Salesforce Agentforce gives CIOs access to powerful AI agents inside the Salesforce ecosystem. But the bigger question is not whether your organisation can add agents. It is whether your operating model is ready for them.
For organisations exploring a Salesforce Agentforce alternative, agentic automation built around Model Context Protocol (MCP) offers a broader approach. Instead of limiting intelligence to one platform, it allows businesses to connect systems, orchestrate workflows and deploy AI where it creates the most value.
The decision is not about choosing between Salesforce and AI. It is about deciding whether your automation strategy is built around a vendor, or around how your business actually works.
The real problem is not a lack of AI agents
More technology will not fix broken processes
CIOs are under pressure to show progress with AI while balancing security, governance, cost and operational risk.
It is tempting to look at Agentforce and think the answer is simple: Salesforce already runs key business processes, so adding AI agents must be the fastest route to transformation.
Sometimes it is.
If your customer service workflows, sales processes and data structures are already mature, adding Salesforce native agents can create meaningful improvements quickly.
But many organisations are starting from a different place.
They have duplicated processes across teams, disconnected systems, unclear ownership and limited visibility into how work moves from one department to another.
In that environment, adding more agents can make the problem bigger.
A sales team might have an AI agent updating Salesforce opportunities, while delivery teams are still manually creating projects elsewhere. Finance might still be chasing information through spreadsheets. Operations might not know which AI actions were taken or why.
The issue is not the agent.
The issue is the operating model around it.
The CRM is not the operating model
Your business does not run inside one platform
Many organisations naturally treat their CRM as the centre of everything. If something happens in Salesforce, it feels official. If it happens somewhere else, it often becomes invisible.
That thinking creates an easy assumption:
If Salesforce has agents, Salesforce should own all automation.
But your operating model is not your CRM.
Your operating model is the complete way work moves through your organisation, from sales and finance to delivery and customer support.
A customer deal does not become successful because it exists in Salesforce. It becomes successful because information moves between teams, decisions happen at the right time and people know what needs to happen next.
The smarter approach is to decide where automation belongs based on the workflow, not the software provider.
Salesforce can remain the source of truth for customer and revenue data without becoming the only place where intelligence exists.
A different approach: choosing the right tool for the right job
Kodah’s “Horses for Courses” framework
Kodah’s approach starts with a simple idea: use the right technology for the work that needs to happen.
Before choosing a platform, organisations need to answer three questions:
- What workflow are we improving?
Are we trying to speed up sales handovers? Reduce manual reporting? Improve customer onboarding? Remove repetitive operational tasks?
The problem comes first.
- Which systems are already part of that workflow?
Most businesses do not run on one platform.
Salesforce might manage revenue operations. ClickUp might manage delivery. Finance systems might manage billing. Teams might use different tools for different parts of the business.
- Where should intelligence sit?
The best place for AI is not always the place where the data started.
It is the place where AI can make the biggest improvement safely.
In practice, this could look like:
- Salesforce managing customer data, opportunities and revenue processes.
- ClickUp managing projects, ownership and delivery workflows.
- Claude providing reasoning, analysis and decision support.
- MCP connecting systems so agents can understand context and take action across platforms.
Agentforce can still play an important role.
The difference is that it becomes part of the operating model, not the entire operating model.
Agentforce vs agentic automation in practice
A simple sales to delivery example
Imagine a company closes a major customer deal.
The sales team marks the opportunity as “Closed Won” in Salesforce.
With a Salesforce focused approach, Agentforce could identify the opportunity, trigger follow up actions and support teams working inside that ecosystem.
For straightforward workflows, this can work well.
But what happens when the process crosses multiple teams?
Delivery needs a project created in ClickUp. Operations needs visibility. Finance needs information. Leadership needs reporting. Governance teams need an audit trail.
This is where a broader agentic automation approach becomes valuable.
The workflow could look like this:
Observe
Salesforce records the Closed Won opportunity.
MCP allows the relevant information to be shared with the agentic layer in a structured way.
Think
Claude analyses the opportunity details, checks requirements, summarises key information and determines the next actions based on business rules.
Act
The agent creates a delivery project in ClickUp, assigns tasks, updates relevant systems and keeps Salesforce informed of progress.
The result is not a replacement for Salesforce.
It is a connected workflow where each system does what it does best.
Salesforce remains the revenue system of record.
ClickUp supports execution.
Claude provides intelligence.
MCP enables communication between systems.
What CIOs should consider before choosing an approach
The key questions are not about features
The biggest mistake organisations make is comparing AI tools before understanding the business problem they need to solve.
Before investing in agents, CIOs should ask:
- Can we clearly map how work moves across departments and systems?
- Do we know who owns AI decisions and outcomes?
- Can we explain what an AI agent did, why it did it and what data influenced that decision?
- Are we solving a real operational problem, or adding AI because a platform introduced a new feature?
These questions matter because AI without governance creates uncertainty.
AI without process clarity creates more complexity.
AI without the right operating model creates expensive experiments that never reach production.
The practical decision for CIOs
Choose the model that supports how your business actually works
Agentforce is a powerful option for organisations that want to extend Salesforce capabilities.
But it should not automatically become the foundation for every automation decision.
A stronger strategy is to start with your operating model, understand where work happens and then decide where different AI capabilities belong.
For some workflows, that may be Agentforce.
For others, it may require a combination of Salesforce, ClickUp, Claude, MCP and other systems working together.
The goal is not to have more agents.
The goal is to have agents that improve the way your business operates.
Start with your operating model, not your AI tool
If you are evaluating Salesforce Agentforce or looking at a broader agentic automation strategy, the first step is understanding your workflows.
Kodah’s Agentic Automation Discovery session helps CIOs identify:
- Where automation opportunities exist across the organisation.
- Which systems should be connected and where agents should operate.
- What governance, security and operating model changes are needed before scaling AI.
Bring us your current workflows, systems and challenges.
We’ll help you map where Agentforce fits, where broader automation is needed and what a practical path to production looks like.