The pattern of becoming AI-native.

The AI-Native Shift, Part 3 — Value-First Newsletter

Preheader: Two protocols connect AI to capability and to human context. Neither tells your business what value even means anymore. This is the third layer.

Part 2 made the case that AI-native businesses need a different operating model than industrial-age B2B. Optimization won't get you there. The shift is structural.

What Part 2 didn't fully name was the operating model itself — the layer that turns AI capability and human context into actual value, repeatedly, for every stakeholder a business serves.

That layer has a name now.

The Headline First

Three protocols are converging to make most of what currently lives in the middle of every business stack unnecessary.

• MCP — the Model Context Protocol — gives AI capability. It standardizes how models reach the tools and systems where work actually happens.
• HCP — the Human Context Protocol — gives AI human context. It standardizes how a person's values, preferences, and trust parameters travel with them across AI systems.
• VCP — the Value Creation Protocol — gives organizations the methodology to turn capability and context into realized value for every stakeholder they serve.

MCP is the wiring. HCP is the signal. VCP is what the wiring and signal are for.

The first two are being built by Anthropic, the Linux Foundation, MIT, Oxford, Microsoft Research, and the Stanford Digital Economy Lab. The third is what the Value-First Team has been building in practice for years, and what this edition is naming publicly for the first time.

If you run a business in 2026 and you've been asking how do we actually use AI to create value, not just automate tasks? — this is the layer you've been missing.

Let me walk you through how all three fit together, why the third one matters most for your business, and what happens when you get the full stack working.

Start With What's Already Built

MCP: Capability Reaches the Work

In November 2024, Anthropic released the Model Context Protocol. A year later it was adopted by the Linux Foundation as an open standard.

Before MCP, every AI integration was a custom build. Every model connecting to every tool required its own bespoke wrapper. Developers were drowning in N-by-M integration problems. If you had five models and ten tools, you needed fifty connectors. Most of them broke the moment anything changed on either side.

MCP solved the connection problem. It gave the industry a shared grammar for how AI capability reaches systems. One protocol. Every model can now talk to every tool through the same interface.

This is already transforming enterprise infrastructure. The bespoke integration middleware that has consumed IT budgets for two decades is collapsing into a single protocol layer. This is the first wave of what the SaaSpocalypse book has called the Great Simplification.

But MCP is deliberately indifferent to what it carries. It doesn't know if a tool invocation aligns with your values, your strategy, or even your best interests. It just moves the payload.

HCP: Human Context Reaches the Model

In mid-2025, researchers at MIT, Oxford, Microsoft Research, and the Stanford Digital Economy Lab published a position paper titled "Robust AI Personalization Will Require a Human Context Protocol."

The argument was clean. Every large model has been trained on aggregated human preference data. That training produces a generalized average user — helpful in the abstract, unhelpful for any actual person trying to accomplish anything specific. Custom instructions and system prompts are fragile stopgaps. They live inside vendor ecosystems. They don't travel. They don't compound. They don't survive the moment you switch from one AI platform to another.

HCP proposes that human context — values, preferences, trust parameters — should be portable the way a passport is portable. You own it. You control it. Any AI system you choose to work with reads it through a standardized protocol, and you get continuity instead of starting from scratch every time.

Stanford's Loyal Agents initiative is building the reference implementation, in partnership with Consumer Reports' Innovation Lab. Mirror is commercializing a version of it. A domain exists at humancontextprotocol.com. This is moving.

HCP addresses the other half of what MCP leaves untouched. MCP makes tools legible to the model. HCP makes the person legible to the model.

Both are necessary. Neither is sufficient.

What's Still Missing

Here's the uncomfortable truth about AI adoption in business right now.

A leadership team can wire their systems with MCP. Their individual users can carry HCP preference vectors from provider to provider. Every model in the stack has access to rich tools and rich human context.

And the business can still create exactly zero additional value.

Worse: most businesses can no longer tell you what value even means anymore.

The industrial-age equation that defined value for the last hundred years — time equals money equals value — broke the moment AI started doing in seconds what humans used to do in hours. If a deliverable that took a team a week now takes the AI an afternoon, what's the value? The hour count? The output? The outcome the customer actually wanted? Nobody has a clean answer. The pricing breaks. The cost accounting breaks. The billable hour breaks. The whole frame for what business value is got destabilized while the protocols above were getting built.

Underneath the broken equation is a deeper inversion.

The default operating system runs cost in → value out. Measure the input, justify the output. Hours billed. Dollars spent. Time spent producing. Industrial-age business assumed input was the thing you measured and value was what flowed from it.

The operating system that replaces it runs value first → resources flow. Name what matters. Then the spending — of time, money, attention, AI capability, human context — comes after.

AI didn't invent that distinction. It just made the old equation untenable. Value was never an output of input; valuing is what drives the spending of input. AI forced the recognition because the input metric finally collapsed too loudly to ignore.

Most businesses haven't fully noticed yet because they're still pricing and operating on the old equation by default. The ones who have noticed are quietly panicking, because you can't redefine value on the fly while running a business — and the protocols above don't help. MCP wires the capability. HCP carries the human context. Neither tells the business what value is now.

So when MCP delivers capability and HCP delivers context, they deliver both into a vacuum where the business hasn't redefined what it's trying to create.

Which means neither protocol answers the question businesses are actually asking:

Given that AI can now reach every tool and understand every person, and given that the old definition of value isn't holding — what does value even look like anymore, and how do we direct capability and context toward creating it?

That's not a protocol problem in the MCP or HCP sense. It's a methodology problem. Two methodology problems, actually: defining value in the AI-native era, and directing operations toward creating it.

In the absence of a named methodology that solves both, organizations default to whatever industrial-age patterns their training data pre-loaded — funnels, pipelines, quick wins, phases, conversion rates. The language gives it away. The outputs confirm it.

You end up with a highly capable, highly personalized AI doing the same things businesses have been doing wrong for twenty years. Just faster.

This is the gap VCP fills. Both halves of it.

Enter VCP: The Value Creation Protocol

The Value Creation Protocol is the protocol identity of the Value-First methodology — the methodology expressed in a form that machine systems can parse, declare, and execute without regressing to industrial-age defaults.

VCP is not a network protocol in the technical sense. It does not have a JSON-RPC specification. It does not compete with MCP or HCP. It sits on top of both and does the two things neither was ever designed to do: name what value actually means in the AI-native era, and direct capability and human context toward creating it.

VCP is not a new methodology either. The methodology is Value-First. VCP is what gives that methodology a place to stand in the AI-native protocol stack — what makes it parseable for the AI systems doing work alongside the team without quietly regressing to the operating model the methodology was built to replace.

If MCP is the protocol layer and HCP is the personalization layer, VCP is the operating-model layer. It turns the stack into a business.

Five load-bearing claims distinguish VCP from industrial-age operating protocols. The first three answer what value is now. The last two answer how to operate toward it.

Mutual value creation. Value is created across stakeholders, not extracted from them. The customer, the team, the operator, the business, and the broader network of partners are all participants in value creation. Any frame that treats one party as the source and another as the target encodes an industrial-age default the AI-native era has already outgrown.

Context as substrate. Context isn't an input to operations — it's the substrate operations run on. Value lives in the unified context of relationships over time, not in discrete transactions captured in fragmented systems.

Relationships over transactions. Relationships are the unit of value creation; transactions are events within them. The deliverable that took an afternoon instead of a week is not the value. The relationship that compounded because the work happened that fast — that's the value.

Configuration over customization. Native platform capability beats custom architecture, every time.

Just-in-time over just-in-case. Capability shows up when needed, not accumulated against hypothetical need.

Five claims, all platform-agnostic. Each one names a place where industrial-age defaults will pull the AI in the wrong direction unless the organization names the alternative explicitly.

A Quick Disambiguation: VCP vs. CVP

Two acronyms that sound nearly identical have to live next to each other in the Value-First world, so let's separate them once and move on.

• VCP — Value Creation Protocol — methodology-as-protocol. What the organization declares.
• CVP — Customer Value Platform — platform category. What the organization runs on. HubSpot is the current best-fit implementation.

CVP is what VCP runs on. VCP is what's declared. They're not synonyms, alternates, or competing names. When in doubt, expand the acronyms — Customer Value Platform and Value Creation Protocol are unambiguous in a way the three-letter forms are not.

The Working Parts

VCP isn't theoretical. The frameworks that make up its working parts have been operating inside the Value-First Team and its client work for years. What's new is naming them as the methodology layer of a three-part stack that the industry is already converging on.

A note on what's mature and what isn't: the methodology itself is locked. The substrate that makes it fully machine-parseable end-to-end — a structured Lexicon, a declarative grammar called VCP-Lang, and a relational specification called the Value Graph — is in active development. What you can run on today is the methodology. What's being built underneath is the encoding that lets AI systems read it natively without translation. Worth being honest about that distinction up front.

Here's what VCP is the protocol form of:

The Value Path

Eight stages of natural progression — Audience, Researcher, Hand-Raiser, Buyer, Value Creator, Adopter, Advocate, Champion. Not a pipeline. Not a funnel. A readable map of how humans actually move toward and through value.

VCP names the Value Path as the primary signal an organization should read. When you design with Value Path stages in mind, your AI agents stop asking how do we convert this person? and start asking what does this person need next to progress naturally? Different question. Completely different system.

The Four Unified Views

Customer, Revenue, Business Context, Team Enablement. The semantic layer that turns scattered data into something an AI — or a human, for that matter — can reason about with purpose.

Most organizations have data. Very few have a model that tells their AI what the data means in the context of creating value. The Unified Views provide that model. VCP names them as the canonical view schema any organization can adapt.

The Five Core Beliefs

Natural Value Flow over Artificial Control. Empowerment over Learned Helplessness. Wholeness over Fragmentation. AI-Human Partnership over Replacement. Emergence over Predictability.

The philosophical foundation of the methodology. Each belief names a Value-First approach paired against the industrial-age default it replaces. The five load-bearing claims earlier in this piece are what these beliefs sound like when expressed as protocol claims AI systems can read and act on. VCP gives the foundation a place to stand in the protocol stack — same beliefs, declared in a form the AI doesn't quietly override.

The TEACH Values

Transparent ↔ Trust. Empathetic ↔ Empowered. Agile ↔ Adaptable. Confidence ↔ Conviction. Humble ↔ Hungry. Five tensions, held productively.

Every relationship your organization has with every stakeholder sits somewhere on each of these five axes. VCP names them as the operating posture for any participant — human or AI — doing work inside the stack.

The Three-Org Model

Customer Org (human-led, AI-supported). Operations Org (AI-led, with human oversight). Finance Org (shared between them). The organizational architecture that makes AI-native operations coherent instead of chaotic. It answers what should humans actually do, and what should AI do, and how do they hand off? without either side losing itself in the other.

VCP names the Three-Org Model as the structural pattern for organizations making the AI-native shift.

None of these frameworks are decorative. They are the protocol's working parts. Together, they give any organization a way to act on MCP capability and HCP context that actually produces value for the people on every side of the work.

How the Three Layers Work Together

Imagine a single customer interaction inside a business running the full stack.

MCP is active. The AI agent working on behalf of your Customer Steward can reach the CRM, the calendar, the payment system, the content library, the analytics platform. Any system required to serve this relationship is one protocol call away. No bespoke integrations. No duct tape. Just capability.

HCP is active. The customer arrives carrying their own human context. Their values, their accessibility preferences, their trust parameters — all portable, all legible to the AI that will interact with them. Your agent doesn't have to guess. It reads.

VCP is active. Your organization has named the Value Path, the Unified Views, the Core Beliefs, the TEACH Values, and the Three-Org Model as the operating model it runs on. The AI knows what stage this customer is in. It knows which Unified View is relevant for the decision at hand. It refuses to pattern-match the customer into a funnel because the Core Beliefs won't let it. It keeps the Customer Steward in the human-led loop where character and judgment matter, and it handles the coordination, documentation, and follow-through that belong in the Operations Org.

The result is not faster sales. That framing is an industrial-age holdover.

The result is compounding value — for the customer, who experiences genuine partnership instead of extraction. For your team, who spend their time on relationships and judgment instead of data shuffling. For your business, which accumulates trust and capability instead of spending both on coordination overhead. For the broader ecosystem of partners and advocates, who become participants in value creation instead of targets of acquisition.

Capability plus context plus methodology. All three layers, doing the work each was designed for.

Why This Matters Right Now

If you are a leader, operator, or builder, here is the immediate implication.

MCP adoption is already happening around you. Your vendors are adding MCP servers. Your team is using AI agents that speak MCP. Your competitors are wiring their systems to the protocol. The capability layer is being installed whether you're paying attention to it or not.

HCP adoption is about to accelerate. As consumers and enterprise users get tired of re-explaining themselves to every new AI they touch, the demand for portable human context will force the issue. Stanford, Consumer Reports, and the commercial startups in this space are building the rails.

VCP adoption is where you still have a choice. You can let your organization default to whatever methodology the training data smuggles in — funnels, pipelines, quick wins, leads, conversions — and watch your AI investments produce marginally faster versions of the industrial-age patterns you were already stuck in. Operating on a definition of value the technology already invalidated.

Or you can make the methodology layer explicit. Name what value means in your business now that time-as-input has stopped holding. Hold the line. Build your AI-native operations on a foundation that knows what value actually looks like for the humans on every side of the work.

The organizations that get VCP right will not just use AI better than their competitors. They will operate inside a fundamentally different business physics. Relationships that compound. Teams that scale capability instead of headcount. Customers who progress naturally through a value path instead of being funneled down a pipeline. Data that has meaning because the model gives it meaning. Operating discipline that holds because the protocol is explicit.

This is not a theoretical future. This is the difference between businesses making the AI-native shift and businesses using AI tools.

An Open Framework

A final note, and this matters to how we want this adopted.

VCP is not a Value-First Team product. It's the protocol identity of a methodology we have been building, teaching, and refining — and the moment has arrived to release the protocol form of it as a named, shareable, extensible framework that anyone can implement, build on, or teach.

The Value-First Team has a specific implementation tuned for HubSpot-centered organizations and AI-native operations. That implementation includes the agent architecture, the data model, the operating skills, and the program delivery infrastructure. We offer it through Office Hours, the Activation Workshop, the four-week AI-Native Shift program, and the full Value-First OS deployment. A clear ladder. Nobody skips rungs.

But the protocol itself is not ours to gate. MCP is an open standard. HCP is being developed as an open framework. VCP sits in the same category. It is teachable. It is extensible. It is yours to use if it resonates.

What we ask of anyone who builds on it is the same thing we ask inside our own work.

• Honor the humans on every side of the work.
• Let capability serve meaning, not the other way around.
• Refuse to collapse transformation back into optimization.
• Build systems that compound trust instead of spending it.

If those commitments sound like the way you already want to work, you are already running VCP — you just didn't have the name for it.

Now you do.

Where to Go From Here

If this is your first encounter with Value-First, start with Office Hours. Free, three times a week, no commitment. You'll meet the methodology in its natural habitat and see whether it fits the problems you're actually trying to solve.

If you're already running MCP internally and wondering why your AI stack hasn't produced the value you expected, book an Activation Workshop. We build something real in an hour — usually a sales deck turned into a branded microsite, or a daily operations skill configured for your business — and the gap between where you are and where VCP lives becomes visible.

If your organization is ready for the full shift, the AI-Native Shift is a four-week cohort program for leadership teams. Your whole team goes through it together. You leave with a working deployed stack, a new mental model, and the capability to continue independently.

If you're a HubSpot partner, consultant, or practitioner who wants to deliver this methodology to your own clients, the Value-First Collective is the path.

The protocols are converging. The middle layer is collapsing. The old definition of value is no longer holding. The organizations that name what value means for them — and the operating model they want to run on — before the default model names them, will be the ones who come out the other side of the SaaSpocalypse stronger than they went in.

Build on it.

Chris Carolan is the founder of the Value-First Team. The Value-First Team helps mid-market B2B organizations make the AI-native shift through the Value-First methodology — implemented on HubSpot as the Customer Value Platform, and given a protocol identity through the Value Creation Protocol. Office Hours, Scoping, and full implementations available at valuefirstteam.com.