Inside the Five-Stage Lifecycle Powering the Agentic Enterprise

In the era of the Agentic Enterprise, the familiar rhythms of traditional software development are giving way to something new. Where once teams followed a linear cycle of planning, building and releasing, today they’re designing intelligent agents that learn, adapt and evolve in real time. This shift demands a new kind of lifecycle: one built for experimentation, continuous optimisation and measurable impact.

That evolution has led to the development of the Agentforce 360 agent development lifecycle: a simple, structured approach to designing, testing and improving agentic agents that deliver meaningful value.

Informed by experience across thousands of Agentforce deployments both worldwide and in Australia & New Zealand, the lifecycle helps teams move from early experimentation to adoption at scale, ensuring that every agent is purposeful, governed, and continually improving.

For a bigger picture look at Agentforce 360 and all the updates announced at Dreamforce 2025, visit Dreamforce Decoded for Australia & New Zealand, your ultimate virtual guide to Dreamforce. Explore exclusive local insights and content, and catch the very best sessions on-demand on Salesforce+.

Now, let’s explore the five stages that constitute the agent development lifecycle.

Salesforce’s Jonathan Hatcher and Jessica Raicebe share the five stages of the Agentforce 360 agent development lifecycle, a proven framework built from lessons learned across thousands of global deployments.

1. Ideate

Every great agent starts with a great use case. The ideation stage is about identifying high-value opportunities where agents can make a measurable difference, like automating repetitive workflows, summarising information or surfacing insights for customers and employees.

The key is balance: focus on low-risk, high-reward scenarios that deliver immediate value while keeping human oversight where it matters most.

2. Configure

Once you’ve defined the “why,” it’s time to shape the “how.” Configuration is all about context engineering: setting clear, unambiguous instructions that guide your agent’s behaviour. Even subtle contradictions can create confusion for a large language model (LLM). Precision, structure and clarity at this stage make all the difference.

3. Test

Unlike traditional software, agentic systems don’t always behave exactly the same way every single time. That means testing isn’t about a definitive pass or fail score, it’s about continuously learning and improving. Teams should simulate real-world scenarios, capture both expected and unexpected outcomes, and then refine prompts, contexts and rules to build resilience into the system.

4. Deploy

When it’s time to go live, start small and scale with intention. Deploy sequentially across channels, whether it’s web to mobile or messaging, and establish feedback loops that help you continuously measure performance and user satisfaction. Agent deployment isn’t the finish line. When you become an Agentic Enterprise, deployment is the start of live, real-time iteration in the field.

5. Optimise

Agents are dynamic digital teammates that learn, adapt, evolve and augment your human team members. Ongoing optimisation is how you deliver ongoing value over time. This means building observability into every deployment, monitoring performance, alleviating pressure points and diagnosing root causes when something doesn’t work as expected. Leading teams are already using observability data to fine-tune context engineering and data pipelines, ensuring every improvement compounds into a smarter, more reliable agent.

The agent development lifecycle represents a necessary mindset shift for the Agentic Enterprise. It’s about moving fast, learning continuously and treating agents not as set-and-forget tools, but as dynamic systems that can evolve and drive real transformation.

To dive deeper, explore our white paper featuring 25 lessons learned from real customer deployments across industries and geographies, and see how your organisation can put these principles into practice.