Warehouse Logistics
Warehouse Operator
$7,500/year → ~$120/year for floor doc access: 50 M365 read-only seats replaced by an internal playbook portal
Challenge
A warehouse logistics operator with roughly 160 employees runs on playbooks, SOPs, and floor procedures across three warehouse sites. When carriers, layouts, or handling rules change, ops managers know what changed on the floor. The hard part was getting that operational change into a procedure the whole company could trust, on every site, without turning the manager into a part-time document librarian.
Procedures lived in Microsoft 365 across several document clusters with no single library. The team counted 500+ files; many were outdated. Nobody trusted the map. People were afraid to edit or delete in case they broke the official copy. The same procedure often existed in different versions on different sites, with no publish path the floor could rely on. Updating a playbook meant finding the right file, guessing which copy was official, and editing inside a system built for email and Office. Warehouse process change control was missing.
Anyone who only needed to read a procedure still needed a paid Microsoft 365 seat. Drivers, pickers, and floor staff who never used email or Teams still required Business Standard licenses just to open a picking or safety SOP on a phone. IT maintained 50 read-only seats for that purpose alone: $7,500 per year at list price on annual commit (50 × $12.50 × 12).
Updates moved slowly. A playbook change could take about a week from request to something floor staff could rely on. Document owners are warehouse managers, ops leads, HSE staff, and training coordinators. Their day job is running the warehouse. Folder trees, version history, and SharePoint permissions sat on top of that work. The old setup forced them to carry both.
Approach
Two problems drove the project: read-only M365 licenses for floor staff, and the same SOP living in different versions on three sites.
We shipped the AI Playbook Agent: custom repo rules, workflow logic, and Slack as the chat surface (Claude from Anthropic handles the LLM calls). A version-controlled repo holds the master copy with full change history. An internal portal publishes what floor staff read, at $10 per month hosting.
Repo rules define topics (inbound, outbound, HSE, training, equipment), naming, cross-links, tone, and required sections. A zone owner describes a change in Slack; the agent finds the right files, applies the edit, flags linked procedures that may need updates, and opens a pull request with a short summary. A stakeholder group approves yes or no. After merge, the portal publishes one version to all three warehouses.
Topic boundaries are enforced on every edit. An outbound lead can update picking and labeling procedures; HSE or training files are left alone unless the change requires a cascade. Training checklists update through cascade when ops changes require it.
Managers can ask the agent what changed over a date range. It answers from repo history and chat context, with section references. Change history lives in the repo and chat; a separate audit tool was unnecessary.
Solution
Floor staff read the portal. Zone owners edit through Slack. Nobody on the warehouse floor needs a Microsoft 365 license to open a procedure.
A typical update starts when an ops or warehouse manager notices a playbook is wrong after a process change. They message the AI Playbook Agent in Slack and describe the change. The agent clarifies zone, role, or equipment if needed, patches the correct topic files in the repo, and opens a PR. The stakeholder group reviews the diff summary and approves or rejects. On merge, the portal publishes to all sites.
When a change affects safety or onboarding, linked HSE and training documents update in the same flow. Everyone reads one portal URL. Owners no longer hunt through scattered M365 clusters or overwrite another team's checklist by mistake.
Stack: AI Playbook Agent, Slack, Claude (Anthropic, LLM layer), version-controlled Git repo, internal portal with company authentication.
Results
Annual cost for floor document access dropped from $7,500 per year (50 Microsoft 365 Business Standard read-only seats at list price) to roughly $120 per year ($10 per month portal hosting). Fifty M365 seats used only for SOP access went to zero.
Playbook update cycle time went from about one week to 30 minutes for typical procedure changes submitted through Slack and approved via PR.
The 500+ document sprawl shrank by roughly 70% after retiring stale duplicates. Three warehouse sites now read the same published version on the portal.
Zone owners edit inbound, outbound, HSE, training, and equipment topics through Slack without touching the repo directly.
Four weeks from kick-off to production.
Learnings
Warehouse managers describe operational changes in Slack. The AI Playbook Agent maintains the documentation: repo structure, file placement, cascade, and PR.
Slack worked as the edit surface because it matched how ops already talks about process changes. The agent maps chat to files; owners never browse a repo or pick a folder. Asking owners to "manage the knowledge base" would have stalled adoption.
Governance mattered more than generation. The calendar bottleneck was ownership and publish rules: who owns which topic, what counts as a valid change, and who can approve a publish. Roughly half the project calendar was stakeholder work: walking zone owners through boundaries, resolving conflicts between sites, and aligning on what "current" means for a given procedure.
During consolidation, deliberate ambiguity surfaced in the old library. Some documents were written loosely on purpose: vague steps that let a site interpret locally, or imprecise wording that avoided locking in a single process. That only became visible once everything sat in one repo with enforced structure. A single formatting and section template raised readability fast. Quality jumped once every SOP shared the same shape.
The agent prepares changes. Publish happens only after PR approval by the stakeholder group. Building the agent took about a week; the rest of the four-week calendar was credentials, approvers, training, and organizational alignment.
Discuss your project
30-minute call. We tell you honestly if we can help.
Need the same pattern? See Operational Agents.