This chapter covers a NEW objective in the CCNA v2 exam (200-301 v2.0, 2026 blueprint): 'Generative AI Prompting for Network Tasks' (Objective 5.2). As networks grow more complex, network engineers are increasingly using AI assistants—like ChatGPT, Claude, or Cisco's own AI tools—to automate troubleshooting, generate configurations, and analyze logs. However, these tools only produce useful results if you ask the right questions. This chapter teaches you how to craft effective prompts for network tasks, a skill now tested on the CCNA because Cisco recognizes that AI fluency is essential for modern network operations. Mastering this objective will help you pass the exam and work smarter in real-world engineering roles.
Jump to a section
Imagine you are a master chef in a busy restaurant kitchen. You have a talented sous-chef (the generative AI) who can execute any recipe perfectly, but the sous-chef has no taste buds and no intuition—they follow instructions literally. If you say 'make a sauce,' they might produce ketchup, because you didn't specify 'tomato basil reduction.' To get the perfect sauce, you must give precise, step-by-step instructions: 'Take 2 cups of crushed San Marzano tomatoes, simmer with 3 cloves of minced garlic in olive oil for 10 minutes, then add 10 fresh basil leaves and simmer 2 more minutes.' This is exactly how prompting works for network tasks. The AI is like the sous-chef: it has vast knowledge of Cisco commands, protocols, and troubleshooting steps, but it cannot read your mind. If you ask 'fix my network,' it might suggest rebooting a router—not because that's the best fix, but because your prompt was vague. You need to provide context (e.g., 'I have a BGP peering issue between Router A and Router B, the link is up but the session is idle, show ip bgp summary shows State = Idle'), specify the desired output format (e.g., 'List the top 3 causes in order of likelihood, with the corresponding debug commands'), and set constraints (e.g., 'Assume IOS XE 17.3, no SD-WAN'). Just as a chef learns to communicate with their sous-chef through experience, a network engineer learns to craft prompts that yield accurate, actionable results. The exam tests your ability to do this efficiently—a skill that saves hours of manual troubleshooting.
What Is Generative AI Prompting for Network Tasks?
Generative AI prompting is the practice of formulating natural language queries to obtain network-related outputs—such as configuration snippets, diagnostic steps, or log analysis—from large language models (LLMs) like GPT-4, Claude, or Cisco's AI Assistant. In the CCNA v2 exam, you won't use a live AI; instead, you'll be given a scenario and asked to select the best prompt from multiple choices. The exam objective (5.2) is defined as: 'Given a network scenario, select the most effective generative AI prompt to produce a desired output.' This means you must understand what makes a prompt clear, specific, and actionable.
Why Cisco Added This Objective
Cisco added AI prompting to the CCNA because network engineers now routinely use AI tools for tasks like: generating ACL configurations, interpreting 'show tech' output, and suggesting troubleshooting paths. The 2026 blueprint allocates 10% of the exam to 'AI and Network Operations and Management,' reflecting the industry shift toward AI-assisted network engineering. Candidates who studied v1.1 are missing this entirely—they learned to troubleshoot manually, but not how to leverage AI as a force multiplier. On the exam, you'll need to recognize prompts that include: role assignment (e.g., 'You are a Cisco network engineer'), context (e.g., 'We have a /30 subnet between two routers'), task (e.g., 'Generate a standard ACL to permit only 10.1.1.0/24'), and constraints (e.g., 'Use Cisco IOS syntax, no zone-based firewall').
The Anatomy of an Effective Prompt
An effective network prompt has four components: - Role: Assign the AI a persona (e.g., 'Act as a senior network engineer with 10 years of Cisco experience'). - Context: Provide relevant network details (e.g., 'We have two Cisco 4321 routers running IOS XE 17.6, connected via a serial link with HDLC encapsulation'). - Task: Clearly state what you want (e.g., 'Generate a configuration to enable OSPF area 0 on both routers with network statements'). - Constraints: Specify format, syntax, or limitations (e.g., 'Use only IPv4, no redistribution, output as CLI commands in a code block').
A poor prompt like 'Write an ACL' might yield an extended ACL with random ports, while a good prompt like 'Write a standard numbered ACL for router R1 that permits only host 10.1.1.1 and denies all else, using access-list 1, and apply it to GigabitEthernet0/0 inbound' gives exactly what you need.
Common Prompting Pitfalls
Vague goals: 'Help me with OSPF' could mean anything. The AI might explain OSPF basics instead of generating a config.
Missing context: Without subnet masks, interface types, or IOS version, the AI may guess wrong defaults (e.g., assuming OSPF network type is broadcast when it's point-to-point).
Overloading: Asking for multiple unrelated tasks in one prompt (e.g., 'Generate an ACL, configure VLANs, and troubleshoot why BGP is down') confuses the AI and produces shallow answers.
Ignoring constraints: If you don't specify 'no NAT' or 'use named ACLs,' the AI might include features you don't want.
Exam-Relevant Prompt Examples
On the exam, you'll see prompts like these and need to choose the best one: - Scenario: A router is not forming an OSPF adjacency. Which prompt is best? - A: 'Why is OSPF not working?' - B: 'Act as a Cisco TAC engineer. Router R1 (IOS XE 17.3) has OSPF configured with network 0.0.0.0 255.255.255.255 area 0. The neighbor router R2 shows state EXSTART. Provide the top 3 causes and the show commands to verify each.' - C: 'Generate OSPF debug commands.' - D: 'Tell me about OSPF states.' The correct answer is B because it provides role, context, specific task, and output format.
Verification and Refinement
Even with a good prompt, always verify the AI's output. For example, if the AI generates an ACL, check that it doesn't accidentally deny necessary traffic. Learn to refine prompts iteratively: if the first output is too generic, add constraints like 'Use only IPv4 ACLs, no object groups.' The exam may test your ability to identify which follow-up prompt would best refine an initial output.
Interaction with Other Network Tools
Generative AI is not a replacement for traditional tools like ping, traceroute, or 'show' commands. Instead, it augments them. For instance, you might use AI to interpret the output of 'show ip ospf neighbor' and suggest next steps. The exam expects you to know when to use AI versus manual troubleshooting. Generally, use AI for: generating boilerplate configs, explaining unfamiliar error messages, and brainstorming possible causes. Use manual methods for: real-time verification, performance-critical changes, and security-sensitive tasks where you need full control.
Define the Network Scenario
Start by identifying the specific network problem or task. For example: 'A VLAN 10 interface on switch SW1 is down, but the physical port is up.' Write down the key details: device model, IOS version, relevant configurations, symptoms, and any recent changes. This context is essential because AI models need concrete information to generate accurate responses. Without it, the AI may produce generic advice that doesn't apply to your situation.
Assign the AI a Role
Explicitly tell the AI to act as a network expert. Example: 'You are a Cisco CCIE with expertise in troubleshooting Layer 2 issues.' This primes the AI to use technical language and focus on Cisco-specific solutions. In the exam, you'll see prompts that include role assignment—recognize that this is a best practice. Without a role, the AI might respond in simple terms or omit important details.
Provide Detailed Context
Include all relevant network parameters: interface types, IP addresses, subnet masks, routing protocols, VLANs, trunking mode, and any error messages. Example: 'Switch SW1 (Cisco 2960, IOS 15.2) has interface GigabitEthernet0/1 configured as access VLAN 10. The end device is a PC with IP 10.0.10.2/24. The interface is up/up but the PC cannot ping the default gateway 10.0.10.1.' The more context, the better the AI's output. On the exam, the correct prompt will include specific details; distractors will omit them.
State the Task Clearly
Tell the AI exactly what you want it to produce. Examples: 'Generate a configuration to fix this issue,' 'List the top 3 causes in order of likelihood,' or 'Provide the show commands to diagnose the problem.' Be specific about the output format: CLI commands, step-by-step instructions, or a troubleshooting flowchart. In the exam, the best prompt will include a concrete task like 'Write the CLI commands to configure trunking on both switches.' Vague tasks like 'Help me with VLANs' are incorrect.
Set Constraints and Format
Specify any limitations: 'Use only Cisco IOS syntax (not IOS XR),' 'Do not include SD-WAN or DNA Center commands,' 'Output as a numbered list,' or 'Assume no firewalls are blocking traffic.' Constraints prevent the AI from suggesting solutions that don't fit your environment. In the exam, look for prompts that include constraints—they are more effective. For example, 'Provide the configuration in CLI format, using named ACLs only' is better than 'Write an ACL.'
Review and Refine the Output
After receiving the AI's response, verify its accuracy. Check that commands match the IOS version, IP addresses are correct, and the solution addresses the original problem. If the output is incomplete or off-target, refine your prompt with additional context or constraints. Example: 'The ACL you generated permits all traffic. I need it to permit only HTTP and HTTPS from subnet 10.1.1.0/24. Please rewrite.' The exam may test your ability to choose a follow-up prompt that effectively narrows the AI's focus.
Scenario 1: Troubleshooting OSPF Adjacency Failure
A network engineer receives an alert that two core routers (R1 and R2) cannot form an OSPF adjacency. The engineer uses a generative AI assistant to speed up diagnosis. They craft a prompt: 'Act as a senior network engineer. Router R1 (IOS XE 17.6) and Router R2 (IOS XE 17.6) are connected via GigabitEthernet0/0 with IPs 10.0.0.1/30 and 10.0.0.2/30. OSPF is configured with network 10.0.0.0 0.0.0.3 area 0 on both. The neighbor state is stuck in INIT. List the top 3 causes and the show commands to verify each.' The AI responds with: (1) mismatched subnet masks – 'show ip ospf interface gigabitethernet0/0', (2) passive interface – 'show running-config | section router ospf', (3) ACL blocking OSPF packets – 'show access-lists'. The engineer runs the commands, finds that interface Gi0/0 is passive under OSPF, removes the passive command, and the adjacency forms. Without a specific prompt, the AI might have suggested generic OSPF troubleshooting steps, wasting time.
Scenario 2: Generating a VLAN Configuration for a New Switch
A junior engineer needs to configure a new access switch. They prompt: 'You are a Cisco network engineer. Create a configuration for a Cisco 2960 switch that creates VLANs 10, 20, 30, names them Sales, Engineering, and Guest, sets interfaces Gi1/0/1-10 as access VLAN 10, Gi1/0/11-20 as access VLAN 20, and Gi1/0/21-24 as trunk ports allowing VLANs 10,20,30. Use IOS 15.2 syntax. Output as a CLI script that can be pasted directly.' The AI generates a correct configuration. The engineer reviews it, adjusts the trunk allowed VLAN list to include native VLAN 99, and applies it. The prompt's constraints (specific IOS version, output format) ensured the config was ready to use. Without them, the AI might have used different syntax or omitted the trunk configuration.
Scenario 3: Interpreting a 'show tech' Output
A core router is experiencing intermittent packet loss. The engineer uploads the 'show tech' output to an AI tool and prompts: 'Analyze this show tech output from a Cisco ASR 1000 running IOS XE 17.9. The symptom is intermittent packet loss on interface TenGig0/1/0. Identify any anomalies in the interface counters, CPU usage, or memory. Provide a summary in bullet points.' The AI highlights high input errors on that interface, suggesting a Layer 1 issue. The engineer then inspects the physical cable and finds a damaged fiber. This approach saved hours of manual log analysis. The key was a focused prompt that specified the symptom and the desired output format.
Common Pitfalls in Production
Engineers often forget to specify the IOS version, leading to commands that don't exist on the actual device. Or they ask for 'the best way to configure OSPF' without mentioning network type, resulting in a generic answer. The exam tests your ability to avoid these mistakes by choosing prompts that are complete and specific.
What the 200-301 v2.0 Tests on This Objective
Exam Objective 5.2 states: 'Given a network scenario, select the most effective generative AI prompt to produce a desired output.' You will be presented with a short network scenario (e.g., a router not forming a BGP session) and four candidate prompts. You must choose the prompt that is most likely to yield a useful, accurate, and complete response from a generative AI model. The exam does NOT test your ability to use a real AI tool; it tests your understanding of prompt engineering principles applied to networking.
The 3-4 Most Common Wrong Answers and Why Candidates Choose Them
Vague prompt: 'Why is my network down?' – Candidates choose this because they think AI can read their mind. Wrong: The AI needs context. A better prompt would specify the device, protocol, and symptom.
Overly broad prompt: 'Tell me everything about OSPF' – This yields a textbook explanation, not a solution. Candidates pick it because they want comprehensive information, but the task requires a specific fix.
Missing role assignment: 'Generate an ACL for 10.1.1.0/24' – Without role, the AI might produce a non-Cisco ACL. Candidates forget that role primes the AI for Cisco syntax.
No output format: 'Help me with VLAN configuration' – The AI might give a paragraph instead of CLI commands. Candidates don't realize they need to specify 'output as CLI commands.'
Specific Values, Defaults, and Command Outputs That Appear on the Exam
The exam will use realistic but simplified scenarios. For instance, you might see a prompt for 'show ip ospf neighbor' output, and you need to choose a prompt that asks for interpretation of that output.
Expect prompts that include specific IOS versions like 'IOS 15.2' or 'IOS XE 17.3' – the correct prompt will match the version.
Common tasks: generating ACLs, configuring VLANs, troubleshooting OSPF adjacencies, and interpreting 'show' commands.
Decision Rule for Scenario Questions
When choosing the best prompt, use the RACT framework: - Role: Does it assign a persona? (e.g., 'You are a CCIE') - Audience/Context: Does it include relevant network details? (e.g., IPs, device models, IOS version) - Constraints: Does it specify format or limitations? (e.g., 'output as CLI commands') - Task: Does it state exactly what to produce? (e.g., 'List the top 3 causes')
The prompt that satisfies the most RACT elements is the best. Usually, the correct answer includes all four; distractors miss one or more.
New in CCNA v2
This objective is entirely new. In v1.1, the exam focused on manual configuration and troubleshooting. Now, Cisco recognizes that AI assistants are part of the network engineer's toolkit. Candidates who studied old materials will be unfamiliar with prompt evaluation. The 5-domain blueprint allocates 10% to 'AI and Network Operations and Management,' which includes this objective. You must practice selecting effective prompts to succeed.
[CCNA v2 NEW] Generative AI prompting for network tasks is a new objective (5.2) in the 200-301 v2.0 exam, reflecting the use of AI in network operations.
An effective prompt includes Role, Context, Task, and Constraints (RACT framework).
Always assign the AI a role, such as 'Act as a senior Cisco network engineer.'
Provide specific network details: device models, IOS versions, IP addresses, and symptoms.
State the desired output format: CLI commands, numbered list, or step-by-step instructions.
Avoid vague prompts like 'Help me with OSPF' – instead say 'Generate OSPF configuration for area 0 on router R1.'
Common exam distractors omit context or constraints, leading to generic AI responses.
Use follow-up prompts to refine AI output if the initial response is too broad or incorrect.
These come up on the exam all the time. Here's how to tell them apart.
Effective Prompt
Includes role: 'Act as a CCIE'
Provides context: device, IOS, IPs, symptom
States specific task: 'List top 3 causes'
Sets format constraint: 'Output as CLI commands'
Yields actionable, accurate solution
Ineffective Prompt
No role assignment
Vague context: 'My router is down'
Broad task: 'Help me with OSPF'
No format specified
Produces generic textbook explanation
Mistake
Generative AI can fully replace manual troubleshooting.
Correct
AI assists but cannot replace human judgment; always verify AI-generated configurations in a lab before production. Cisco expects engineers to validate outputs.
Candidates overestimate AI capabilities because of marketing hype.
Mistake
Any natural language prompt works equally well with AI.
Correct
Effective prompts are structured with role, context, task, and constraints. Vague prompts yield generic or incorrect answers.
People assume AI understands intent like a human, but it follows literal instructions.
Mistake
You don't need to specify IOS version; the AI will guess correctly.
Correct
Different IOS versions have different command syntax (e.g., 'ip ospf network' vs 'ipv6 ospf'). Always include the version to avoid wrong commands.
Candidates think all Cisco IOS is the same, but features and syntax vary.
Mistake
The best prompt is the shortest one because AI can infer details.
Correct
Longer, more detailed prompts produce better results. The exam rewards prompts that include multiple context elements.
Brevity is valued in human communication but not in AI prompting.
Reveal each answer, then mark whether you got it right. Score 60%+ to unlock the next chapter.
No. The exam does not require you to interact with an actual AI. Instead, you will be given multiple-choice questions where you select the best prompt for a given network scenario. You need to understand prompt engineering principles—like providing context, specifying format, and assigning a role—to choose the correct answer. Familiarity with tools like ChatGPT or Cisco's AI Assistant can help, but it's not mandatory.
RACT stands for Role, Audience/Context, Constraints, and Task. It's a mnemonic to remember the four essential components of an effective prompt. Role primes the AI to act as a network expert. Context provides relevant network details. Constraints limit the output (e.g., 'use Cisco IOS syntax'). Task states exactly what you want. On the exam, the best prompt will include all four elements.
No, and attempting to do so would violate Cisco's exam policies. The exam is proctored and designed to test your knowledge, not your ability to use AI. This objective is about evaluating prompts, not using AI live. You must learn the material yourself. AI is a study aid, not a shortcut.
Cisco added this objective because network engineers increasingly use AI tools for daily tasks like generating configs, interpreting logs, and troubleshooting. The 2026 blueprint reflects the industry shift toward AI-assisted operations. By testing prompt evaluation, Cisco ensures that certified engineers can leverage AI effectively and responsibly.
You'll likely get the question wrong. The exam presents one clearly superior prompt that includes role, context, task, and constraints. Distractors are vague, missing one or more elements. For example, 'Help me with OSPF' is a distractor because it lacks context and a specific task. Always look for the most detailed and structured prompt.
Not always, but often. For configuration tasks, specifying 'output as CLI commands' ensures you get ready-to-use syntax. For analysis tasks, you might want a bullet list or step-by-step instructions. The key is to match the output format to the task. The exam will test whether you choose the appropriate format for the scenario.
The 'AI and Network Operations and Management' domain is 10% of the exam (about 10-12 questions). Objective 5.2 (prompting) is one part of that domain, so you can expect 2-4 questions on this specific skill. It's a small but important slice—don't skip it.
You've just covered Generative AI Prompting for Network Tasks — now see how well it sticks with free CCNA 200-301 practice questions. Full explanations included, no account needed.
Done with this chapter?