The AI category has, in eighteen months, produced one of the largest concentrations of capital deployment without execution architecture that the mid-market has seen in a decade. Companies are buying tools, hiring consultants, standing up committees, and producing strategies. Most of it is not converting into operating leverage.
This pattern is not new. The category is new. The mechanics are familiar.
For years, Fulcrum has diagnosed the same failure inside stagnant companies, sub-scale services businesses, and sponsor-backed portfolios under pressure to demonstrate value creation. Capital deploys. Advisors are engaged. Recommendations are issued. Nothing executes. The business does not advance. The investment sits as overhead.
Those companies were buying advice. They needed operating leverage.
A meaningful share of AI spend right now is the same purchase under a newer label.
The pattern, named precisely.
The pattern has a name in our practice. We call it advisory without authority. Capital deployed into a relationship that produces diagnosis, recommendations, and strategic frameworks, but lacks the implementation mandate, the operating cadence, or the decision-rights structure required to convert any of it into throughput.
The pattern fails in a specific sequence. Diagnosis is delivered. The leadership team agrees with the diagnosis. A plan is produced. Execution is delegated to an existing team that is already at capacity, lacks the relevant capability, or does not own the decisions required to act on the plan. The plan stalls. New advisors are engaged. New diagnosis is produced. The cycle continues.
By the time we are typically called in, the company has been running this cycle for two to four years. The reinvestment is real. The execution is not. Money has moved, often a substantial amount. The operating layer is largely where it was when the cycle began.
“AI accelerates the pattern because the tools are cheap enough to deploy without governance, the category is moving fast enough that almost any spend can be justified as keeping up, and the lack of internal capability creates structural dependence on outside advisors.”◆ Fulcrum Operating Partners
AI accelerates this pattern in three specific ways.
First, the tools are cheap enough to deploy without governance. A vendor relationship that costs $40,000 a year does not trigger the same scrutiny as a capital project. Multiple such relationships can accumulate inside a single function without surfacing at the portfolio level.
Second, the category is moving fast enough that almost any spend can be justified as keeping up. The framing is defensive. No CEO wants to explain to their board that they are behind on AI. This produces a procurement environment where saying yes is easier than saying not yet, and the not yet decision is the one that usually creates value.
Third, the lack of internal capability creates structural dependence on outside advisors who have the same incentive structure that produced the original pattern: paid for diagnosis, not for installed capability. The advisor’s economics improve when the engagement extends. The company’s economics improve when the engagement converts to operating output. These two incentives are not aligned, and the misalignment compounds over time.
Why the pattern produces overhead instead of leverage.
Three operating mechanics explain why this pattern produces overhead instead of leverage. They are consistent across categories, company sizes, and capital structures.
There is no owner. Every AI initiative we have observed inside companies operating in this pattern has the same structural feature. It does not appear in anyone’s accountability. The CFO sponsors the spend. The CEO endorses the direction. A consultant runs the workstream. No one inside the company is on the hook for the operating outcome. When ownership is distributed across that many parties, none of them can be removed when the initiative fails. Distributed ownership is unaccountability with extra steps.
The decision rights are unresolved. Even when an AI initiative produces a usable output, the question of who can decide to deploy it, integrate it, change a workflow around it, or shut it down sits in an unresolved layer of the organization. Operating cadence is what resolves decision rights in companies that work. Most of the companies running expensive AI initiatives do not have operating cadence that can absorb the output.
The implementation infrastructure does not exist. Deploying a new capability requires the same operating infrastructure as deploying any other change: defined workflows, accountable owners, measurement, a feedback loop, and a path from pilot to production. When this infrastructure is missing, the AI initiative produces an artifact (a tool, a model, a process map) that has no clean path into the operating layer. The artifact sits adjacent to the business, not inside it.
Each of these mechanics is recognizable to anyone who has tried to run change inside a mid-market company. They are not specific to AI. They are specific to the operating conditions that produce the broader pattern of advisory without authority. AI is currently making the conditions visible at scale because the AI category is large enough, fast enough, and expensive enough to surface what was always there.
The structural alternative.
The alternative is not to spend less on AI. The alternative is to install the operating architecture first, then deploy capital into the architecture.
Operating architecture, in our practice, means three things. A defined cadence that creates the regular forums where decisions get made. A decision-rights structure that names who owns what and at what threshold. An execution discipline that closes the gap between decision and throughput.
When operating architecture is in place, AI initiatives convert. When it is not, no amount of AI spend will create leverage. The variable is not the technology. It is the operating architecture into which the technology lands.
When this architecture is in place, AI initiatives convert. The cadence creates the forum for review. The decision rights create the path from output to deployment. The execution discipline produces the throughput that demonstrates operating leverage rather than overhead. The same capital deployed into the same vendors produces a measurably different result.
When this architecture is not in place, no amount of AI spend will create leverage. The capital will continue to convert into reports, slides, and prototype tools that do not move the business. The pattern continues. The next budget cycle approves more of the same.
This is the operating distinction Fulcrum was built to install. It is also the distinction that determines whether the current AI cycle produces value creation or simply produces a new category of overhead inside otherwise capable companies.
Five operating moves that matter more than any AI strategy.
For sponsors, operating partners, and CEOs reading this with active AI initiatives underway, five operating moves matter more than any AI strategy a company will write this year.
- Audit the current cadence. Before any new AI initiative is approved, document the operating forums where decisions about capability investments actually get made. If those forums do not exist, build them before deploying capital. The cadence is the precondition.
- Name the owner. Every AI initiative should have a single accountable name inside the company, with the authority to make the deployment decisions and the obligation to report on operating impact. If no such name can be assigned, the initiative is not ready for funding.
- Define the operating outcome, not the technology outcome. The output of an AI initiative is not a model, a tool, or a vendor relationship. The output is a measurable change in operating performance. Initiatives that cannot be tied to a specific operating metric within a defined window should be paused.
- Structure for installation, not advisory. When external capability is required, it should be engaged on terms that produce installed capability, not standalone recommendations. The advisor’s compensation, scope, and engagement structure should reward execution, not diagnosis.
- Screen for the pattern. Sponsors evaluating portfolio AI spend should screen for the advisory-without-authority pattern explicitly. If the current spend is producing diagnosis without execution, redirect the capital. The pattern does not correct itself. It compounds.
The operating thesis.
The AI category will produce real operating leverage in companies that are positioned to absorb it. It will produce expensive overhead in companies that are not. The variable is not the technology. It is the operating architecture into which the technology lands.
This is the operating thesis Fulcrum will return to throughout this section. Every piece that follows assumes it. Every framework we publish in AI In Operations is built on it. The pattern is older than the category. The discipline required to break it is the same discipline that produces compounding value inside the hold period.
We have been operating against this pattern since before AI was the label on it. We expect to be operating against it long after the category has matured. That is the work.