The challenge
A licensed cannabis producer was at the very start of its AI journey. There was real interest across the business, but no shared view of where AI could help, what was realistic, or where to begin. Ideas lived in individual heads and team conversations. Some overlapped with capabilities already arriving through a recent ERP migration. Others depended on data or processes that weren't ready. Without a way to compare them, leadership had no confident place to start.
What we did
We interviewed the business. We sat down with teams across Finance, Commercial, Cultivation, Supply Chain and shared enterprise functions, and surfaced every idea, not just the obvious ones. That produced 50 candidate use cases, each traced to the team that raised it.
We organized every idea on the axes that matter. Each use case was assessed on:
- Value: the business benefit, with the KPI it would move.
- Effort: how hard it would be to build or adopt.
- Readiness: whether the process is defined and the data is available.
- Dependencies: what has to land first, whether that's another use case, a data foundation or an ERP capability.
- Governance and risk: applied as a veto, not a ranking factor.
- Stakeholder agreement: whether business and IT views were corroborated (29), contested (10) or not yet validated (11).
Business and IT scored each idea separately, and the scores were weighted out of 50 so benefit and effort drove pilot selection.
We sorted every idea into a clear next step:
- Pilot (9): small enough to prove value in weeks, with the data and the process already in place.
- Adopt (6): the capability already exists or is already licensed. The work is rollout, training and process, not a new build.
- Foundational (2): data groundwork that makes much of the rest work better.
- Future wave (7): real opportunities, deliberately sequenced behind the pilots.
- Blocked (14): each held up by a specific, named blocker.
- Routed (12): already covered by the ERP roadmap or existing BI, so tracked rather than proposed.
The results
- Confident first steps: 9 pilots across Finance, Commercial, Cultivation and Supply Chain, each with scores, a rationale and a baseline to measure against. One example is an AP inbox triage agent expected to save about a day a week and ensure no invoice goes unactioned.
- A roadmap for the months ahead: adopt-now wins, foundational data work and future-wave opportunities sequenced behind the pilots, with every blocker named.
- Money not spent: 12 ideas were routed to existing ERP and BI capability instead of new builds. For example, invoice extraction and PO matching were left to the ERP's own accounts payable agent rather than duplicated.
- A shared view of AI across the business: every team could see where its ideas landed, and why.