framework 05 · enterprise technology

The Technology Model

The most expensive bottleneck in most enterprises is the one nobody admits to: leaders don't understand how technology works, and technologists don't understand how it makes money.

By · V-Squared AI

The most expensive bottleneck in most enterprises is the one nobody admits to: leaders don't understand how technology actually works, and technologists don't understand how it makes money. So the business funds things engineering can't build while engineering builds things the business can't monetize.

The shared model that closes the gap

The Technology Model is an explicit account of how technology gets built, why it's built that way, and how it gets monetized. Once that model is shared, the C-suite can fund against how technology actually creates value, and technologists can build against how the business actually makes money. The translation layer between them stops being a person who happens to be in the room and becomes a model the whole enterprise can use.

Where the roadmap starts

Understanding the technology model is where the Enterprise AI Transformation Roadmap starts, because you can't align strategy to something you can't see. The full treatment, including how it connects to opportunity discovery, maturity, and platform monetization, lives at Strategy.vin.

You can't align strategy to a model you can't see.
the roadmap · strategy.vin

Understanding the technology model is where the roadmap starts

The full treatment, covering how the technology model connects to opportunity discovery, maturity, and platform monetization, lives at Strategy.vin.

Go to Strategy.vin
the evidence · earnings calls, july 2026

The shared model got stated out loud, on the record, by the vendor.

The Technology Model is the explicit account of how technology gets built, why it gets built that way, and how it makes money. Most enterprises never write it down. In July 2026 Microsoft and Meta both narrated theirs to investors, and the two accounts disagree in a way worth studying.

MicrosoftFY26 Q4 · Jul 29, 2026

Keep the harness separate from the model

Nadella's architectural instruction to enterprises: keep the harness separate from the model, so memory, context, and action space sit outside any model family and every model stays substitutable. Use frontier models where they earn it, cheap models where they do not, and train your own when you want neither, because you already hold the outputs, the traces, and the context. He said Microsoft intends to evangelize that design pattern.

what it confirms

This is the build half of the technology model, stated plainly enough for a CFO to fund against. The durable asset is the harness. The model is a substitutable input. An organization that cannot articulate that distinction will keep funding model access and wondering why nothing compounds.

MicrosoftFY26 Q4 · Jul 29, 2026

The cost-to-outcome curve, with the routing ratio published

A small cyber model outperforming a much larger frontier model at half the cost, with 90% of tasks handled by the small model and 10% escalated. An 89% GPU cost reduction in Dynamics 365 and up to 84% in PowerPoint. 10% lower median token usage on GitHub Copilot. Copilot workload throughput up fourfold since January. Maia 200 at 30% better performance per dollar.

what it confirms

Here is how technology gets built and how it makes money in one set of numbers. Note what is being optimized: not capability, but cost per delivered outcome. Any technology model that still treats the frontier model as the default is funding a curve that the vendor itself has stopped riding.

MicrosoftFY26 Q4 · Jul 29, 2026

Token spend became a licensed product

Amy Hood described the E7 SKU as selling both observability and manageability of token spend across business processes. Alongside it, Agent 365 registered close to 40 million agents across tens of thousands of companies in two months.

what it confirms

The monetization half of the technology model is no longer abstract. Consumption is the unit, governance of consumption is a product, and an enterprise that cannot see its own token spend by business process has a modeling failure with an invoice attached.

MetaQ2 2026 · Jul 29, 2026

Same capacity, two monetization models, and the choice is technical

Zuckerberg said it would be foolish to simply sell all of the compute, and that he expects a significantly higher margin from selling intelligence than from selling compute directly. He described running an efficient auction over compute the way Meta auctions ad inventory, and said Meta expects to evolve business agent products so that companies only pay when Meta delivers results for them.

what it confirms

Two companies with comparable infrastructure reached opposite monetization conclusions, and the deciding input was an architectural read, not a finance one. That is precisely the decision an enterprise cannot make well when leaders and technologists hold different models of how the technology works.

Sources: Alphabet Q2 2026, Microsoft FY26 Q4, and Meta Q2 2026 earnings calls and releases, plus Amazon Q2 2026 results reported July 30, 2026. Figures are as stated by company executives on those calls. Amazon reported after market close on July 30, so Amazon figures here come from the release and initial call remarks rather than the full transcript.

the wider evidence · both halves, measured

The gap got measured in 2026, in the boardroom and in the pricing model.

This page claims leaders don't understand how technology works and technologists don't understand how it makes money. Both halves were measured directly this year. One at the body that approves the funding, one in how the market now charges for value.

peer-reviewedsurvey researchcompany filingsmarket dataindependent audit
Boston Consulting Groupsurvey researchMay 2026 · 625 leaders

CEOs and their boards don't share a model of the technology

BCG surveyed 351 CEOs and 274 board members at companies with at least $100 million in revenue. 61% of CEOs said their boards are rushing AI transformation. Around 75% of board members rated their own AI understanding at or above their peers, an assessment their CEOs didn't share. Board members with less confidence in their own AI knowledge were more likely to believe their organizations were moving too slowly. Roughly 80% of both groups agreed prospective directors should have to demonstrate measurable understanding of how AI reshapes their industry.

Source: BCG, Split Decisions: The BCG CEOs and Boards Survey, first edition, May 2026.
what it confirms

This is the technology model bottleneck sitting where it does the most damage, at the body that approves funding. Watch the direction of the error. The less a director understood, the more urgency they felt. A missing shared model doesn't produce caution. It produces confident acceleration in an unexamined direction.

Boston Consulting Groupsurvey researchMay 2026

The accountability gap, with a number on it

In the same survey, CEOs estimated 35% of their performance evaluation depends on achieving AI return on investment. Board members put it at 27%.

Source: BCG, Split Decisions: The BCG CEOs and Boards Survey, May 2026.
what it confirms

Eight points between what a CEO thinks they're accountable for and what the board thinks they're accountable for. Neither side is wrong, because there's no shared model to be wrong about. Every funding argument downstream inherits that ambiguity.

Enterprise software marketmarket data1H 2026

The monetization half moved, and seats lost

Futurum Research surveyed enterprise software decision makers in the first half of 2026 and found fewer than 20% of buyers still prefer per user pricing. 43% preferred consumption based models and 27% favored outcome based structures. Zendesk and Intercom now bill for successful AI resolutions rather than AI seats. Decagon prices around resolved interactions. Deloitte published accounting guidance on revenue recognition for outcome based pricing in agentic AI products in June 2026.

Sources: Futurum Research 1H 2026 Enterprise Software Decision Makers survey; Deloitte Technology Spotlight on outcome-based pricing, June 4, 2026.
what it confirms

The accounting guidance is the item to watch. Pricing experiments are opinions until the accounting profession writes rules for them. After that they're infrastructure. If your technology model still assumes value gets captured per seat, it no longer matches how your own market prices work.

Executive AI usagesurvey research6,000+ executives

Leaders are mandating a technology they don't use

A survey of more than 6,000 senior executives across four countries found nearly 70% of CEOs, CFOs, and senior executives use AI at work less than one hour per week, including 28% who never use it. Many of the same organizations set adoption mandates and track employee usage.

Source: survey of senior executives conducted with Stanford economist Nicholas Bloom, reported March 2026.
what it confirms

You can't build an accurate model of how a technology creates value from a demo and a vendor deck. This is the modeling failure measured in hours per week, and it explains why funding decisions and engineering reality drift apart so reliably.

The BCG survey covers companies above $100 million in revenue, so these findings describe the large enterprise segment rather than the whole market.