Services: The New Software — Why Agencies Win the AI Era

In March 2026, Sequoia partner Julien Bek published an essay that reframed the entire AI opportunity in one sentence: “The next $1T company will be a software company masquerading as a services firm.” If you sell digital services to enterprise clients — and you have been doing it for a while — that sentence lands differently than it does for a software founder. It describes the business you have been running.

The 6:1 Ratio That Defines the Real Market

Bek's central data point: for every dollar companies spend on software, they spend six on services. The total labor market for outsourced services dwarfs the SaaS market. IT managed services alone account for over $100B in outsourced spend annually. Management consulting runs $300–400B. Supply chain and procurement services hit $200B+. Accounting and audit: $50–80B in outsourced spend in the US alone.

These are not niche markets waiting to be discovered. They are mature industries that have been absorbing enterprise budgets for decades, precisely because the work requires expertise that most companies do not want to hire and retain in-house. The digital services that PixelEruption delivers — AEM and CMS implementations, SEO programs, HTML5 banner production — sit in this same category. Enterprise pharma, banking, and automotive companies do not want to hire a team of AEM architects. They want their platform built and running.

Copilot vs. Autopilot: Where the Value Actually Sits

The essay's sharpest distinction is between copilots and autopilots. A copilot helps a professional do their job faster — Cursor helps a developer write code, Harvey helps a lawyer draft documents. The human is still in the loop; the software is a seat license. An autopilot sells the completed outcome. The client does not care about the tool; they care about the closed books, the approved banner, the indexed page.

Bek's framing is direct: “If you sell the tool, you're in a race against the model. But if you sell the work, every improvement in the model makes your service faster, cheaper, and harder to compete with.”

That dynamic inverts the competitive threat. When a new Claude release ships, a software tool vendor has to scramble to stay differentiated. A services firm that uses AI internally to deliver outcomes just got faster and more profitable. The model improvement is captured by the firm, not leaked to the client as a reason to cancel.

Why Outsourced Work Is the Right Starting Point

Bek identifies a specific entry strategy: “Start with the outsourced, intelligence-heavy task. Nail distribution. Expand toward the insourced, judgement-heavy work as the AI compounds.”

The logic is clean. When a company already outsources a function, they have already accepted that external parties can do this work. There is a budget line. The procurement motion is a vendor swap, not a restructuring. The buyer has been purchasing an outcome, not a tool.

The pharmaceutical, banking, and automotive clients PixelEruption serves are exactly this profile. A global pharma brand does not have 12 AEM architects on payroll. They have a CMS vendor contract and an agency relationship. When we build a multi-market AEM platform that reduces MLR review cycles from 11 days to 3.5, we are not competing with a software tool — we are replacing an outsourced services contract with a better one.

PixelEruption benchmark: Our pharma AEM implementations average a 68% reduction in time-to-publish for regulated content, with component reuse rates above 75% across market variants. That is the kind of outcome metric that wins and renews enterprise contracts — not the feature list of the CMS platform underneath it.

The Structural Shortage That Accelerates the Shift

One of the more striking data points in the essay: the US has lost roughly 340,000 accountants over five years while demand has grown. 75% of CPAs are nearing retirement. That structural supply shortage is forcing firms to accept AI-delivered work faster than almost any other profession.

The digital marketing and CMS talent market has the same dynamic at a smaller scale. Qualified AEM architects are scarce. MLR-compliant content operations require rare combinations of technical and regulatory knowledge. Senior SEO specialists who understand pharmaceutical content restrictions are not easy to hire. When the talent is structurally scarce, the enterprise buyer is already primed to accept an outcomes-first engagement model. They have no other choice.

Judgement Is the Last Moat — and Where Agencies Live

Bek's distinction between intelligence and judgement is the most important framing for anyone running a services business. Intelligence is the ability to execute a known task correctly — writing code, translating clinical notes, formatting a document. AI is now cheaper than humans for most intelligence work and improving fast.

Judgement is knowing what to build, which SEO strategy fits a regulated industry, how to structure a multi-market CMS architecture for a client entering three new markets simultaneously. “Writing code is mostly intelligence. Knowing what to build next is judgement.”

An agency that positions at the intelligence layer — we are fast and cheap at execution — will be commoditized. An agency that positions at the judgement layer — we know what to do and can execute it — becomes harder to replace as AI makes the execution layer faster. The more AI handles the intelligence work, the more the client depends on the judgement layer to direct it.

PixelEruption's engagement model has always been built this way. We do not get hired to push buttons in AEM. We get hired because we know which architecture decisions avoid a painful migration in year three, which SEO structure survives a pharmaceutical rebrand, which banner production workflow actually hits a 6-week campaign deadline across 12 markets. That is not intelligence. That is experience compressed into judgement.

What This Means for the Next Three Years

Bek's prediction: “In 2025, the fastest-growing AI companies were copilots. In 2026, many will try to become autopilots.” The transition is happening now. Pure-play autopilot firms — ones that sell completed work backed by AI — are the ones capturing the 6× services market, not the 1× software market.

For enterprise digital services, the implication is direct. Clients will increasingly ask not “which CMS do you recommend?” but “can you just run our digital operations?” The scope of the engagement expands from project-based to program-based. The relationship shifts from vendor to partner. The contract stops being a fixed-scope SOW and becomes an outcomes SLA.

That is a natural evolution for PixelEruption. We already deliver multi-language CMS programs, ongoing SEO programs, and production retainers for banner campaigns. The structure Bek describes is the direction we are already moving toward — and AI makes the unit economics significantly better.

If you are evaluating how to structure your enterprise digital operations for an AI-first world — whether that is a CMS migration, an SEO program, or a global banner production workflow — talk to us. We have been doing the work while others were debating the thesis.