# Initiative inventory: you can only prioritize what you see > Before you prioritize, list everything already running: owner, cost and baseline for each initiative. Without an inventory, the queue reflects who asked. - Author: Marlon Trettin - Published: 2026-10-20 ยท Updated: 2026-10-20 - Language: en - Canonical: https://yowpi.com/en/blog/initiative-inventory-you-prioritize-what-you-can-see --- You can only prioritize the initiatives you can see. In many companies, the list that reaches the meeting holds the ones somebody presented well, not everything already running with budget, people and a deadline. Before you order the queue, the first step is an inventory with owner, cost and baseline for each initiative. ## TL;DR - About 11% of the functional leaders surveyed by [Gartner](https://www.gartner.com/en/newsroom/press-releases/gartner-survey-finds-only-22-percent-of-organizations-have-successfully-scaled-ai-across-multiple-business-units) (1,303 respondents) did not know how much their function spent on AI in 2025. Without that number, what you have is a list of requests, not a queue. - In the same study, the top performers, which treat initiatives as a portfolio and review them regularly, reported positive returns on 81% of them. Among low performers, the return on 29% of initiatives was unknown. - At [Bain](https://www.bain.com/insights/your-ai-budget-is-growing-your-returns-arent-heres-why/), 44% of companies fund today's AI with savings from earlier automation programs. Savings without a baseline become budget without backing. - The minimum inventory has five fields per initiative: owner, cost, problem with baseline, system where it lives, and status. ## Why does the priority queue start out incomplete? Because nobody sees the total. Initiatives are born in different departments, by different routes: a vendor pilot, a subscription on a corporate card, an automation an analyst built during month-end close. Each one makes sense inside its own department. None passes through a place where someone counts them. Survey data shows it. In Gartner's survey, run between January and April 2026 among 1,303 functional leaders at organizations with annual revenue above US$50 million, [about 11% of them did not know how much their function spent on AI in 2025](https://www.gartner.com/en/newsroom/press-releases/gartner-survey-finds-only-22-percent-of-organizations-have-successfully-scaled-ai-across-multiple-business-units). At the same time, 85% planned to raise that spending in 2026. The cost of not knowing grows with the budget. The sample is organizations with revenue above US$50 million, and it covers AI. The mechanism tends to repeat in any operation that grew by accretion. The smaller the operation, the shorter the list, and the easier it is to build by hand. This is not a technology gap. It is a missing architecture decision about where initiatives get recorded. ## What does the savings from earlier automation have to do with it? It has to do with funding. In Bain's survey of 951 companies, published in June 2026, [44% said they fund AI with savings from earlier automation programs](https://www.bain.com/insights/your-ai-budget-is-growing-your-returns-arent-heres-why/). Among those that measured cost savings, nearly 40% landed below 10%, while the most common target was 11% to 20%. And 90% of those companies are raising their budgets again. Put the numbers together with care: a good part of today's AI is paid for by yesterday's savings, and among companies that measured those savings, nearly 40% fell below the floor of the most common target. Bain sells consulting in this market, so read the percentages as orders of magnitude. The pattern they describe is exactly what an inventory exposes. [McKinsey](https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai) shows the same gap from another angle. In its survey of 1,719 participants in 97 countries, run in May and June 2026, eight in ten say AI improved their own productivity, but 37% attribute any EBIT impact to AI, a share similar to the year before. The survey does not explain the distance between the two perceptions. A reasonable hypothesis is that, without a record of what was done and what existed before, it is hard to connect an individual gain to the company's result. ## What goes into an initiative inventory? Five fields per initiative, and none of them is sophisticated. The value is in every initiative answering the same questions. The owner is a person with a name, not a department. It is whoever answers when the initiative fails or when someone asks whether it should continue. Without a name, the initiative is an orphan, and orphans tend to become permanent workarounds. The cost is what it consumes per month: licenses, hours of the people who operate it, hours of the people who fix it. People's hours count even when they never touch an invoice. It is the field that tends to be missing, and the one that weighs most when the board asks whether it is worth continuing. The problem with its baseline says which pain the initiative attacks and what the number was before it: hours of rework per week, days of delay in the close, error rate. Without that number, the declared savings cannot be checked. Bain's data shows what happens when they are checked: the distance between target and result appears. The system where it lives shows whether the initiative is already part of the operation or still an experiment. An automation running on one person's laptop and another running in a controlled environment carry different risks, and the queue needs to know the difference. Status closes the list: running, in pilot or stopped. A stopped initiative that is still costing money is the first place to look for savings. To find what is hidden, two sources cover a good part of it: recurring invoices in finance and access requests in IT. The rest shows up in a twenty-minute conversation with each department head, asking what they built on their own so they would not have to wait. It is the same ground as [shadow IT](/en/blog/shadow-it-when-every-department-picks-a-tool), looking at whole initiatives instead of loose tools. ## How do you build the inventory without it becoming an audit? Treat it as a survey, not an investigation. As a planning reference, and not as a measurement, two weeks tends to be a realistic timeframe for a mid-sized company, if the scope is limited to initiatives with a cost or a system dependency. In the first week, finance and IT hand over what they see. In the second, each leader validates, corrects and fills in what only they know. A blank field stays blank and becomes data: an initiative with no owner or no known cost is already a finding. With the inventory in hand, [prioritization by dependency and bottleneck](/en/blog/prioritize-innovation-and-get-past-the-pilot) stops ordering what was presented and starts ordering what exists. The [shutdown criterion](/en/blog/when-to-shut-down-an-innovation-project) also gets material to work with, because you can only shut down what you know exists. Gartner describes this discipline among those reporting returns: organizations that treat initiatives as a portfolio and review performance regularly, reallocating or discontinuing the ones that yield less, reported positive returns on 81% of their AI initiatives. Low performers did not know the return on 29% of theirs. The survey shows an association, not a cause, but the starting point is the same in both cases: to review, someone has to have the list. ## A practical example: the board thinks there are six The scenario below is hypothetical, and the numbers are illustrative assumptions, not measurements. A packaging manufacturer with 180 employees believes it has six innovation initiatives underway, the ones the innovation manager presented over the last two quarters. A two-week inventory finds eleven. The other five are two vendor pilots nobody closed that still bill a subscription, two automations that planning analysts built and only they know how to run, and a dashboard the sales team keeps in parallel to the board's. None of the eleven has a recorded baseline, and three have no owner. The queue changes. The two forgotten pilots are shut down. The two analyst automations get an owner, documentation and a place in a system, because they feed the close. And the initiative the board wanted most, among the original six, moves down, because it depends on a master record that one of the orphan automations maintains. Without the inventory, that dependency would only have surfaced midway through the project. In the [Repap On](/en/cases/repap-on) case, the design of the document management system starts from the process landscape, the map of the company's processes: first you see what exists and how it connects, then you decide what the system controls. ## Frequently asked questions **My team resists change. Won't an inventory become a witch hunt?** It will, if it is presented as an audit. Present it as the step that protects everyone's work: a listed initiative with an owner has a stronger case in a budget cut. Ask what the person built to solve a real problem and treat the finding as demand the structure was not serving. The tone decides whether the list comes out complete or half done. **Will the inventory take too much of my team's time?** Little, if the scope is limited. The two main sources, invoices and access requests, already exist, and the rest is a short conversation with each leader. The biggest effort is the consolidator's, and that is usually you. The alternative is discovering the hidden initiatives in the middle of a project, when each surprise costs more. **How do I measure results if no initiative has a baseline?** Record what exists today, even approximately, with the date and the origin of the number. A dated estimate can be checked later. No estimate cannot. For new initiatives, make the baseline a condition for entering the queue. For old ones, measure from now on and do not rewrite the past. ## The next step If the list of your initiatives does not yet exist in one place, the Operational Architecture Diagnostic is a 30-minute conversation to sketch it with you and see what it changes in your queue: [book it through our contact page](/en/contact).