Case Study · Group Benefits Brokerage

How a 5-person benefits team built an AI-powered back office in 30 days.

AFC-AIS partnered with Integrated AI to deliver 11 production Claude Projects, a signed AI Governance Policy, and a two-day onsite training arc, turning hours of template-driven busywork into minutes of advisor-reviewed AI output.

ClientAFC-AIS Group Benefits
EngagementPhase 1 Foundation · 30 days
Delivered byBrad Weber, Integrated AI
StatusPhase 1 complete · April 2026
11
Production AI Workflows
One Claude Project per use case, each with a custom system prompt.
7–13 hrs/wk
Time Recovered
Conservative range across the 5–6 person team.
$32.5K
Annualized Capacity
Upper-band recovered capacity, conservative projection.
<9 wks
Phase 1 Payback
From signed governance to measured ROI window.
The Client

A small benefits team, a real book, and no margin for repeatable manual work.

Associated Financial Consultants & Investor Services (AFC-AIS) advises employers on Group Benefits: medical, dental, vision, life, disability, and a long list of voluntary products. Their advisors are senior. Their book is real. Their bottleneck was the time spent on format-heavy, low-leverage work.

If I could get two hours back every week, I'd use it for actual client work.
AFC-AIS team member, Discovery interview · February 2026
IndustryEmployee Benefits Brokerage
Team Size5–6 advisors & account managers
Carrier MixUHC · Humana · Cigna · Aetna
ChampionLogan Mondshein, Principal
StackClaude Team + 11 Custom Projects
EngagementPhase 1 Foundation (30 days)
The Challenge

High-skill people doing low-leverage, format-heavy work.

Discovery surfaced the same root issue across every conversation, and the same downstream cost: never enough capacity for the higher-margin advisory work that clients actually pay for.

Template-driven busywork

Benefits guides, Benefit Zone templates, OE handouts, and contribution sheets, all populated by hand from carrier proposals every renewal cycle. A single benefits guide could eat 2–3 hours.

High-stakes data reconciliation

Census-to-bill matching across multiple carriers each month. Errors mean clients overpay; missed catches erode trust. Tedious, error-prone, and impossible to skip.

Compliance overhead with no upside

Compliance reminders, plain-English guidance on ERISA, ACA, COBRA, PCORI. All necessary, none of it billable, all of it pulling advisors out of client conversations.

The Approach

Pilot first, scale later: discovery, governance, training, then sustainment.

Most AI rollouts fail because they start with shiny agents before the team has a working relationship with the tool. the Integrated AI method inverts that: build the foundation, train on real client work, and measure before scaling.

Step 01 · Discovery
Two structured discovery sessions, end-to-end workflow mapping
Walked the team through their actual week. The 11 use cases, the data classifications, and the priority order all came from the team, not a generic playbook.
Step 02 · Governance
AI Governance Policy signed before training day
Green/Yellow/Orange/Red data classification with HIPAA-adjacent rules. Signed by every attendee. Encoded into every Project's system prompt.
Step 03 · Onsite Training
Two days, in person, on the team's real client work
Day 1 ran the full agenda: governance, fundamentals, every demo, lightning tour of all 11 use cases, and two hands-on labs on live client files. Day 2 went Beyond the Prompt: connectors, custom tool building, scheduled automations, plugins & skills, and delegation, each block ending with a real build on the team's own workspace.
Step 04 · Sustainment
Office hours scheduled before the trainer left the room
Recap email day +3, office hours week +1 and week +3, 30-day metrics review, 60-day Phase 2 decision. Phase 2 (AI agents) stays earned, never promised in advance.
Inside the Onsite

Day 1 and Day 2: what we ran and what we delivered.

Two days, six attendees, every team member building real deliverables from their own active client list. Each day shipped its own complete document pack into the team's shared workspace.

Day 1 · Foundation & the 11 Use Cases

Get fluent with Claude, then run all 11 use cases on real client work.

Day 1 ran the entire onsite training agenda end-to-end: governance, fundamentals, every demo, and two hands-on labs on the team's actual client files. By 5pm, every AM had used Claude on live work and shipped at least one deliverable from their own queue.

Sessions Run

  • Welcome & "why we're doing this": Logan and Brad set the bar
  • AI Governance Policy walkthrough: Green/Yellow/Orange/Red, HIPAA-adjacent rules
  • Claude Fundamentals: workspace, Projects vs. chats, the 5-part prompt, recovery
  • Live Demo #1: Carrier Quote Comparisons (Use Case 1)
  • Live Demo #2: Census-to-Bill Reconciliation (Use Case 2)
  • Live Demo #3: Benefits Guidebooks (Use Case 3)
  • Live Demo #4: Contribution Comparisons & Deduction Sheets (Use Case 4)
  • Live Demo #5: Open Enrollment Communications (Use Case 5)
  • Lightning tour: Use Cases 6, 7, 8, 9, 10, 11
  • Hands-On Lab #1: top-priority workflows on live client files
  • Hands-On Lab #2: every member ships a real deliverable from their own queue

Documents Delivered

  • AI Governance PolicySigned before Day 1 starts; Green/Yellow/Orange/Red classification + incident response.
  • Onsite Training AgendaThe full agenda: every session, demo, lab, and break, run end-to-end.
  • Claude Team Setup GuideWorkspace provisioning and access for all six seats.
  • Claude Fundamentals DeckProjects vs. chats, the 5-part prompt, iterate-in-the-same-chat, recovery patterns.
  • Prompt Template Library30+ reusable prompts mapped to the 11 use cases.
  • Quick-Reference CardsOne-page card per use case, designed for laminated desk reference.
  • 11 Project System PromptsOne per use case: role, voice, data rules, output standards, QA reminders.
Day 2 · Beyond the Prompt

"Yesterday you got fluent. Today you get force-multiplied."

Day 2 ran the Beyond the Prompt master deck: five blocks, each ending with the team building something real on their own workspace. Every block followed the same rhythm: short walkthrough, live build, hands-on exercise, recap. Outcome: each AM shipped one live build before end of day.

Sessions Run · Five Blocks

  • Frame: Use → Extend → Own: the three levels of working with AI
  • Block 01 · Connectors: each AM connects 2 of 3: Calendar · Gmail drafts · Drive folder; Connector Safety Check used by every AM, every time
  • Block 02 · Building Tools: "Stop asking. Start commissioning." One paragraph in, a working benefits tool out: Plan Compass OE decision tool built live
  • Block 03 · Automations: Monday Morning Prep + Friday Renewal Scan: a prompt + a schedule + a delivery channel, scheduled end-to-end
  • Block 04 · Plugins & Skills: install 2 ready-made skills that save time today; build the afc-voice house skill that enforces AFC-AIS tone in every draft
  • Block 05 · Delegation: hand off multi-step work with goal, access, guardrails, and check-in protocol: "Prep 5 Q3 renewals for Logan's Friday review"
  • Sustainment plan and 30/60/90 commitments, closed by Logan

Documents Delivered

  • Day 2 Master Deck: "Beyond the Prompt"37 slides, five blocks, every block ending with a hands-on build.
  • Connector Safety FrameworkFive-question safety check for every connector. No AMS connector this quarter. PHI/BAA out of scope.
  • Plan Compass & Tool-Build PromptsThe exact prompts the team used to commission working OE decision tools from a single paragraph of intent.
  • Automation Specs: Monday Prep & Friday Renewal ScanScheduled tasks the team set up live. Four common silent-failure modes documented.
  • afc-voice House SkillBuilt live in front of the team. Encodes the AFC-AIS tone so it ships on every draft without re-typing the rules.
  • Delegation PlaybookGoal, access, guardrails, check-in protocol: the four pieces every delegated workflow needs.
  • 30/60/90 Sustainment PlanOffice hours week +1 and +3, 30-day metrics review, 60-day Phase 2 decision gate.
What We Built

11 Claude Projects, one purpose-built system prompt each.

Tap a card for the role, the guardrails, and the standard output. Every project ships with voice rules, data-handling guardrails, output standards, and QA reminders baked into the system prompt.

Before & After

Three workflows, end-to-end.

Switch tabs to see how the team's three highest-volume workflows changed shape after Phase 1.

Before
3–4 hours per renewal
  • Open each carrier PDF and hand-key plan details into the comparison sheet
  • Manually flag rate changes; write narrative commentary from scratch
  • Reformat the table for client presentation
  • Format and depth varied across team members
After
~60 minutes: review, refine, send
  • Drop carrier PDFs into the project; Claude returns a normalized table
  • Meaningful design gaps flagged with explanation, not just diffs
  • Renewal narrative drafted in client voice, ready for advisor edit
  • QA reminder forces a 3-rate spot-check before send
  • Output format identical across the team
Before
2+ hours per client per month
  • VLOOKUP gymnastics in Excel between census and invoice
  • No standard categorization for discrepancy types
  • Carrier follow-up emails written from scratch each time
  • Pattern recognition across months relied on memory
After
~20 minutes: verify, send, log
  • Census & invoice in, categorized discrepancy report out
  • Term-removals, missing enrollments, tier and rate mismatches separated
  • Pre-drafted carrier emails reference invoice number, period, deadline
  • Multi-month pattern analysis names the recurring offenders
  • SSN/DOB hard-stop enforced on every input
Before
1–2 days per client cycle
  • Each email written from scratch in the client's tone
  • FAQ documents recreated per client with copy/paste from prior year
  • Spanish translation outsourced or skipped entirely
  • Bandwidth meant some clients got the full campaign and some didn't
After
~2 hours: refine and schedule
  • Five-email sequence delivered as one batch
  • 1-page FAQ generated per client with current dates and platform link
  • Latin American Spanish translation included by default
  • Print-ready break-room flyer in the same workflow
  • Every client gets the full campaign, regardless of team load
By the Numbers

Conservative projections from the proposal, to be replaced with measured outcomes at Day 30, 60, and 90.

Phase 1 just wrapped. The numbers below are the conservative impact projections in the AFC-AIS proposal. They will be replaced with measured outcomes captured at the 30, 60, and 90-day reviews.

Estimated Weekly Time Recovered

Hours per week, by workflow, low/high range.

Engagement ROI Snapshot

Annualized capacity recovered vs. Phase 1 investment.
  • Production AI workflows shipped11
  • Reusable prompts in shared library30+
  • Team trained, hands-on, onsite6 / 6
  • Days of onsite delivery2
  • Conservative weekly capacity7 hrs/wk
  • Upper-band weekly capacity13 hrs/wk
  • Annualized capacity (low)$17,500 / yr
  • Annualized capacity (high)$32,500 / yr
  • Phase 1 payback period< 9 weeks
Note: These are the conservative impact projections from the AFC-AIS proposal. Measured numbers replace projections at the 30, 60, and 90-day reviews.
Voices from the Team

Verbatim quotes captured at Day 30 office hours.

Phase 1 just wrapped. Real testimonials from Logan and the AFC-AIS team will be captured at the first 30-day office hours review and added here.

Pending · Day 30 Office Hours

Quote collection in progress

The first office-hours session post-onsite is the moment to ask the team what changed. The two prompts below have produced the highest-signal quotes in past engagements:

"What's the most time you've gotten back in a single week so far?"
"What's one thing you used to dread that's now no big deal?"
What Made This Engagement Different

Six choices that separated this rollout from the median AI pilot.

Boutique agility, enterprise expertise, and a method designed to compound after the trainer leaves the room.

01 · Discovery

Discovery first, tools second.

Two structured discovery sessions named the actual workflows before a single project was built. The 11 use cases came from the team, not from a generic playbook.

02 · Governance

Governance on day one.

AI Governance Policy with HIPAA-adjacent data rules signed before training started. Every Project enforces it in its system prompt.

03 · Onsite

Onsite, on real client work.

Two days, in person, every team member building real deliverables from their own active client list. The team left having shipped.

04 · System Prompts

System prompts, not just chats.

Each Project has a custom system prompt encoding role, voice, data rules, output standards, and QA reminders. Disciplined output for every team member.

05 · Sustainment

Sustainment scheduled, not promised.

Office hours week +1 and week +3. Metrics review at Day 30. Phase 2 decision at Day 60. Follow-through booked before the trainer left the room.

06 · Phase 2

Phase 2 stays earned.

No agents promised until Phase 1 ROI is measured. The team chooses Phase 2 scope from a menu of validated workflows, not from a sales pitch.

Run This Playbook on Your Firm

If your back office is the bottleneck on growth, this is the engagement that gets your team's week back.

For mid-market professional services teams (benefits, financial advisory, accounting, insurance, legal) where the people you trust most are spending their time on work the firm shouldn't be paying them to do.