Case Study · Employee Benefits Brokerage

How a tech-enabled benefits brokerage put 10 renewal-season workflows on Claude in six weeks.

LaSalle Benefits partnered with Integrated AI to deliver 10 production Claude Projects, a signed HIPAA-aware AI Governance Policy, and a two-day onsite training built around live client work. The engagement ran discovery to trained team in about six weeks, timed to land ahead of the open enrollment season.

ClientLaSalle Benefits · Northbrook, IL
EngagementFoundation build · ~6 weeks + sustainment
Delivered byBrad Weber, Integrated AI
StatusOnsite complete July 2026 · sustainment in progress
10
Production AI Workflows
One Claude Project per use case, built in LaSalle's own templates.
20 → 10
Surfaced → Built
Twenty use cases surfaced in discovery; the ten highest-value locked for the build, ten roadmapped.
40
Prompts in the Library
30 shipped at go-live; the team grew it to 40 within a week of training.
2 days
Onsite, Hands-On
July 7–8 in Northbrook. Every team member shipped real deliverables.
The Client

A firm that markets itself as tech-enabled, and wanted the operations to match.

LaSalle Benefits is a tech-enabled corporate benefits firm in Northbrook, IL, founded in 2017 by Bill Gimbel, a 30-year industry veteran and TRUE Network of Advisors member. The firm advises small-to-midsize employers across professional services, technology, manufacturing, and financial services. The team is senior and lean, and the bottleneck was clear: too many hours going into rebuilding spreadsheets, retyping forms, and hand-assembling documents every renewal cycle.

The team's earlier attempt with another AI tool failed at something as basic as changing the year from 2025 to 2026 across a booklet.
Discovery finding, AI Use Case Inventory · June 2026
IndustryEmployee Benefits Brokerage
Founded2017 · Northbrook, IL
ChampionBill Gimbel, Founder & President
NetworkTRUE Network of Advisors member
PlatformRippling benefits administration
StackClaude Team + 10 Custom Projects
The Challenge

Senior people doing template work by hand, at renewal-season volume.

The discovery session surfaced the same pattern across every workflow: the inputs already exist, the output follows a template, and a skilled person spends hours re-keying between the two.

The Q4 booklet crush

50 to 200 open enrollment booklets every fourth quarter, each built by hand off the marketing report, then reviewed by a second person, then checked a final time. Dates and rates updated throughout, one document at a time.

Renewal quotes re-keyed by hand

Five or six carrier proposals come back per RFP. Someone pulls the key figures out of each one and types them into a side-by-side comparison, choosing among four spreadsheet templates by rating type. Hours of data entry, on the numbers clients use to decide.

Compliance sent generic

Post-renewal compliance notices and Medicare Part D notices went out uncustomized, because customizing each one by hand was never feasible. A required deliverable, delivered at the minimum, purely for lack of time.

The Approach

Discovery, governance, build, train, sustain. In that order.

Most AI rollouts start with the tool and hope the workflows follow. This engagement inverted that: map the team's actual week first, put the data rules in writing before anything is built, then train on live client work.

Step 01 · Discovery · June 8
One structured session, twenty use cases surfaced
A 60-minute working session walked the team's day-to-day, task by task, in their own words. Twenty candidate workflows surfaced. Ranked together by time saved and ease of build, the team locked ten for the engagement and kept the other ten as a roadmap.
Step 02 · Governance · June
HIPAA-aware AI Governance Policy, signed before training
A four-tier data classification with one core rule: no PHI or identifiable member data enters the Claude Team account, because the plan carries no Business Associate Agreement. Real member data is de-identified or replaced with synthetic data first, and a per-use-case data guidance table tells the team exactly what each project may touch.
Step 03 · Build · June–July
Ten Claude Projects, built in LaSalle's own templates
Every project was loaded with LaSalle's actual source materials: the age-rated and composite rate workbooks, the benefit guide template, real (sanitized) worked examples, carrier enrollment templates, and the firm's branding. Each shipped with custom instructions, a copy-and-paste prompt set, and a one-page quick reference card.
Step 04 · Onsite Training · July 7–8
Two days in Northbrook, on the team's real client work
All ten use cases demoed live. Two extended hands-on labs where every team member built deliverables from their own active workload. Day 2 closed with the Beyond the Prompt track: connectors, custom tools, automations, skills, and delegation.
Step 05 · Sustainment · July–September
Follow-through booked before the trainer left the room
Recap email at day +3, office hours at week +1 and week +3, a 30-day metrics review with Bill, and a 60-day review to decide on the next phase, drawing from the ten roadmapped use cases. Phase 2 stays earned, never promised in advance.
Inside the Onsite

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

Two days, 8:30 to 4:00, the full LaSalle team. Every session was either a live demo on real materials or a hands-on lab on the team's own client files. Homework between the days: bring one real deliverable from your current workload to build on Day 2.

Day 1 · Foundation & Highest-Impact Use Cases

Governance, fundamentals, and the renewal-season workhorses.

Day 1 built confidence in order: the rules first, then the tool, then the six use cases the team touches most. By end of day, every team member had run a morning use case on their own live client work in Lab #1.

Sessions Run

  • Welcome: Bill on team priorities, Brad on the 10 use cases and why these ten
  • AI Governance Policy walkthrough: data tiers, human review, what never goes in, policy signing
  • Claude Fundamentals: workspace tour, Projects vs. chats, the 5-part prompt, iteration and recovery
  • Live Demos, Round 1: Carrier Quote Comparison, Benefit Guides, OE Presentations
  • Hands-On Lab #1 (90 min): everyone runs a morning use case on live client work
  • Live Demos, Round 2: Illustration Document, Census to Enrollment Forms, Combined Enrollment Forms
  • Debrief and homework: pick the real deliverable you'll build tomorrow

Documents Delivered

  • AI Governance PolicyFour-tier classification, de-identification standard, per-use-case data rules. Signed by all attendees.
  • AI Use Case InventoryAll 20 discovery use cases documented: how it runs today, how Claude helps, the payoff.
  • Claude Fundamentals DeckProjects vs. chats, the 5-part prompt, iterate-in-the-same-chat, recovery patterns.
  • Prompt Template Library30 copy-and-paste prompts across the 10 use cases, live in the shared drive.
  • Quick Reference CardsOne card per use case: inputs, steps, checks before send.
  • 10 Configured Claude ProjectsCustom instructions, source materials, and data rules loaded per use case before Day 1.
Day 2 · Real Work & Beyond the Prompt

“Yesterday you got fluent. Today you get force-multiplied.”

Day 2 finished the demo set, then gave the team almost three hours to build and share real deliverables from their own queues. The afternoon ran the Beyond the Prompt track from the 37-slide master deck: Use, Extend, Own.

Sessions Run

  • Live Demos, Round 3: Rippling Population, Contribution & Deduction Sheets, Compliance Notices, Medicare Part D
  • Hands-On Lab #2 (165 min total): every member builds one real deliverable from homework, then shows the result
  • Connectors: a scoped pipe to calendar, email, and Drive, with a safety check before every connection. No ben admin connector: PHI and BAA out of scope
  • Building Tools: stop asking, start commissioning. One paragraph in, a working benefits tool out
  • Automations: a prompt plus a schedule plus a delivery channel; Monday-morning prep that runs without you
  • Plugins & Skills: install two ready-made time-savers; build a LaSalle house-voice skill
  • Delegation: goal, access, guardrails, check-in protocol
  • Bill closes: team commitments and the first week back at the desk

Documents Delivered

  • Day 2 Master Deck: “Beyond the Prompt”37 slides, five blocks, every block ending with a live build on the team's own workspace.
  • Connector Safety FrameworkThe safety check run before every connection, printed as a handout.
  • PII Masking ToolkitOne-click census masking and carrier quote sanitizing, with a local token key so real files never enter Claude un-masked and outputs can be restored afterward.
  • Onsite Training AgendaThe full two-day agenda: every session, demo, lab, and break.
  • Sustainment Plan & Success CriteriaOffice hours weeks +1 and +3, 30-day metrics review, 60-day Phase 2 decision, quarterly policy review.
What We Built

10 Claude Projects, one per workflow, each with its own guardrails.

Tap a card for what the project does and the data rule it enforces. Every project ships with custom instructions, LaSalle's own templates and worked examples, a prompt set, and a quick reference card. A person reviews every output before it reaches a client.

Before & After

Three workflows, end-to-end.

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

Before
Hours of hand-keying per RFP
  • Quotes come back from five or six carriers per RFP
  • Key figures pulled out of each proposal by hand
  • Typed into one of four spreadsheet templates, chosen by rating type
  • Transcription risk on the exact numbers clients use to decide
After
A drafted comparison to review and spot-check
  • All carrier proposals drop into the project alongside the templates
  • Claude extracts benefits and rates from every proposal
  • Detects age-rated vs. composite-rated and picks the correct template
  • Side-by-side comparison and renewal analysis, in LaSalle's format
  • Advisor spot-checks rates against source before send
Before
50–200 booklets per Q4, each built by hand
  • Every booklet built off the marketing report, dates and rates updated throughout
  • One person builds, a second reviews, a final copy check after that
  • The team's biggest time sink of the year, concentrated in one quarter
  • A prior AI tool failed at rolling the plan year from 2025 to 2026
After
A populated first draft per client
  • The project holds LaSalle's booklet template and pulls plan data from SBCs and plan documents
  • First draft generated per client with dates, rates, and plan details already placed
  • Same data feeds the matching OE presentation, no second manual build
  • Review passes shift from re-keying to verifying
Before
Generic notices, because there was never time
  • Post-renewal notices sent as a generic document of links, narrowed to carriers but not to the group
  • Medicare Part D notices sent uncustomized for the same reason
  • A required deliverable, delivered at the minimum
After
Customized per group, no added manual time
  • Notice content and links live in the project
  • Claude assembles a customized notice set per group, only the relevant carriers
  • Medicare Part D notice customized to the group and its coverage
  • A quality upgrade on a compliance deliverable that used to go out generic
By the Numbers

The onsite just wrapped. Measured outcomes land at the 30 and 60-day reviews.

Training finished July 8. Rather than projecting hours saved, this engagement measures them: a 30-day metrics review with Bill covers hours saved, adoption, and incidents, and a 60-day review decides the next phase. The success criteria below were set before the trainer left the room.

30-Day Success Criteria

Agreed at the onsite · reviewed with Bill at week +4.
  • 100% of the team has logged into Claude Team and used at least 3 use cases
  • At least 2 fully AI-assisted deliverables completed per team member
  • Zero data classification incidents
  • Measurable time savings reported on the top 3 use cases
  • Team has proposed at least 2 improvements to the prompt library Already met: the team contributed 3 new workflow prompts within a week of training.

Engagement Snapshot

Delivered scope, discovery to sustainment.
  • Use cases surfaced in discovery20
  • Production Claude Projects built10
  • Use cases roadmapped for a future phase10
  • Prompts in the shared library40 and growing
  • Quick reference cards10
  • Days of onsite delivery2
  • Full team trained, hands-on100%
  • PHI entered into ClaudeNone, by design
Note: Measured time savings replace this snapshot's scope metrics at the 30 and 60-day reviews. No projections are shown here on purpose: the numbers on this page will be the team's real ones.
Voices from the Team

Verbatim quotes arrive with the 30-day review.

The onsite wrapped July 8. Real testimonials from Bill and the LaSalle team will be captured at the 30-day metrics review and added here.

Pending · 30-Day Review

Quote collection in progress

The early signal is already visible: within a week of training, the team was extending the prompt library on their own, adding three new workflow prompts from their live renewal work. The 30-day review will ask:

"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.

Including the one most vendors skip: telling a benefits firm what AI must not touch, before showing what it can.

01 · Discovery

Discovery first, tools second.

Twenty use cases surfaced from the team's own words, ranked together by time saved and ease of build. The ten that got built were the team's picks, not a generic playbook.

02 · Governance

Honest about the PHI line.

The Claude Team plan carries no BAA, so the policy says so plainly: no PHI, no identifiable member data, de-identify or go synthetic. Signed before training. A per-use-case table makes the rule concrete.

03 · Tooling

Masking built in, not bolted on.

A PII masking toolkit ships with the engagement: one-click census masking and quote sanitizing with a local token key, so following the policy is the easy path, not the slow one.

04 · Templates

Built in LaSalle's own formats.

Every project produces output in the firm's existing templates: the rate workbooks, the benefit guide design, the illustration shell, the carrier forms. Nothing to reformat, nothing to re-learn.

05 · Onsite

Trained on live client work.

Two days in Northbrook with homework in between: bring a real deliverable from your queue. Every team member left having shipped work, not having watched demos.

06 · Sustainment

Measured, then earned.

Office hours at weeks +1 and +3, a 30-day metrics review, a 60-day decision on the next phase drawn from the ten roadmapped use cases. No Phase 2 promised until Phase 1 proves out.

Run This Playbook on Your Firm

If renewal season runs on re-keying, this is the engagement that gets your team's hours back.

For employee benefits brokerages and the professional services firms that look like them: where the people you trust most spend their busiest quarter on work a template already knows how to do.