810-110 AITECH study guide: topics, format and a 4-week plan

The Cisco AI Technical Practitioner exam (810-110 AITECH v1.0) is a 60-minute, English-language exam that leads to the Cisco AI Technical Practitioner certification. Its topics are general AI concepts rather than Cisco products: it tests generative AI models, prompt engineering, AI ethics and security, AI-assisted data analysis, AI for code and workflow automation, and agentic AI. It suits technical IT professionals, solutions architects and technical leads who use or champion AI tools, and it has no formal prerequisites.

810-110 at a glance

Details as of September 2026 — confirm on the official exam page before booking.

Who this exam is for

Cisco lists three groups: technical IT professionals modernising their workflows with AI-assisted coding and data analysis, solutions architects evaluating how to customise and deploy AI, and technical leads or managers overseeing AI adoption. The companion training course widens this to network engineers, data analysts, AIOps specialists and business process analysts.

Cisco states there are no prerequisites for the training, and none are listed for the exam. You do not need to be a developer, but you should be comfortable reading short code snippets and reasoning about how an AI tool fits into a workflow. Cisco says the certification is valid for three years.

What the exam covers

The 810-110 exam topics list six domains. Cisco notes that related topics outside the list may also appear.

1.0 Generative AI Models (20%)

Model families such as large language models and diffusion models, and what each is used for. Expect trade-off questions on cloud-hosted versus locally hosted models (cost, latency, privacy, scalability), on context windows and token limits, on choosing a model from a model hub for a use case such as reasoning or multimodal input, and on how RAG uses embeddings and a vector database. Learn the reasoning behind each trade-off rather than specific model names or sizes.

2.0 Prompt Engineering (15%)

Prompt structure (roles, instructions, constraints) and techniques such as iterative, chained and few-shot prompting, including how prompts differ for text, image and audio output. The domain also covers prompt injection types and defensive prompting, including ways to reduce hallucinations. Know the difference between direct and indirect injection and which mitigations address each.

3.0 Ethics and Security (15%)

Responsible AI principles (fairness, transparency, accountability, bias mitigation and safety), protecting corporate data in AI systems, AI-specific threats including misinformation, and governance through policy, risk management and compliance. Questions tend to present an organisational situation and ask for the most appropriate control or policy.

4.0 Data Research and Analysis (10%)

The smallest domain: how AI supports exploratory data analysis, automated data preparation (quality checks, formatting, transformation, cleaning), privacy controls that stop sensitive data being exposed to an AI tool, and AI-assisted research, ideation and drafting. Focus on the privacy task, because it connects to the ethics domain and is where scenario questions have the clearest wrong answers.

5.0 Development and Workflow Automation (20%)

AI across the software development lifecycle, from requirements and prototyping to testing and deployment. It covers code generation and rapid prototyping, designing and monitoring AI workflows, how token usage and context management affect cost, latency and quality, and how AI helps with debugging, error handling and documentation. The token task overlaps with 1.3, so study them together.

6.0 Agentic AI (20%)

What makes an application agentic rather than simply generative, agent design principles and orchestration, the Model Context Protocol (MCP) and its primitives, human-in-the-loop strategies, and how agents transform and map data between tools. For MCP, be clear on the roles of host, client and server and what each server primitive exposes. For human-in-the-loop, know where approval points belong and why.

A 4-week study plan

Five domains carry 15–20% each and one carries 10%, so no area can be skipped.

Week 1: models and prompting (35%). Study model families, hosting trade-offs, context windows and RAG. Then practise prompt structure, few-shot and chained prompting on a real assistant, and write down one example of each prompt injection type with its mitigation.

Week 2: ethics, security and data (25%). Map each responsible AI principle to a practical control. Read your own organisation's AI or data policy if it has one, and work through how an AI-assisted analysis could leak sensitive data and what would prevent it.

Week 3: development and agents (40%). Use an AI coding assistant to build and test a small script, noting where it helped and where it failed. Read the MCP specification's overview of hosts, clients, servers and primitives, and sketch an agent workflow with a human approval step.

Week 4: mocks and repair. Take a full-length mock under a 60-minute limit, review every explanation, spend two days on your weakest domain, then take a second mock.

Where candidates lose marks

• Answering from one vendor's product. The exam is about concepts. When an option names a feature of one tool, check whether the question asked for a principle instead.
• Blurring generative and agentic. A model that drafts text is generative; a system that plans, calls tools and acts toward a goal is agentic. Task 6.1 tests exactly this boundary.
• Treating RAG and fine-tuning as interchangeable. RAG grounds a model in current external data at query time; fine-tuning changes the model's behaviour. Scenario wording usually signals which one fits.
• Mixing up MCP roles and primitives. Tools, resources and prompts are exposed by servers for different purposes. Confusing them is an easy way to drop marks in the agentic domain.
• Choosing autonomy over oversight. In human-in-the-loop questions, the right answer often places a person at the irreversible or high-impact step, not at every step.
• Relying on fixed numbers. Context window sizes and prices change often. The exam topics ask about the effect of token limits, not specific figures.

Practising for 810-110 on CertifyAIx

CertifyAIx has 471 original 810-110 questions, written against Cisco's v1.0 exam topics and organised by its six domains, each with an explanation for every option. [How our practice questions are written](/blog/how-certifyaix-practice-questions-are-written) sets out the method.

• 3 full-length mock exams of 50 questions in 60 minutes (Cisco does not publish the question count, so 50 is our estimate)
• 21 mixed domain papers and 25 topic papers of up to 25 questions each (60 minutes per 25)
• Single- and multiple-answer formats

See the [810-110 practice questions](/certifications/cisco-ai-technical-practitioner-810-110). For a wider view of the options, read [how to choose an AI certification](/blog/how-to-choose-an-ai-certification).

Frequently asked questions

How hard is the Cisco AITECH exam?

It is a broad, concept-level exam rather than a deep technical one. Candidates who already use AI assistants in their work usually find prompting and development familiar and spend most of their effort on MCP, governance and security.

How long should I prepare for 810-110?

Three to six weeks suits most technical professionals. Allow longer if agentic AI and MCP are new to you, since those topics account for a fifth of the exam and have their own vocabulary.

What is the passing score for 810-110?

Cisco does not publish passing scores, stating that questions and passing scores can change without notice. Results are reported as pass or fail.

How long is the certification valid?

Cisco states the certification is valid for three years. Cisco also notes that the exam can be used toward recertification.

Are CertifyAIx questions real Cisco exam questions?

No. Every question is original, written to rehearse the topics in Cisco's published exam topics document. We never use live or past exam content.

CertifyAIx is independent and not affiliated with Cisco. Cisco AI Technical Practitioner is a trademark of its owner. All CertifyAIx practice questions are original.