Fresh from Dreamforce 2026 — Agentforce, configured properly
Salesforce built Dreamforce 2026 around the "Agentic Enterprise" — seven pre-built AI agents and a keynote with Anthropic on Salesforce in Claude. Turning an agent on is easy. Configuring it to actually fit your business, your data, and your guardrails is the part that takes real Salesforce architecture experience.
The seven agents
Agentforce launched with seven named, pre-built agents, each built for a specific job function and connected directly to your Customer 360 data.
Turning one on takes a few clicks. Configuring it to make the right calls on your actual data, without overstepping what it should decide alone, is the work that determines whether it helps your team or creates a new mess to clean up.
Core capabilities
The parts of an Agentforce rollout that determine whether it actually works.
Why Radix2
These agents are only as good as the Customer 360 data underneath them. We have configured that data model 38 times — we know exactly where it usually breaks.
An agent given too much autonomy on live customer data is a liability, not a shortcut. We scope every rule in Agent Script before an agent goes live, not after something goes wrong.
Agentforce agents inherit your org's sharing rules and field-level security. We verify that inheritance is actually doing what you expect before rollout, not assume it.
We are not paid by Salesforce to push agent adoption. Our job is telling you honestly which of the seven agents fit your business today, and which are not worth the setup time yet.
How we deliver
We walk through your actual customer service, sales, and operations workflows and map them against the seven pre-built agents to find genuine fits, not assumed ones.
We audit the objects, fields, and record types the chosen agents will read and write, and clean up whatever would otherwise feed the agent inconsistent data.
We write the deterministic rules governing what each agent can decide autonomously, what requires human approval, and what it should never touch.
Agents go live to a small group first, with monitoring on their decisions before a full rollout — especially for long-horizon agents running unattended for days or weeks.
We document every rule and configuration decision so your own admin can maintain and adjust the agents after we are done, not depend on us indefinitely.
Frequently asked questions
Agentforce is Salesforce's lineup of seven pre-built, named AI agents (Casey, Paige, Carter, Hunter, Marshall, Piper, Fin), each built for a specific business function like customer service, sales, or supply chain. No — most businesses get real value from one or two agents configured well before it makes sense to add more.
Agent Script is Salesforce's language for defining deterministic rules alongside an agent's AI reasoning — it is how you constrain what an agent can decide on its own versus what needs a human. Writing these rules well requires understanding both the scripting language and your actual Salesforce data model, which is where a lot of DIY configurations go wrong.
They run on your existing Customer 360 data model, so their quality depends entirely on how clean and consistent that data model already is. Part of our setup process is auditing and aligning the data these agents will actually read and act on before configuring their behavior.
That is exactly what Agent Script guardrails are for — defining what an agent can decide autonomously versus what always requires human approval. We scope those boundaries before any agent goes live, and stage rollouts to a small group first so mistakes are caught early, not after a full deployment.
Agentforce licensing is separate from standard Salesforce editions — confirm your specific entitlements with your Salesforce account team. We can help you understand what configuration work is worth doing once licensing is confirmed.
A single, well-scoped agent (for example, inbound lead qualification) can go from assessment to a staged rollout in 3 to 4 weeks. Configuring multiple agents with interdependent rules takes longer — we scope the real timeline after the process and data model assessment, not before.
Dreamforce 2026 (September 15-17, San Francisco) centered on what Salesforce calls the "Agentic Enterprise" — Agentforce 360 across Sales, Service, and industry clouds, plus a keynote conversation between Salesforce CEO Marc Benioff and Anthropic CEO Dario Amodei on "Salesforce in Claude," which is moving from pilot to open beta this September. None of that changes the core work: picking the right agents for your business and configuring them properly still takes real Salesforce architecture experience.
Let's figure out which of the seven agents actually fit your business, and set them up with guardrails that keep them useful instead of risky.
Discuss your process, your data model, and which agents are worth it.