Most remote onboarding fails the same way. Week one is a scramble of logins and welcome calls. Week two the new person shadows whoever happens to have time. By week three nobody can say which tasks they own, and the account manager who was supposed to get relief is doing the same work plus answering questions about it.
The cause is almost never the hire. It is that onboarding got scheduled as a series of meetings instead of designed as a series of handoffs, each with an owner, an artifact, and a test that can be passed or failed.
Below is a 30-day plan an operations manager can run as written. It assumes one new remote team member, one internal owner with authority to hand work over, and four to six hours per week of a senior person's time in weeks one and two, tapering after that. Examples use insurance servicing and general customer support, but the structure holds for any back-office role.
Define "done" before day one
Write the finish line down before the start date. Three artifacts, one page total:
- The workflow list. Three to five named workflows the new person owns by day 30, named the way your team names them: certificate of insurance requests off a standard template, renewal document collection, loss run requests, refund-status replies, carrier download entry.
- The quality bar. What "correct" means per workflow, in checkable terms. Right form, right holder name, right effective dates, note logged in the agency management system, client informed the same business day.
- The escalation rule. What they never decide alone: coverage changes, binding, premium adjustments, refunds above a stated threshold, anything a client is unhappy about.
If you cannot fill in the workflow list, onboarding is not the problem you are solving yet. Pick the work first.
The 30-day plan at a glance
| Week | Focus | Internal owner does | Exit test |
|---|---|---|---|
| 1 | Access and context | Grants tier 0 and tier 1 access, records 6 to 10 SOP videos | Runs one workflow end to end on a test account, from the SOP alone |
| 2 | Shadowing both directions | Narrates live work, then watches and corrects | Ten completed items, no correctness errors, all notes logged |
| 3 | First owned workflow | Reviews every item before it leaves, logs error types | Volume target hit three days running at the quality bar |
| 4 | QA sampling and scope | Drops to sampled review, scores the scorecard | 30-day review with a documented decision on scope |
Print it. Put dates on it. An onboarding plan without exit tests is a calendar, not a plan.
Week 1: access in tiers, and SOPs captured from real work
Grant access in tiers, not all at once
Least-privilege setup is faster than the alternative, because it settles in advance whether a new person can send, submit, or bind anything in week one. They cannot. Structure access as tiers and move up on evidence.
- Tier 0, read only. Read access to the agency management system or CRM (AMS360, Applied Epic, EZLynx, HawkSoft, NowCerts, QQCatalyst, Salesforce, HubSpot), view-only shared mailbox, internal wiki, SOP library.
- Tier 1, draft but do not send. Save drafts, work only tickets assigned to them in Zendesk, Gorgias, or Freshdesk, no bulk actions, no public replies without review.
- Tier 2, send on owned workflows. Named individual logins for the carrier portals those workflows touch. Never shared credentials, never a mailbox password passed around in chat.
- Tier 3, restricted permanently. Payments, binding, commission records, bulk export of client data. Most remote roles never need tier 3, and saying so up front prevents a slow drift into it.
Two habits make this stick. Write the revocation steps on the same day you write the grant, in the same document, so offboarding is a checklist and not an investigation. And keep one access register listing person, system, permission level, date granted, and approver. Signed NDAs, a locked-down workstation, and single sign-on where your stack supports it take an afternoon.
Capture SOPs by recording the work, not describing it
The fastest SOP is a five to eight minute screen recording of a senior person doing the real task, narrated as they go, plus a one-page written summary with the exception rules. Descriptions written from memory skip the parts that cause errors: which template, which portal tab, where the note goes, what to do when the holder name does not match the named insured.
A worked example for planning, with assumptions you should replace with your own: three workflows, three recordings each covering the standard path and the two most common exceptions, about 20 minutes of senior time per recording including review. That is near three hours of internal effort in week one, and it produces the reference library every later week depends on.
If the SOP does not exist before the start date, the first month gets spent building it under time pressure, by the person least qualified to build it. Capture the recordings during the notice period, not during week one.
Week 2: shadowing that produces a paper trail
Split the week. Days 6 and 7, the new person watches the internal owner work and takes notes in the SOP itself. Days 8 through 10, the direction reverses: the new person drives, the owner watches quietly, and corrections go into a shared log instead of into the moment.
Three rules keep this from becoming passive video-call time:
- Every session ends with an SOP edit. If nothing changed, the session was not close enough to real work.
- Every question goes in one log, with the answer, the date, and whether it became an SOP edit. A question asked twice is a documentation defect, not a training defect.
- Documentation gets scored from day 6. A correct outcome with no note logged counts as an error. Insurance servicing is judged on the record as much as the result, and that habit is easier to build than to retrofit.
By the end of week 2, aim for ten completed items per target workflow with no correctness errors. If errors still cluster in one workflow, hold it back and let the others advance. Onboarding does not have to move at one speed.
Week 3: the first owned workflow
Pick the workflow with the highest volume and the smallest blast radius. In agency servicing that usually means certificates of insurance off an existing, unchanged policy, evidence of property requests, loss run requests, and renewal document chasing, not endorsements that change coverage, cancellations, or first notice of loss intake. In general support it means order status, shipping exceptions, and refund-status replies rather than refund approvals or account credits.
Ramp volume deliberately. As an illustrative model, assume an experienced person completes the workflow in about 10 minutes and the new person starts at roughly three times that. A defensible ramp is 30 percent of a day's queue on day 11, 60 percent by day 13, and the full queue by day 15, with handling time near 12 to 14 minutes by day 30. Substitute your own baseline; the shape matters more than the figures.
Review every item before it leaves until two consecutive days come back clean, then move to sampling. Log error types rather than error counts, in four buckets: wrong data, missing documentation, wrong process step, and tone or clarity. The buckets tell you whether to fix the SOP, the training, or the access.
Week 4: QA scorecards and the 30-day review
By week 4 the question shifts from "can they do it" to "can you see whether they did it." That needs a scorecard applied to a sample on a fixed cadence, with recorded results the new person can read.
| Dimension | Illustrative weight | What is checked | Sample |
|---|---|---|---|
| Accuracy | 40 percent | Correct data, correct form, correct dates and parties | 10 items per week |
| Documentation | 25 percent | Note logged, activity closed, attachments filed correctly | Same 10 items |
| Responsiveness | 20 percent | Handled inside the agreed window, no item aging silently | Full queue report |
| Communication | 15 percent | Clear, correct, on-brand client and internal messages | 5 messages per week |
Those weights are a starting point, not a standard. If your complaints are mostly about response time, weight responsiveness higher. Publish whatever you choose before scoring against it.
Run the 30-day review as a 45-minute working session with four items: the scorecard trend across weeks 3 and 4, the error-type breakdown, the question log converted into remaining SOP gaps, and one decision. The decision has three options. Expand scope and add the next workflow. Hold scope for two weeks and fix a named gap. Or correct course with a documented plan and a follow-up date. Ambiguity here is what turns a good hire into a stalled one.
What breaks this plan
- No internal owner with authority. If the person hosting the shadowing cannot decide to hand work over, week 3 never starts.
- All access on day one. Tier 3 permissions granted early get used early, and the first mistake is expensive rather than instructive.
- Training on the exception first. Teach the standard path to fluency, then the exceptions. Reversing that order produces hesitation on everything.
- Feedback that lives only in chat. If corrections are not scored and recorded, week 4 has no trend and the review becomes an opinion.
- A workflow list of one. A single owned workflow leaves the new person idle when volume dips and gives you nothing to compare against.
Where to start
- Write the one-page definition of done this week: three to five named workflows, the quality bar for each, and the escalation rule.
- Book two hours with your strongest internal person and record the first six SOP videos on live work, before anyone new starts.
- Build the access register and tier map for your stack, including revocation steps for every system.
- Put the four exit tests on a dated calendar with a named internal owner for each one.
- Choose the week 3 workflow now and confirm it is both high volume and low risk. If it is not both, pick a different one.
Run this once and the SOP library and the scorecard are already built for the next hire. That reuse is the part most teams skip, and it is what makes the second onboarding faster than the first.