Aligning AI Projects with Organizational Priorities
The first filter for an AI project is whether the problem is one of the customer's top five priorities; the source says that if it is not, the engagement will fail.
A Working System That Never Ships
Imagine a team that has built an AI system which works well in testing. The model is accurate, the code is clean, and the demonstration impresses everyone who sees it. Yet the project never reaches daily use. Nothing in the technology is broken.
This article answers two questions. First, why can a well-built system still fail? Second, what does the source identify as the first thing to check before effort is spent on building it? The answer to the second question is a filter that is applied to the problem, not to the technology.
Testing the Problem Before the Technology
The first filter for an AI project is whether the problem is one of the customer's top five priorities. According to the source, if it is not, the engagement will fail.
The source puts the principle bluntly: the success of an AI project depends on its alignment with the organization's top priorities. Before asking whether the technology can work, a proposal has to be tested against what the organization has decided matters most.
The test has a specific shape. You ask whether the problem appears among the customer's five most important priorities. The question is about the problem and its place in the organization's agenda. It is not a question about model accuracy, code quality, or which algorithm is used. The filter comes first because it decides whether the rest of the work has a path to production at all.
Passing the first filter is a necessary starting condition. It does not by itself guarantee success. A project that clears the filter still has to win the support of several groups before it reaches daily use.
Tracing a Proposal Through the First Filter
The following scenario is illustrative. It was written for this article and is not a case from the source. It shows how the first filter is applied step by step, and it lets you check at each step whether a proposal is still on track.
Two proposals meet the same filter
A regional insurance company has five priorities that its leadership has ranked as most important this year: reducing claim processing time, improving fraud detection, meeting a new regulatory reporting deadline, consolidating customer support channels, and retaining experienced staff. Two AI proposals arrive on the same day. Proposal A is a tool that drafts claim summaries to shorten processing time. Proposal B is a tool that generates marketing copy for a seasonal campaign, which is not on the list. Apply the first filter to each.
- 1Step 1: Identify the problem each proposal addresses
Proposal A addresses claim processing. Proposal B addresses marketing copy. The filter asks about the problem, so the first task is to name the problem clearly before comparing it with the list.
- 2Step 2: Compare each problem with the top five priorities
Claim processing time appears on the list, so Proposal A passes the first filter. Marketing copy does not appear among the five priorities, so Proposal B fails the first test.
- 3Step 3: Apply the consequence the source describes
The source states that a problem outside the top five means the engagement will fail. Proposal B is therefore predicted to fail at this stage, regardless of how good its model is.
- 4Step 4: Keep Proposal A in view without declaring success
Proposal A has cleared the first filter, which is a necessary starting condition. It still has to win the support of the four groups discussed later in this article before it can reach production.
Proposal A passes the first filter and may continue. Proposal B is predicted to stall because its problem is not among the customer's top five priorities.
Why Priority Decides Whether Work Continues
The source gives its reason in one phrase: the reason is organizational physics, not technology. A project's fate is set by people and institutions, not only by the quality of the model. A system that is technically correct still has to be deployed, and deployment changes how an organization works.
This is why the filter asks about priorities. Changes that touch processes, systems, and daily routines are costly for an organization to absorb. The source notes that none of the support that production deployment requires happens for problems that are not high on leadership's list. Priority is what makes the organization willing to spend the attention and political capital that deployment demands.
| Barrier type | What it asks | Where it sits |
|---|---|---|
| Technical barrier | Can the model be accurate and the code be sound? | Inside the system itself |
| Organizational barrier | Does the problem rank among the organization's top priorities? | In leadership decisions and institutional attention |
| Deployment barrier | Will the groups affected by the change agree to absorb it? | Across senior leadership, legal, IT, and the line workforce |
A technically sound system can still fail at the organizational barriers, which the source identifies as the reason the first filter comes before technical effort.
Four Groups That Must Act
Once a problem has passed the first filter, the source describes what it takes to put an AI system into production. Four groups must act, and each contributes something different. The diagram below shows how these requirements connect to deployment.
Predicting a Stall Before Building
What do you think happens?
A team proposes an AI system whose problem is not one of the customer's top five priorities, but the model is highly accurate and the demonstration is impressive. What does the source say will happen?
Reveal answer
The engagement will fail.
The source states that if the problem is not one of the customer's top five priorities, the engagement will fail. Model accuracy does not override the first filter, because the filter tests the problem against the organization's priorities rather than testing the technology.
A customer's top five priorities for the year are reducing supplier delays, improving warehouse safety, adopting a new inventory system, expanding into two new regions, and reducing energy costs. Three AI proposals are on the table: (1) a model that forecasts supplier delays, (2) a model that generates social media posts for brand awareness, and (3) a model that flags unsafe warehouse conditions from camera footage. For each proposal, state whether it is likely to stall at the first filter, and explain your reasoning using the source's rule.
Hints
- Start by naming the problem each proposal addresses, then compare that problem with the five listed priorities.
- Remember that a problem outside the top five means the engagement will fail, while a problem inside the list only passes the first filter.
Mistakes When Applying the Filter
Common mistake Judging a proposal by the quality of its model
What it looks likeA team argues that its accurate model should be built because accuracy is impressive.
Why it failsThe first filter tests the problem against the customer's top five priorities. Accuracy does not change whether the problem matters to the organization.
Do this insteadAsk first whether the problem appears among the five priorities, and only then evaluate the technology.
Common mistake Treating the filter as a guarantee of success
What it looks likeA team whose problem is on the priority list assumes the project will reach daily use without further work.
Why it failsThe source says passing the filter is a necessary starting condition and does not by itself guarantee success.
Do this insteadAfter passing the filter, plan for the four groups whose agreement or support production deployment requires.
Common mistake Assuming the barriers are mainly technical
Why it failsThe source attributes the reason for the filter to organizational factors, not technology.
Key Takeaways
- The first filter for an AI project asks whether the problem is one of the customer's top five priorities, and the source says that if it is not, the engagement will fail.
- The filter is organizational rather than technical: a project's fate is set by people and institutions, and a technically sound system can still fail to reach production.
- Production deployment requires senior leadership attention over months, legal and compliance agreement to a process change, IT function access to systems, and line workforce acceptance of daily routine changes.
- Passing the first filter is a necessary starting condition, not a guarantee, so the four groups must still be engaged before deployment.