Generative AI is shifting from hype to hard ROI. During our recent session, Greg Wasowski, SVP of Consulting & Strategy at Aquiva Labs, and Kam Sangha, Customer Success Director at Salesforce, walked through what it takes to operationalize Agentforce responsibly and at scale.
Organizations want sharper efficiency, better customer experiences, and a faster path to value. Agentforce is giving them exactly that when the right foundations are in place.
You can watch the full recording here, or checkout the recap below.
1. Consumption-based pricing works when you understand actions—not features
Salesforce’s move to Flex Credits has been one of the biggest shifts in the platform’s commercial model. Instead of licensing seats, you now pay for what agents do.
Kam explained it plainly:
“Rather than selling big bundles of usage, Flex Credits let us sit with customers and map out what actions their agents will take. You get transparency and real control over cost.”
An action—like looking up an order, summarizing a case, or updating an address—costs 20 credits. Internal-facing agents can still use per-user licensing, but externally-facing agents consume credits based on usage.
For customers running serious POCs, Salesforce Foundations changes the game. Enterprise Edition customers get:
- 200k Einstein Request Credits
- 1k Conversations for Agentforce
- 250k Data Services Credits
- 1 TB Data Storage
- 10k Segmentation & Activation Credits for Data Cloud
This lets teams prove value before they ever sign a check.
2. A successful AI Center of Excellence starts with clarity—not technology
A COE isn’t optional anymore. It’s now the operating system for AI inside the enterprise.
Greg put the reality on the table:
“This is a new world. These are the same pillars you’d use to stand up any Salesforce COE—strategy, governance, best practices—but applied to agents that can take real action.”
The strongest COEs share the same structure:
A defined AI mission aligned with business strategy
Scalable data and integration architecture
Clear governance and risk oversight
A repeatable way to source, test, and prioritize use cases
Training that raises AI literacy across teams
And while many organizations are still experimenting, the maturity curve is fast. Kam noted that what started as admin-led tinkering has turned into structured AI leadership roles.
His observation:
“We’re now seeing formal COE leadership emerge. People are realizing experimentation only gets you to a point—you need ownership.”
3. Clean, connected data is the real unlock
Agentforce can only perform as well as the data that grounds it. Kam and Greg were aligned here: don’t turn Agentforce into a data project, but don’t ignore data maturity either.
Kam shared what he sees across enterprise customers:
“Start small with the data you need. Don’t try to bring everything in from day one. You can grow into Data Cloud over time.”
Early agents can succeed using what’s already in Service or Sales Cloud. But higher-value use cases—semantic search across knowledge, personalization, cross-system workflows—require more.
Greg reinforced the same point:
“Your first POC shouldn’t demand a full Customer 360 that leverages data from multiple systems. But long-term, unified profiles unlock a different level of intelligence.”
4. Testing and governance matter more than the build
Everyone loves building agents. Few teams dedicate the equivalent effort to testing. That’s where risk—and failure—lives.
Kam didn’t mince words:
“The build is this much. The testing is bigger. You’ve got to think about every scenario someone might throw at the agent.”
A modern testing cycle includes:
Functional output validation
Guardrail and policy enforcement
Bias, toxicity, and safety checks
Human handoff behavior
Performance and reliability under load
Salesforce’s new Agent Test Center, announced at Dreamforce, gives teams much stronger visibility. But the COE must still own the governance framework.
Greg connected governance directly to value:
“If governance isn’t solid, you won’t get to scale. It’s that simple.”
5. The organizations winning with Agentforce start small—and go fast
Both speakers pointed to a clear behavior pattern among successful customers.
- They don’t overthink the starting point.
- They don’t run 40-use-case workshops before building anything.
- They ship an MVP agent that solves one real problem.
Greg framed it this way:
“Start with something valuable but contained—often service, quoting, or field ops. Prove or disprove fast, then expand.”
Kam added his own experience:
“We built a ‘dumb’ agent first. No integrations, simple data. It proved the concept. Then we built the real version. That’s the model.”
A typical journey looks like:
- MVP: One agent, one channel, small user group
- Custom Agent: More actions, deeper integrations, broader reach
- Scale: Multi-channel, multi-role agents driving measurable business value
The north star? Efficiency and speed. Organizations want to do more without adding headcount, and Agentforce is proving it can deliver.
Where to go from here
If you joined the live session, you already have the foundation. If not, here are the next steps Greg and Kam emphasized for every org:
Align use cases with business value
Prepare the minimum data needed to get started
Stand up your COE early—even if lightweight
Build your first agent and test aggressively
Measure the impact and refine
- Follow the AIM framework to help with change management
Agentforce is reshaping how work gets done. Organizations that start now will be the ones defining how AI scales across their business.
If you want the pricing breakdown, use-case framework, or COE templates discussed in the session, reach out—we’re happy to share.
Author
Aquiva Labs
When you count on Salesforce, count on Aquiva.
