GaneevEcoWhat the exercise actually is
For each department, list the specific, repetitive tasks people do manually — with how often it happens and a rough time estimate per instance, gathered directly from the people doing the work rather than estimated by someone several levels removed from it. Not a strategy session, not a workshop — just an honest inventory, usually surfaced most naturally through the department interviews described elsewhere in this series.
The simplicity of this exercise is exactly why it's so often skipped — it doesn't feel strategic or impressive enough to warrant a dedicated initiative, so it quietly falls through the cracks between more ambitious-sounding projects that get prioritized instead, despite this unglamorous inventory often being the single most useful input to any AI planning process that follows it.
Why this alone changes decisions
- It replaces guesswork about "where AI could help" with an actual ranked list, grounded in real numbers rather than impressions or whoever argued most persuasively in a planning meeting.
- It often surfaces the same task independently mentioned by multiple departments — a strong signal for where to start, since a shared pain point across teams tends to indicate a genuinely systemic, high-value problem.
- It gives you a genuine before-and-after baseline to measure ROI against later, instead of an anecdotal "it feels faster" that doesn't hold up to any real scrutiny months down the line.
Keeping the inventory useful over time
An inventory built once and never revisited slowly loses its value as work genuinely changes — new tools get introduced, processes shift, and what was the most painful manual task a year ago may have already been partially solved by something else in the meantime. Treating this as a living document, refreshed roughly annually, keeps it a genuinely useful planning input rather than a historical artifact from whenever the original assessment happened to run.
It's also worth revisiting the inventory specifically whenever a new department or function joins the scope of your AI rollout, rather than assuming the original inventory from a different part of the organization transfers cleanly. Each function tends to have its own distinct pattern of manual work worth understanding directly, on its own terms, rather than assumed by analogy from a different team's experience.
It's worth pairing this inventory with a simple, shared spreadsheet or document that stays genuinely accessible to whoever's planning the next phase of a rollout, rather than a one-time report that gets filed away and forgotten shortly after it's produced. Accessibility over time is what turns a good one-time exercise into a genuinely lasting planning asset for the whole organization.
It's also worth explicitly asking, during the inventory exercise, which tasks people have already tried to solve themselves with some existing tool or workaround, and why that attempt didn't fully succeed. That specific question often surfaces valuable context about what a new solution actually needs to get right that the previous attempt missed, rather than starting the next attempt with the same blind spots that limited the first one.