Agentic AI Security: Tool Poisoning and Audit Trails
Explore agentic AI security risks, including MCP tool poisoning, and learn how governance, permissions, approvals, and audit trails reduce exposure.
So here's a question we get constantly: do you actually need Data Cloud to run Agentforce, or is that just Salesforce padding the invoice? Fair question. The two get sold together so often that most people just assume it's one purchase. It isn't. Below is the honest version: what Data Cloud does, when it's worth the money, what it actually costs right now in 2026, and where you can probably skip it, at least for a while.
Agentforce builds AI agents that live inside your Salesforce org. Case resolution, lead qualification, order lookups, that kind of thing. The pitch is that these agents don't just answer questions, they act: update a record, kick off a workflow, hand things off to a human when they hit a wall.
Nobody puts this on the pricing page, but here's the catch. Out of the box, an agent can only see what's already sitting in the Salesforce record right in front of it. Fine if the task is simple. Not so fine once you realize most customer relationships are scattered across way more than one system.
Data Cloud got renamed Data 360 sometime in late 2025, so if you see both terms floating around in different articles, that's why: same product. Its job is pulling data out of Sales Cloud, Service Cloud, Marketing Cloud, plus whatever else you're running (ERP, support desk, a spreadsheet someone's been maintaining since 2019) and mashing it into one profile per customer that actually stays current.
What that means for Agentforce: the agent stops guessing off a partial record and starts working off the real picture. Past orders. Open tickets. Whether they clicked the last email campaign. Here's roughly how it fits together.
Figure 1. Data Cloud unifies fragmented data before Agentforce reasons and acts on it.
Depends what you're building and where your data already lives, honestly. There's no universal answer here, so use the table below as a gut check, not a rulebook.
| Situation | Do You Need Data Cloud? |
|---|---|
| Small team, single-object use case (e.g., FAQ deflection, simple case summarization) | Likely not required. Standard Salesforce record context may be enough for a pilot. |
| Customer data lives in multiple systems (CRM + support + marketing + external apps) | Yes. Data Cloud is what unifies these into a single profile the agent can reason over. |
| Production deployment with real customers and reputational risk | Strongly recommended. Reduces hallucinations and grounds responses in verified data. |
| Already own Data Cloud for another initiative (CDP, personalization, analytics) | Cost is largely sunk. Reuse it: Agentforce becomes an incremental, not new, investment. |
| Tight first-year budget, need to prove ROI before scaling | Consider a scoped pilot without Data Cloud first, then add it once value is proven. |
Here's where budgets go sideways. Agentforce's own pricing looks fine when you first see it: a couple bucks a conversation, or credits you buy in bulk. Then Data Cloud lands on the same quote and suddenly what was a pilot budget is a six-figure ask. It happens almost every time.
| Cost Component | Typical Range (2026 list pricing) | Notes |
|---|---|---|
| Agentforce (Flex Credits) | ~$500 per 100,000 credits (~$0.10/action) | Pay-as-you-go usage model |
| Agentforce (conversation pricing) | ~$2 per conversation | Simpler to forecast for support use cases |
| Agentforce per-user add-on | From $125/user/month (Enterprise); $550/user/month (Agentforce 1) | Stacks on top of credit costs |
| Data Cloud (Data 360) | Starter SKUs from ~$60,000/year; ~$108,000/year for 10M unified profiles | Often the single largest recurring line item |
| Required Salesforce edition (Sales/Service Cloud) | From $80/user/month (Professional); $165/user/month (Enterprise) | Prerequisite if not already licensed |
| Implementation & systems integration | $50,000 – $300,000+ | Scales with data complexity and use-case count |
| Training & change management | $2,000 – $5,000 per user | Often underestimated in project budgets |
Rough total: for a mid-market rollout, say 500 users, first-year cost across everything, Agentforce, Data Cloud, the required Salesforce editions, implementation, training, usually lands somewhere between $150K and $450K. If your org already runs Data Cloud for something else, personalization, analytics, whatever, that number drops a lot. You're not paying twice for the same infrastructure.
Not everyone needs to jump straight in, and frankly most teams shouldn't. If it's a small pilot, one use case, a handful of users, nothing that blows up if the agent gets a detail wrong, running on native Salesforce record context is a reasonable way to start. See if agents even move the needle before you commit real money to the data side. Just don't let the pilot's architecture quietly become the production plan. Plenty of teams skip Data Cloud at first and end up adding it six months later once they try to scale, so at least sketch out what the data strategy would look like, even if you're not paying for it yet.
Data Cloud isn't something every Agentforce project needs on day one. But once you're touching real customers and the data's spread across more than one system, it stops being optional in any practical sense. The cost is real, usually the biggest number on the quote. The difference it makes is real too. Best move is to just be deliberate about it: know your data, run a pilot, get actual numbers, and skip the assumption that it's all bundled in for free.
Data Cloud, now called Data 360, needs to be provisioned and enabled for Agentforce use, but that does not mean every company needs a large, fully implemented Data 360 project on day one. Salesforce distinguishes between simply enabling Data 360 and implementing it with connected, unified data. The level of implementation you need depends on what your Agentforce agents are expected to do and how much data they need beyond standard Salesforce records.
The cost depends on the Data 360 capabilities, data volumes, credits, Salesforce products already in place, and the level of implementation required. There can also be separate costs for Agentforce usage, integrations, data preparation, implementation, and training. Because Salesforce pricing and contracts vary by customer, businesses should calculate the full cost of the Agentforce and Data 360 setup rather than looking at the Data Cloud price alone.
Yes, there is an important difference between enabling Data 360 and fully implementing Data 360. Salesforce says Data 360 must be enabled for Agentforce, but a full implementation involves connecting external sources, mapping data, unifying customer profiles, and setting up capabilities such as identity resolution and data transformations. A smaller Agentforce use case may not require that full data-unification project at the beginning.
Data Cloud becomes more valuable when an Agentforce agent needs information that is spread across Salesforce and other systems. For example, an agent may need customer records, purchase history, support cases, marketing activity, or external data to answer a request properly. If your use case only needs a small amount of information already available in Salesforce, a large Data 360 implementation may not provide enough additional value to justify the investment.
For companies planning to move from a small Agentforce pilot to larger production use cases, it is worth reviewing the data architecture before scaling. Data 360 can give agents access to unified data and support capabilities such as data libraries, RAG, and real-time access to information outside the core Salesforce records. Starting with a focused pilot and then expanding the data layer based on actual gaps can help avoid paying for data infrastructure that the business does not yet need.
Get Started
Share a few details about your goals: a Mirketa consultant will follow up within one business day.