What Is AIforce and Why Does It Matter?
AIforce is the Day 1 headline. It is a live interface layer that exposes Salesforce's data, workflows, business logic, semantics, permissions, security and governance to any AI interface — not just the Salesforce UI. Users can work with Salesforce inside Claude, Slack, Google Gemini Enterprise, Amazon Quick or Lightning.
Built on Zero Data Retention, Salesforce explicitly denied that customer data trains third-party AI models. The architecture rests on Data 360, Customer 360 and Agentforce. Benioff framed it as an interface revolution, comparable to the shifts from DOS to GUI to web to mobile.
The package includes Claudeforce (37 prebuilt sales skills powered by Anthropic Claude), Slackforce (Salesforce context in Slack), Agentforce Coworker (an AI teammate in Lightning) and a Headless Toolkit (MCPs, APIs, plug-ins and a skills library for builders).
Koa: A CRM-Specific Reasoning Model
Salesforce and NVIDIA announced Koa — Salesforce's first proprietary CRM reasoning model for Agentforce. It is built by post-training NVIDIA Nemotron 3 Super (an open-weight foundation model) with supervised fine-tuning and reinforcement learning, including Group Relative Policy Optimization.
The training corpus is synthetic: it simulates 27 years of CRM deployment knowledge across 14+ industries. Salesforce explicitly states no customer data was used for training, and inference runs inside Salesforce-controlled infrastructure.
Salesforce reports that Koa matches or exceeds leading models on its CRM Benchmark with 3× fewer errors on CRM actions like updating an opportunity, routing a case or scheduling a follow-up. Read KVP's full Koa analysis →
Partnerships: Google Cloud, Siemens and AWS
Google Cloud: Hyperforce now runs natively on Google Cloud with live production traffic. Salesforce will migrate select US customers in Q4 2026. The partnership enables cross-platform agent reasoning between Salesforce and Google Gemini Enterprise.
Siemens: Agentforce is combined with Siemens' Teamcenter Service Lifecycle Management, bringing engineering-grade digital-twin answers into sales and service workflows. Salesforce says Agentforce now qualifies 100% of inbound leads for Siemens' 18,000-person sales force.
AWS: Expanded partnership tied into the multi-surface AIforce strategy, with Amazon Quick as a target interface.
Missionforce: Nemotron-based models and NVIDIA accelerated computing for government and regulated customers, deployable in private clouds and air-gapped networks.
Customer Story: Adecco Group at Enterprise Scale
Adecco Group announced a global rollout of Agentforce Coworker across 40+ countries, reaching 27,000 employees. Powered by Anthropic's Claude, Coworker draws on 2.5M+ agent-candidate interactions since April 2025. Agentic AI already covers recruitment workflows in 10 countries — roughly 50% of Adecco's revenue. CEO Denis Machuel joined Benioff on the keynote stage.
This is one of the largest enterprise Agentforce deployments disclosed publicly. It moves agentic AI from pilot decks to a global operating model — and signals that the technology can scale when the data, permissions and workflow design are ready.
Key Takeaways for Business Leaders
The interface is no longer the product
AIforce says Salesforce's value is the data model, permissions and logic — not the screen. Users will increasingly work inside Claude, Slack or Gemini rather than Salesforce itself.
Specialised reasoning beats general chat for CRM
Koa gives Agentforce a model trained on CRM sequences, tool calls and policy — not broad conversation. The promise is fewer broken hand-offs when an agent must reason, act and update records across multiple turns.
Infrastructure choice is expanding
Hyperforce on Google Cloud (in addition to AWS) means customers can choose where their Salesforce data lives. Missionforce extends to private clouds and air-gapped networks for regulated and government workloads.
Enterprise-scale proof is emerging
Adecco's 27,000-employee rollout across 40+ countries, Siemens' 18,000-person sales force and the partnership with Google Cloud signal that agentic AI is moving from pilots to production.
How These Announcements Impact KVP as a Salesforce SI
As a Salesforce Platinum Partner, KVP sees six immediate implications for our practice and our customers:
Governance is the new consulting practice
When actions start outside Salesforce, customers need audit trails, approval controls and policy enforcement that travel with every surface. SIs should lead with governance design, not feature demos.
Skill libraries replace app builds
AIforce shifts value from building Lightning pages to configuring headless skills, MCPs and plug-ins. SIs need packaged skill libraries and reusable governance frameworks.
Multi-cloud architecture planning
With Hyperforce on Google Cloud and AWS, plus Missionforce for regulated sectors, SIs should help customers plan infrastructure choice, portability and data residency.
Industry reference architectures
The Siemens–Teamcenter pattern is a template SIs can replicate for other manufacturing and industrial clients — embedding domain-specific data into Agentforce workflows.
Cost and pricing advisory
Per-user, per-month Claudeforce billing and unresolved AI cost questions mean customers should budget conservatively and negotiate consumption terms before broad rollouts.
Pilot-first methodology
Start with one bounded workflow, define the evaluation set, test exceptions and permissions, compare quality and cost against the current model, and scale only after evidence.
What This Means for Customers
Work where you already are
AIforce means sales, service and ops teams can act on Salesforce data inside Claude, Slack or Gemini — without context switching to a separate CRM tab.
Better agent reasoning for CRM
Koa gives Agentforce a model specialised for multi-step CRM work — qualifying opportunities, routing cases and scheduling follow-ups with fewer errors.
Infrastructure flexibility
Hyperforce on Google Cloud and AWS means customers can choose where their data lives. Missionforce extends to private clouds for regulated industries.
Trust and governance first
Zero Data Retention, Einstein Trust Layer and inherited permissions mean agentic workflows can meet enterprise compliance — when designed correctly from the start.
KVP View: Design for the Surface, Govern the Workflow
The interface is becoming composable. AIforce means Salesforce's value is the data model, permissions and logic — not the screen. Customers should plan for a world where work starts in Claude, Slack or Gemini, not in a Lightning tab.
Governance must travel with every action. When work originates outside Salesforce, audit trails, approval controls and policy enforcement must hold up across every surface. That is a design question, not a feature toggle.
Budget for per-user pricing and cost uncertainty. Claudeforce is billed per-user, per-month. Customers should model the cost of agentic AI before committing to a broad rollout — and negotiate consumption terms early.
Pilot one bounded workflow first. Pick a process with clear users, inputs and next actions. Test exceptions and permissions across surfaces. Compare quality and cost against your current model. Scale only when outcomes improve.
Dreamforce Day 1 in Six Slides
Download the customer-ready deck, share one slide, or expand it for mobile reading.
Dreamforce Day 1 · KVP POVAIforce: The Interface Revolution
Salesforce's data, logic and governance exposed to any AI surface — Claude, Slack, Gemini, Lightning.
Dreamforce Day 1 · KVP POVKoa × NVIDIA
Salesforce's first proprietary CRM reasoning model, post-trained on Nemotron 3 Super for agentic workflows.
Dreamforce Day 1 · KVP POVGoogle Cloud, AWS, Siemens
Hyperforce on Google Cloud, cross-platform agent reasoning, and industrial digital-twin sales with Siemens.
Dreamforce Day 1 · KVP POVFrom suggestions to autonomous action
Agentforce Coworker embedded in Lightning, Claudeforce with 37 prebuilt sales skills, Slackforce in every channel.
Dreamforce Day 1 · KVP POVAgentic AI goes vertical
Siemens brings engineering-grade data into sales. Adecco rolls out Coworker to 27,000 employees across 40+ countries.
Dreamforce Day 1 · KVP POVGovern the workflow, not just the interface
Design governance that travels with actions across every surface — then pilot one bounded process and measure.