KM4BI KM4BI

5 Levers That Separate Winning AI Investments from Disasters

Nate B. Jones argues that successful AI investments start with workflow clarity, then choose whether to automate, build, buy, hire, or wait.

Nate B. Jones offers a practical decision framework for avoiding AI investments that look impressive in demos but fail to create business value. His starting point is straightforward: many failures do not come from agentic AI itself, but from a poor understanding of the work a company is trying to transform.

The core idea: invest in workflows, not in “AI”

Before discussing models, vendors, or dashboards, a company has to describe the workflow: what information comes in, what the system is allowed to do, what good output looks like, who checks it, what gets escalated, and who owns the result. Without that clarity, one project can hide many different needs and turn into a vague vendor search.

The five investment levers

Jones lays out five possible moves. A company can automate a repetitive workflow that is easy to verify; build an agentic loop when the work is specific and full of internal context; buy a solution or primitives when the market is mature enough; hire when a missing human capability is blocking progress; or wait when another workflow offers more immediate leverage.

The point is not to slow down AI transformation. It is to sequence it. Change-management capacity is limited, so the first investments should target the workflows where AI creates disproportionate leverage.

Build, buy, or hire depends on the shape of the work

Buying makes sense when the work is common and market solutions are mature. But when the workflow is company-specific, it may be better to buy building blocks — models, connectors, orchestration, or services — while keeping ownership of the operational loop and quality standard.

Hiring follows the same logic. Instead of searching for a perfect AI unicorn, companies should identify the missing capability: domain trust, workflow engineering, evaluation design, executive ownership, or standard-setting.

The rule to remember

The most useful line from the video is: do not automate what you cannot describe. If inputs, outputs, exceptions, standards, and owners are unclear, the AI project is already fragile before the model or vendor decision even begins.

Source