AI pillar — Agentforce & Claudeforce
Get your Salesforce agentic journey right
Agentforce, Salesforce's native agent, and Claudeforce, the Salesforce x Anthropic partnership announced in August 2026, are together changing how you access your CRM: an agent — native or powered by Claude — can now query your data, reason over your pipeline and trigger governed actions. Our conviction: the value of an agentic deployment is decided before the first demo — in the permissions audit, the traceability of actions and the governance framework.
Discovery
Identify and prioritise your use cases
Assessment
Know where you stand and what exposes you
Compliance
Make the org auditable and the team self-sufficient
Proof of value
Measure real usage and real cost
Scale-up
Encode your business rules, frame what comes next
Documented use cases
What AI is already changing, by function
The examples below are Agentforce and Claudeforce deployments publicly documented by Salesforce and Anthropic, not 2Gezer projects. We share them to give a concrete idea of what is possible, before building with you the use cases that make sense for your organisation.
Salesforce's native agentic engine, built for repeatable, high-volume tasks in your org: ticket qualification, field updates, actions triggered by business rules.
The Salesforce x Anthropic partnership announced in August 2026: Claude on the human side for exploratory, multi-source reasoning (synthesis, preparation, decision support) and, optionally, as the reasoning model inside Agentforce.
It is not a choice between the two: the right mix depends on your use case — which is exactly what our Offer 6 below is for.
Qualification, meeting prep, pipeline
- UberAgentforce — inbound requests qualified automatically: RFPs processed 83% faster, +60% conversion on leads handled by the agent.
- Prudential FinancialAgentforce — meeting summary and follow-up agent: roughly half a day saved per week per sales rep.
- Salesforce in ClaudeClaudeforce — 37 ready-to-use sales skills (morning brief, meeting prep, pipeline review), already piloted at GitLab, Siemens and Legora.
Ticket resolution and deflection
- XeroAgentforce — 62% of support requests resolved autonomously, across more than 800,000 tickets a year.
- CVS HealthAgentforce — 58% chat containment, up to 12% less call time.
- TrustpilotAgentforce — ticket backlog cut by 99%.
Personalisation and campaign creation
- 360 MMSAgentforce — campaigns created 16 times faster, using Data Cloud for segmentation.
- Fewer named, quantified cases than in Sales or Service so far: an area where the evidence is still being built — a chance to be among the first to document it in your organisation.
Sources: Salesforce customer stories (Agentforce Customer Stories, Agentforce Metrics, 2026) and the Salesforce x Anthropic press release of 26 August 2026. Curious what it could look like for you? That is exactly what our Offer 0 below is for.
Our offers
Eight offers, sequenced to limit risk
AI use case mapping
You leave with a short list of use cases prioritised by value and complexity, built with your business teams — not guessed from a generic catalogue.
What we do
- Workshops with business leaders (Sales, Marketing, Service, Ops) to map repetitive, time-consuming or high-stakes processes
- Comparison with use cases already documented in the Agentforce/Claude ecosystem, to challenge and enrich the list
- Scoring of each use case on business value, technical complexity and governance risk
- Prioritisation workshop with business and IT sponsors
Deliverables
- Map of AI-eligible processes, by function
- Prioritised use case short list (value / effort / risk)
- Recommendation of the 2 to 3 top-priority cases, passed on to Offer 1 for technical validation
Agentic readiness audit
You will know precisely what an agent could read, modify and trigger in your org today — and what needs fixing before you let it.
What we do
- Review of profiles, permission sets and sharing
- Mapping of exposed automations (Flows, Apex, validation rules)
- Assessment of data quality and structure
- Inventory of existing connections and MCP servers
- Regulatory exposure analysis (GDPR, traceability)
Deliverables
- Audit report with risk matrix
- Prioritised, costed remediation plan
- Recommended pilot scope (read-only)
- Steering committee readout
Supervised Claudeforce pilot
A quantified answer to two questions: what concretely changes for your teams, and what it really costs in your organisation.
What we do
- Connection: Claude enabled as the reasoning model in Agent Builder, or via Salesforce in Claude on the team side; dedicated scope, read-only start
- Pilot cohort (5-10 users), 2-3 high-signal use cases
- Instrumentation of consumption and usage
- Gradual opening of write access, step by step
Deliverables
- Operational, documented pilot environment
- Usage and consumption dashboard
- Value report and real cost projected at scale
Supervised Agentforce pilot
A native Agentforce agent up and running on your repeatable, high-volume processes, configured in your org on your business rules.
What we do
- Scoping of Topics and Actions with the relevant business teams
- Configuration in Agent Builder, grounding on your data via Data Cloud
- Pilot cohort (5-10 users), 2-3 high-volume use cases
- Regression testing of existing automations (Flows, Apex) triggered by the agent
Deliverables
- Configured, documented Agentforce agent
- Topics and Actions tested on real cases
- Usage report and extension recommendation
Business skill engineering
Your business rules encoded, not generic prompts — so the answer is the same whoever asks the question.
What we do
- Workshops to capture precise business definitions
- Decisions on reference fields and priority rules
- Writing, testing and versioning: Claude skills or Agentforce Topics/Actions, depending on the chosen platform
- Integration into the governance framework
Deliverables
- 3 to 5 documented, tested and versioned skills
- Maintenance guide and regression test suites
- Handover session for administrators
Agent governance & observability
Know what an agent did, and be able to prove it — using licences you may already be paying for.
What we do
- Field Audit Trail: long-term history of fields touched by agents
- Event Monitoring: tracking of access, APIs, exports, dashboards
- Transaction Security Policies: blocking, step-up authentication, notification
- Privacy Center and connector register
Deliverables
- Targeted Shield configuration
- Operational transaction security policies
- Agent action monitoring dashboard
- Connector governance charter
Administrator training
Your administrators become self-sufficient on Privacy Center, Field Audit Trail and Event Monitoring, with workshops on your own environment.
What we do
- Half-day 1: data governance, Privacy Center + workshop
- Half-day 2: Field Audit Trail, Event Monitoring + governance plan
- Hands-on sandbox workshops, with cases drawn from your context
Deliverables
- Delivery of the 2 half-days (7 hours)
- Training materials and certificate of completion
- Report on configuration issues identified
AI FinOps & agentic roadmap
A cost model, a clear rule for deciding between Agentforce and Claude by use case, and a 12-month roadmap.
What we do
- Cost model: consumption, licences, implementation, maintenance
- Use case triage matrix (repeatable vs. exploratory)
- Target architecture note and 12-month roadmap
Deliverables
- Quantified, configurable cost model
- Prioritisation matrix (value / effort / risk)
- Architecture note, roadmap, readout deck
Scope, duration and budget are defined together based on your context — let's talk. Excludes Salesforce licences (Shield, Privacy Center, Agentforce) and Claude inference costs, billed on consumption.
Training may be eligible for OPCO funding, subject to eligibility. Salesforce in Claude is in beta at the time of writing: our commitments cover our support and method, not the product's availability.
An agent inherits the permissions you give it
Audit before demo. Let's talk about your Salesforce agentic journey.