How a CPG Brand's AI Reviewer Caught 14 of 15 Brand Faults

A North American CPG brand selling through grocery, natural, marketplaces and its own store had five years of orders in an ERP staff disliked and that couldn't sync with anything else. Two reporting tools from an outside consultancy produced two different reports, and much of the work was manual: order entry, vendor setup, hand-closed work orders, nightly cycle-count files and payments posted one at a time. The daily sales report was assembled each morning by one person from five or six reports and didn't go out if she was away. AI use was individual, on free accounts, with no guidelines, training or connection to work systems.

North Group took on the role of a fractional AI department. Discovery ran through an all-hands, department kickoffs and interviews with nearly everyone in the office. Training covered a half-day in-person workshop on a shared Claude team plan, a second session on connectors, scheduled automations and skills with one automation built live, and smaller sessions for finance, fulfillment and quality. Planning produced a workflow map, a memo on two paths for the ERP question, and a six-month automation roadmap with the executive dashboard pulled into month one. Governance covered an AI use policy, an AI-in-hiring policy red-teamed before release, a team AI plan with spend managed by finance, and an HR policy assistant in the company's chat workspace. The first automations were a marketing creative-review skill that checks work against brand guidelines and a due-date follow-up automation, both tested and handed to the team in late August.

The engagement opened with a sixty-day brief and discovery period: an all-hands meeting, department kickoffs, and interviews with nearly everyone in the office, which fed a workflow map of where time went. Training came next, starting with a half-day in-person workshop on a shared Claude team plan and a second session on connectors, scheduled automations and skills, where one automation was built live in the room. Finance, fulfillment and quality each got a smaller follow-up session. Governance ran alongside: an AI use policy, an AI-in-hiring policy that was red-teamed before release, and a team plan with spend managed by finance. The roadmap sequenced automations over six months and moved the executive dashboard, requested separately by five departments, into month one. The first two automations, a creative-review skill and a due-date follow-up, were backtested against planted faults and real cases, then handed to the team in late August. The due-date automation passed all 15 of its test cases. The executive dashboard is a working prototype with its data layer designed.

Mid-sized consumer brands with in-house manufacturing, several sales channels and lots of manual back-office work, where AI use is scattered across free accounts and leadership wants one partner to set policy, train the team and start automating.