Frequently asked questions about Salesforce, Agentforce and AI

Agentforce is Salesforce's AI agent platform: assistants that query your data, reason over your processes and trigger governed actions, right inside your CRM. Unlike a traditional chatbot, an agent can act — create a record, follow up on a case, orchestrate a step. Since the Salesforce–Anthropic partnership of August 2026, Claude powers Agentforce's reasoning engine.

Three criteria really matter. Depth of expertise — the number of certified consultants, not just the number of consultants. Verifiable customer satisfaction — public reviews on the AppExchange are collected by Salesforce, not by the integrator, so they are hard to fake. And team continuity : ask who will actually deliver, and whether it will be the same people from start to finish. A large integrator reassures with its size, a small firm with its closeness — look for the one that offers both.

Those whose business processes have outgrown what a spreadsheet or a single tool can handle: several teams handing cases to one another, customer data scattered across systems, or a need to run marketing, sales and service from the same numbers. Complexity matters more than size: we work with growing SMEs as well as multi-entity groups.

It depends on the scope, but our formats are calibrated: a technical and functional audit takes 8 to 12 days spread over 3 to 4 weeks, a supervised pilot 15 to 20 days over 6 to 8 weeks. A full rollout is measured in months. We always work in short phases with a deliverable at each step, rather than one big project whose results only show at the end.

Our starter packs range from €4,500 to €15,000 excl. VAT depending on scope. Beyond that, our services are quoted individually. Every proposal follows an initial scoping conversation, with no commitment.

Technically yes — but rarely without an audit first. An agent inherits exactly the permissions you give it. An over-provisioned profile, harmless behind a slow interface navigated by hand, becomes real exposure as soon as an agent can query everything in three seconds. So we always start by auditing the exposed permissions and automations, then open up write access one use case at a time.

It is not a choice, it is a division of labour. Agentforce handles repeatable, high-volume processes inside your Salesforce org. Claude handles exploratory, multi-source reasoning on the user's side. The same technical plumbing and the same permission model serve both: the real question is which use case goes where — and that is exactly what our AI FinOps scoping is for.

Claude is served via Amazon Bedrock inside the Salesforce Trust Boundary: your data never leaves Salesforce's trust perimeter. This is often what unlocks a security review in banking, insurance or wealth management. On top of that, we set up traceability of agent actions (Field Audit Trail), monitoring (Event Monitoring) and blocking rules on sensitive actions.

Your data, partly — but not your metadata. Your Flows, profiles, permission sets and most of your configuration are not covered by the native backup. And free export tools export without restoring: they preserve neither the configuration nor the relationships between records. The deciding question: when did you last actually run a restore?

Yes. 2Gezer is Qualiopi-certified for training, which makes our sessions eligible for funding by your OPCO, subject to the eligibility conditions in force. Our training is delivered in-house, with hands-on workshops run on your own environment rather than a textbook case.

No. Our teams are spread across Bordeaux, Paris and Aix-en-Provence, and we work throughout France. Our consultants work on site or remotely depending on the project phase — scoping workshops benefit from being in person, while implementation runs very well remotely.