Agentforce ARR grew 240% in Salesforce's Q2 FY27 results. Zero Copy data grew 731%. Only one of those numbers tells you what customers are actually buying — and it isn't the one in most of the headlines.
Both figures come from the same Salesforce earnings release. But they answer different questions. The Agentforce number says licensing is accelerating. The Zero Copy number says where the money is going first: into the data foundation agents depend on. If you are a CFO or COO sizing an FY27 AI budget, the second number should shape your spreadsheet.
What happened: In Salesforce's reported Q2 FY27 results (August 26, 2026), Zero Copy data volume grew 731% year over year — roughly three times faster than Agentforce ARR's 240% growth. Zero Copy is Salesforce's approach to federating data where it already lives instead of copying it into the CRM.
Why it matters: Customers — especially regulated firms — are funding data-foundation work before agent deployment. The practical translation: the line item you are calling an "Agentforce budget" is mostly a data-foundation budget, and the sequencing is not optional. Agents inherit whatever the data layer hands them.
Here is the compact version of what Salesforce reported for Q2 FY27, so we can get past the numbers to what they mean:
The first four numbers got most of the coverage. The fifth is the budget signal. A 240% growth rate on agent ARR says customers are buying agents. A 731% growth rate on Zero Copy says they are reorganizing how data connects to Salesforce — at roughly three times the pace — before or alongside those purchases.
For a practitioner view of the full quarter, see our Salesforce Q2 FY27 earnings recap. This post is about the one line with the most direct implications for how you fund the work.
Zero Copy — also called data federation — lets you query, unify, and activate data where it already lives instead of replicating it into Salesforce storage: define connections to external platforms, map their schemas to Salesforce's data model, and read the data in place when needed.
That is a fundamentally different posture from the classic pattern, where every system that needs customer data gets its own synchronized copy. Copies drift and multiply storage costs — and in a regulated environment, every copy is a new surface to inventory, permission, and defend in an audit.
When Zero Copy volume grows 731% in a year, a fast-growing share of customers have decided federation is the default: data stays put, and the CRM comes to the data. That architectural commitment shapes everything built on top of it — including agents. We covered the governance side of this pattern in Data Cloud in practice: zero-ETL, governance, and quick wins.
Put yourself in a regulated seat. At a wealth management firm, a bank, or an insurer, much of the data an agent needs does not originate in your CRM. It lives in custodial platforms, core banking systems, and policy administration systems — systems you do not control and often cannot freely extract from.
Copying that data into a CRM is not just an integration project. It opens a new compliance question: which copy is the system of record, who can see it, how long it is retained, and how you prove it to an examiner. Every duplicated field is a small regulatory liability that compounds with scale.
Zero Copy is the answer that avoids the question. The custodial record stays at the custodian; the CRM queries it in place, with permissions enforced at query time. That is what a regulated firm does when data cannot be copied into a CRM without opening a new compliance question — and it is why federation is growing at 3× the rate of agents: keeping data where it lives is the precondition for everything else the firm wants to do with AI.
This is clearest in multi-custodial environments, where federation turns a years-long migration program into a configuration and governance exercise. Our multi-custodial integration architecture guide walks through those patterns, and the org consolidation playbook for M&A covers the related case where duplication lives across multiple Salesforce orgs.
Here is the honest translation for a $2B RIA — or any firm of similar complexity — sizing an FY27 AI line item.
The line in your draft budget probably says "Agentforce" and pictures licenses, some implementation help, and a go-live. But the earnings mix says your peers are spending roughly three parts on data availability for every part spent on agents. The deployment you are imagining depends on work that is not in your line item: federation decisions, data model design, household resolution, permissioning, and governance.
One note on interpretation, since this cluster of posts makes the same release do different work. Our companion piece on Agentforce crossing $1B uses the ARR numbers to make a scale argument — early adopters earn their advantage through sequencing. This post makes a narrower argument: the 240%-versus-731% gap is a spend-mix tell, not a scale tell. It shows what is actually being purchased right now, and in what order.
The sequencing point deserves emphasis: agents inherit whatever the data layer hands them. An agent answering a client question from a federated custodial record is only as accurate as the federation mapping, only as safe as the permission model, and only as useful as the household definitions underneath it. You cannot fix a weak data layer with a better agent. You fund the data layer first and let the agent benefit from it.
There is also a cost argument CFOs appreciate: federated data means fewer sync pipelines and reconciliation jobs before you count a single productivity gain. For the licensing side, our Data 360 and Agentforce pricing guide explains how flex credits actually bill.
Here is the sequence in the order the work has to happen.
Start with an inventory, not a purchase. List every system holding data your future agents will need — custodial, core banking, policy admin, portfolio accounting, document stores. For each, make an explicit decision — federate, migrate, or out of scope — based on regulatory constraints, freshness requirements, volume, and ownership. Document each decision and its reason; auditors and your future self will both want it.
Before federation mapping matters, agree on what the data means. In wealth management that is household definitions — how individuals roll up into households and accounts into relationships. Everywhere, it is a canonical model for client, account, contact, and interaction. This definitional work is where most "agent gave a weird answer" stories begin: an agent joining a custodial account to the wrong household is a modeling failure that predates the agent.
Humans forgive permission gaps and work around them. Agents do neither — they operate at machine speed on every record they can reach, doing exactly what the permission model allows. Governance therefore has to be built for agent scale, not human scale: field-level security mapped to real roles, sharing rules audited against federated sources, retention extended to every surface an agent can touch. This is the part your compliance officer cares about most — and the part most often deferred to "phase two." It is phase one.
Only after those three layers exist do agents enter the plan — against specific, measurable use cases whose data is federated, modeled, and governed. This is where Agentforce deployment work starts, and it is much faster when layers one through three are real.
For the CFO or COO building the budget, the practical move is to split the AI line item into two visible components. The first is the data foundation: federation architecture, data modeling, governance — work worth doing even if agents did not exist, because it cuts duplication cost and compliance surface. The second is agent licensing and deployment, sized against the use cases that survive a hard look at data readiness.
The split sets honest timing expectations and protects the foundation work when the AI budget gets scrutinized. The earnings numbers make the case for you: the market is funding data foundations at three times the rate it is funding agents.
Not sure whether your FY27 AI number is an agent budget or a data-foundation budget? Vantage Point's Data 360 readiness assessment maps your source systems, scores your federation and governance readiness, and delivers a sequenced plan — so the agents you deploy inherit a foundation that makes them accurate, compliant, and trusted from day one.
Explore our Salesforce Data Cloud services or talk to our Salesforce implementation and advisory team to get started.
Zero Copy is a Salesforce Data 360 capability — also called data federation — that lets you query, unify, and activate data where it already lives, in platforms like Snowflake, Databricks, or BigQuery, without copying it into Salesforce. In Salesforce's Q2 FY27 results, Zero Copy data volume grew 731% year over year.
Because the data foundation has to exist before agents deliver value, and customers are buying in that order. Federation outpacing agents roughly 3-to-1 is a spend-mix signal: the market is funding data availability first, at scale.
It means the line item labeled "AI" or "Agentforce" is mostly a data-foundation budget. Fund federation architecture, data modeling, and governance first, then size agent deployment against the use cases your data layer can actually support.
No. Migration and ETL copy data into a destination, creating a second record to sync, secure, and maintain. Zero Copy reads data in place — one record, one permission model, no drift.
Copying custodial, core-banking, or policy-admin data into a CRM opens compliance questions about systems of record, supervision, retention, and access. Federation keeps data at the source with permissions enforced at query time, avoiding the new compliance surface.
Four things, in order: make federate-or-migrate decisions per source system; agree on data model and household definitions; complete governance and permissioning at agent scale; then deploy agents against bounded, measurable use cases.
It front-loads the work that determines whether agents succeed, so deployment goes faster and sticks. Skipping the foundation usually costs more time in stalled pilots and re-work than sequencing the data layer would have taken.
Agentforce's 240% growth made the headlines. Zero Copy's 731% growth should make your FY27 budget. Agents are the visible purchase; the data foundation is the real one. Fund the foundation first, and your agents will inherit something worth acting on.
All figures cited are from Salesforce's reported Q2 FY27 results as published in its August 26, 2026 earnings release.
About Vantage Point: Vantage Point is a boutique CRM consulting firm helping organizations transform with Salesforce, HubSpot, and AI. With 150+ clients and 400+ engagements, our senior-only, US-based team specializes in data foundations, integrations, and AI deployments that work on arrival. Learn more at vantagepoint.io.