How a Chicago logistics platform turned three Claude power users into a shared org memory

A freight-tech company's team had adopted Claude organically until every dashboard in every meeting was a Claude document. The knowledge lived in individual chat histories. A three-person pilot moved it into shared memory: customer analysis dropped from hours to minutes and new-PM ramp time halved, at lower monthly spend.

12 min
median customer-specific analysis, down from hours
9 days
new-PM ramp to parity, down from 4-6 weeks
27
private artifacts moved into shared memory
-26%
monthly AI spend at higher usage

The situation

Nobody had to push this team onto AI. Over a year, adoption happened by itself: any dashboard projected in a meeting was a Claude document, and the old BI tool quietly stopped being the source of truth. The analytics team had even built a shared skill over forty curated warehouse tables, documenting KPIs, context, and jargon.

The gap was everything customer-specific. How to analyze one particular shipper account lived in one PM's head, and when that PM was out, the analysis waited. Dashboards PMs built for their own areas were private by default. Every week the org worked this way, individual memories compounded faster than the shared one did.

The turn

The Head of Product chose a deliberately small start: three power users, running Second Axis in parallel with their existing Claude subscriptions so the difference could be felt side by side. Onboarding captured each user's customer-analysis heuristics as explicit workflows without asking anyone to write documentation, and every correction the pilot users made taught the workflows their preferences.

By the end of month one, 27 previously private artifacts, dashboards, account playbooks, and analysis recipes, lived in shared memory. The test came when the team's senior PM took a two-week holiday: a shipper review that would previously have waited ran in twelve minutes off her captured workflow, by a PM who had never analyzed that account before.

Rather than migrating the whole organization, I picked three power users and we set it up. Four weeks later the question flipped from whether we roll it out to how fast we can.
Head of Product, Analytics & Design

How Second Axis fits

01

Same usage patterns, different memory

Daily behavior did not change. What changed is where it accumulates: every session compounds into shared org memory instead of a personal chat history.

02

Heuristics lifted into runnable workflows

Customer-analysis knowledge locked in one PM's head became workflows anyone can run, refined by the team's own corrections week over week.

03

Automated model routing

Routine tasks stopped burning premium-model tokens. Routing alone cut monthly spend 26% while usage grew.

04

Parallel run, then a decision

Nobody gave up Claude during the pilot. The team measured the compounding difference directly before switching anything off.

Results

AreaBeforeAfter
Customer-specific analysis timeHours, single-owner12 min median, any PM
New-PM ramp to full analysis parity4-6 weeks9 days
PM artifacts available to the whole team3 shared27 in shared memory
Monthly AI spend$4,200$3,100 at higher usage
Cross-PM knowledge transferAd hoc, manager-pushedContinuous, by default

What came next

Week 0Three power users selected. Pilot runs in parallel with existing Claude seats.
Week 1Onboarding maps daily patterns and customer heuristics into shared workflows.
Week 427 private artifacts in shared memory. Holiday-cover analysis runs in 12 minutes.
Month 2Side-by-side review: shared-memory workflow beats parallel Claude runs on speed and reuse.
Month 3Full team rollout approved. Individual Claude licenses wound down in phases.

The rollout decision took one review meeting, because the pilot had already produced the comparison the team needed. The org kept the analytics team's beloved warehouse skill, now versioned and shared inside the workspace it should have lived in all along.

Company details anonymized at the customer's request.