coverage question, the quiet failure mode, sits exactly in the gap between escalation awareness and task judgement. That's not a coincidence; it's the same underlying gap showing up twice, once in what gets escalated and once in what gets drafted unsupervised in the first place.
My recommendation, and the one this course builds live for a real team, is to fix task judgement before touching prompt discipline, not alongside it. A written rule for what's safe to draft unsupervised is what gives the team something concrete to standardise their prompting against. Reverse that order, and you end up training people to prompt consistently for a boundary nobody has actually agreed on yet.
PRAXORALAB Fluency Gap Map
Built for one live use case brought into the room. Northfield Mutual Insurance's example use case: an AI drafting-and-triage copilot piloted on first-notice-of-loss (FNOL) claims intake. Every task below is sorted into exactly one of three fluency levels, not left as a general policy statement.
Level 1 — Runs unsupervised
Acknowledgement and status-update letters for claims under S$2,000 with no injury and a single named party.
Summarising recorded intake calls into the claim file for the adjuster's first read.
Level 2 — AI drafts, an adjuster verifies before it goes out
Initial coverage determination letters for claims between S$2,000 and S$25,000.
Fraud-indicator flags surfaced for a senior adjuster's review, not acted on automatically.
Level 3 — Not yet, fully human
Any claim involving bodily injury, a fatality, or active litigation.
Any coverage question where the policy language is genuinely contested or ambiguous.
Claims above S$25,000 in exposure.
The two dollar thresholds on this page, S$2,000 and S$25,000, are judgement calls the room reached, not numbers a formula produced, and it's worth being explicit about why those specific figures earned their place. The Level 1 boundary sits at the point where a wrong unsupervised draft is cheap and reversible: a S$2,000 acknowledgement letter that needs a correction costs Northfield an apology email, not a reopened file. Push that boundary to S$10,000 without evidence the team's task judgement has actually improved, and "cheap and reversible" quietly stops being true, because a correction on a larger claim takes real adjuster time to unwind and can damage the claimant's trust in the process.
The Level 3 exclusions, bodily injury, fatality, active litigation, aren't dollar-driven at all, and I'd treat that distinction as the single most important line on this page. They're excluded because the cost of a wrong automated step there isn't financial, it's a family dealing with a claims process that got a sensitive case wrong, or a litigation exposure the insurer created for itself. That's not a boundary a claims team gets to loosen as trust in the tool grows; accuracy improving over time doesn't change what the consequence of an error looks like for a bereaved policyholder.
What genuinely earns scrutiny is the Level 2 middle tier, because "an adjuster verifies before it goes out" is only a real control if a specific adjuster is named and accountable, not a generic instruction the team nods along to. The roadmap opposite names weeks one through four for exactly this: naming the person, not just the step. Without that name attached, Level 2 reads as a policy statement rather than a working control, and I have sat through more than one post-incident review where "someone was supposed to check that" turned out to mean no one specifically was.
The fraud-referral flag inside Level 2 deserves its own comment, because it changes what the Fraud Referral Analyst role is for, not just what it does, and that distinction matters more here than the dollar thresholds. What's genuinely missing from this page, and what I'd build next for a real team, is a record of which claims actually moved between tiers over time, kept long enough that Northfield can prove the boundary is holding, not just that it was written down once in week one and never revisited.
Northfield Mutual Insurance, the Claims Processing team and its claim thresholds are invented for this sample only, to show the shape of the output, not a real client's actual policy. In the session, this fluency map is built from the use case a participant brings, not assigned from a template.
PRAXORALAB Roles & Tasks Agentic Systems Change First
The specific roles and tasks in the Claims Processing team an agentic system changes first, named against the fluency levels opposite, not left as an abstract "some roles will change" warning.
This is the first task fully inside Level 1 of the fluency gap map opposite, and it is also the task most junior adjusters currently learn the job on. Within two rollout cycles, most of this role's current day-one work is done by the assistant before a person sees the file.
A role that historically existed largely to produce this correspondence shifts toward editing and sending, not drafting from a blank page. The task moves fastest of any on this list because it is already sitting inside Level 1.
This role does not disappear, it inverts. Today the analyst is the one spotting indicators from scratch; inside Level 2, the assistant surfaces the candidates and the analyst's value moves entirely into judging which flagged claim is actually worth escalating.
The assistant can recommend a reassignment, but sign-off stays fully human, because this task touches performance management, not just workflow. Included here as the one role on this list that changes shape without shrinking, since it is the mistake most rollouts make to assume every role on a list like this one gets smaller.
Read as a set, these four roles tell a story I'd want Northfield's leadership to sit with before they announce anything to the wider claims floor, because the natural instinct with a list like this is to read it as a list of jobs at risk, and that's the wrong frame for at least half of it. The Junior Claims Adjuster and Claims Correspondent entries genuinely are about task volume shrinking fast, because both sit squarely inside Level 1 of the fluency gap map: FNOL intake and routine correspondence are exactly the tasks the team already scored well enough, on tool fluency, to run unsupervised. If Northfield does nothing else from this session, retraining those two roles toward review-and-edit work rather than draft-from-scratch work is the highest-leverage move available in the first eight weeks.
The Fraud Referral Analyst entry is the one I'd spend the most time on in the room, because it is the clearest example of a role that doesn't shrink, it inverts, and organisations consistently miss this distinction. Today the analyst's value is in noticing; inside Level 2, the assistant does the noticing and the analyst's entire value proposition becomes judging which flagged claim is actually worth escalating. That is a harder skill than the one it replaces, not an easier one, and I would be genuinely concerned about any transformation roadmap that treated this role the same way it treats the Junior Claims Adjuster role, with a training session and a new set of instructions. It needs a different kind of development: exposure to false positives, calibration against real fraud outcomes, judgement built over months, not a single afternoon.
The Claims Team Lead entry is on this list deliberately, and it's the one I'd insist on keeping even though it's the least dramatic of the four, because it is the mistake I watch rollouts make most often: leadership assumes a list of "roles that change" only ever points down the org chart. The lead's caseload-triage task changes exactly as much as the analyst's does, from doing the sorting to approving the assistant's suggested sorting, and the accountability for a bad reassignment doesn't move to the assistant just because the assistant proposed it. If HR only budgets retraining time for the two entry-level roles on this list and treats the lead's shift as automatic, Northfield will have solved the easy sixty percent of this transition and left the accountable person in the room least prepared for what actually changed under them.
The four roles above are invented for Northfield Mutual Insurance only, to show the shape of the output. In the session, this list is named against the real team a participant brings, not assigned from a generic org chart.
PRAXORALAB Sequenced Workforce Transformation Roadmap
The literacy baseline, fluency gap map and named role list above, brought together into one sequenced plan, ready to hand to HR and L&D, ordered by dependency rather than by department.
- Score the team's literacy baseline against this scorecard and lock in the fix for the priority gap identified.
- Write down, in one page, which claim types sit in each of the three fluency tiers, so the boundary stops living in individual adjusters' heads.
- Name the specific person accountable for reviewing any Level 2 draft before it goes out.
- Run a hands-on session for the Claims Correspondent and Junior Claims Adjuster roles on the specific tasks moving into Level 1.
- Set one shared prompt standard for coverage determination letters, so two adjusters stop getting two different answers from the same file.
- Retrain the Fraud Referral Analyst role around reviewing AI-surfaced referrals, not screening from scratch.
- Package the literacy baseline, fluency gap map and named role list into one roadmap document HR and L&D can act on without a facilitator in the room.
- Re-score the five literacy dimensions and compare against this baseline.
- Schedule the Claims Team Lead's caseload-triage role change with HR specifically, since it is the one role changing shape rather than shrinking.
The reason this plan runs baseline, then fluency, then handoff, and not some other order, is that each phase produces exactly one artefact the next phase can't function without, and I have watched teams try to reverse this sequence more than once. Weeks one to four exist to produce two things nothing later in the plan can substitute for: a written boundary for what sits in each fluency tier, and a named, accountable reviewer for Level 2 drafts. Skip straight to week five's training session without those, and you're training the Claims Correspondent and Junior Claims Adjuster roles against a boundary that still lives in individual habit, which means the training just teaches everyone to be confidently inconsistent instead of unconfidently inconsistent. That is not progress, it just looks like progress on a status update.
Weeks five to eight are the phase most likely to get compressed under deadline pressure, because none of its three items produce a document leadership can point to the way a finished scorecard or roadmap can. Retraining the Fraud Referral Analyst role around reviewing AI-surfaced referrals rather than screening from scratch does not show up on a Gantt chart as cleanly as "deploy the assistant to twelve more claim types" does, and I would flag that specific item as the one most likely to be quietly dropped, and also the one whose absence shows up latest and most expensively, months after go-live, when the analyst is still screening from scratch alongside a tool nobody retrained her to actually use.
Weeks nine to twelve exist because a roadmap that only lives in a facilitator's head, or in Northfield's HR team's memory of one workshop, doesn't survive staff turnover or a change in reporting line. Packaging the literacy baseline, the gap map, and the named role list into one document HR and L&D can execute against without me in the room is not a formality, it's the actual deliverable this course exists to produce; everything on the first three pages of this sample is diagnostic, and this page is where the diagnosis becomes someone else's job to run.
What "done" looks like at week twelve isn't a perfect score, it's a Claims Team Lead whose caseload-triage role change was scheduled deliberately with HR rather than assumed to happen on its own, because that is the item most rollouts I've seen skip, and the one Northfield's own version of this plan should not.
Every score, quote and figure on these four pages is invented for Northfield Mutual Insurance, a fictional company, so the format of what a participant leaves with can be judged before enquiring. It is not a real client's deliverable, and no organisation named Northfield Mutual Insurance is a Praxora Lab client. The session itself scores your own team's own use case, in the room, on the day.
3.5 hours, one session, facilitated by Sameen Khan.