🎯Guidance

Our guidance feature will help you set up your AI agent to respond exactly how you want it to, tone, scenarios and when you want it to handover to your team.

What is guidance?

Guidance is a way to instruct and control your AI agent's responses so that they can more closely align with your brand's communication style, support policies and business processes.

This all results in higher quality responses, better brand identification and consistency and faster resolutions for your users or customers.

You can use guidance to ensure your AI agent uses the right terminology, does the right thing and escalates issues to your human agents at the right time.

You'll always be able to see what guidance was used by your AI agent and you can change it at any time.

Guidance is currently only available for the following channels: Intercom (chat and email) and Zendesk (messaging and tickets/emails). More channels coming soon.

What are the types of guidance and when should I use them?

There are 3 types of guidance you can add to your AI agent:

Communication style

This guidance controls the phrases, vocabulary/terminology and writing style of your AI responses to ensure that your AI agent responds like your best agents.

All of the communication style guidance added will be used in every response by your AI agent.

You should use this for instructions like:

Use inclusive terms such as “folks” or “everyone” and avoid gender-specific pronouns when uncertain.

Spell out acronyms on first use, then use the acronym only (e.g., “Natural Language Processing (NLP)”).

Use one or two emojis in your responses to create a friendly and engaging tone.

Context & clarification

This guidance should be used to define follow-up questions that should be asked by the AI agent in certain scenarios in order to gather more information to better answer questions.

Only one piece of the context & clarification guidance will be used in each response by your AI agent.

You should use this for instructions like:

If a user is unsure how to reset their API key, verify whether they want to regenerate the existing key or create an additional one.

If a user asks how to export their data, first clarify whether they need a CSV file or a full JSON export.

When a user reports a bug, first ask for the following information before continuing troubleshooting: 1) what device are you using? 2) URL where the bug is occurring; 3) when was the bug first noticed.

Handover & escalation

This guidance should be used so your AI agent knows when a specific topic of conversation should be handed over to one of your human agents for resolution.

Only one piece of the handover guidance will be used in each response by your AI agent.

You should use this for instructions like:

If a user requests a custom contract, MSA, or NDA, always hand the conversation over to a human agent.

If a subscriber submits a GDPR “right to be forgotten” request, always hand the conversation over to a human agent.

If a user requests a return of a product, always hand the conversation over to a human agent.

How do I add or enable guidance?

To add or enable guidance for your AI agent:

1. Login to your Dashboard.

2. Go to Improve > Guidance.

3. Select the type of guidance you want to add and click the "+ New" button.
4. Enter your guidance and edit the title to make it easier for you to find later.

5. Click "Save".

6. Your guidance will be applied instantly, to test it out, ask a question in the adjacent chat window and see how it responds.

How do I write the best guidance?

When writing your guidance, there are several things you should consider to get the best results and the most obedient AI:

Outcome-oriented

Work backwards from what you want the AI to achieve to come up with your instruction. For example:

❌ Before (vague and ineffective):

“Make sure the AI customer support agent understands our different product types before answering.”

✅ After (clear and structured):

“If a customer asks about the ‘add file’ feature, first ask which product they are using before responding. Then, provide product-specific instructions based on their answer.”

Use simple and unambiguous language

Overly complex instructions can lead to inconsistent responses. Write as if you’re training a new support agent—be direct, specific, and easy to understand.

❌ Before (unclear and open to interpretation):

“Be professional but also friendly, keeping responses engaging.”

✅ After (defined and actionable):

“Use a professional yet approachable tone. Keep responses concise, avoid jargon, and use reassuring language when addressing customer frustrations (e.g., ‘I understand how that can be frustrating. Here’s how we can resolve it’).”

Provide context and examples

Guidance performs best when it can be understood as to when and how to be applied. Use words like “if,” “when,” and “then” to define conditions, and include clear examples.

❌ Before (lacks context):

“When customers ask about pricing, make sure you answer correctly.”

✅ After (context-driven and detailed):

“If a customer asks about pricing, first check if they mention a specific plan. If they don’t, ask which plan they’re interested in before providing details. Always refer to prices as ‘starting at [lowest tier price]’ unless the customer specifies a plan.”

One instruction per piece of guidance

Each piece of guidance should address a single objective. If one instruction tries to convey more than one objective, it should be split into separate instructions. Avoid mixing multiple instructions, as this can make it harder to apply them correctly.

❌ Before (too broad and unfocused):

“Use a fomal tone, clarify questions before answering, and escalate returns issues to an agent.”

✅ After (one clear purpose per entry):

Communication guidance: “Maintain a fun, friendly tone by using positive language and avoiding jargon.”

Context & clarification guidance: “If a customer asks about pricing but does not specify a plan, ask for their ticket volume before proceeding.”

Handover & escalation guidance: “If a customer mentions ‘return’ or ‘cancellation,’ immediately escalate the conversation to a human agent.”

Speak directly to the AI agent in your guidance

When writing guidance, avoid referring to the AI customer support agent in the third person or commenting on how it should modify its responses. Instead, write as if you are speaking directly to it, telling it exactly what to do.

❌ Before (third-person and indirect)

“If the AI answer tells the customer to uninstall and reinstall the app, then rewrite the answer to remove that information, as reinstalling is not a valid troubleshooting step.”

✅ After (direct and actionable)

“Never tell the customer to uninstall and reinstall the app. Reinstalling is never a valid troubleshooting step, and the AI customer support agent should never communicate this to the customer.”

Provide example phrasing

When writing instructions, it is often a good idea to demonstrate correct and sometimes incorrect behaviours as part of the instruction using examples.

❌ Before (tells AI customer support agent how to respond without an example):

"If a customer is frustrated, acknowledge their feelings and use calming language to show you care."

✅ After (helps AI customer support agent understand how it should and should not respond):

"If a customer is frustrated, acknowledge their feelings and use calming language to show you care. For example: "I understand this is frustrating, and I'm sorry for the trouble. Let's work through this together to find a solution." "I know this isn't ideal, and I'm sorry for the inconvenience. I'm here to help and will get this sorted as quickly as possible.""

Correct formatting

Always ensure proper grammar is used, including proper sentence grammar and punctuation, including full stops at the end of each instruction.

Keep guidance concise

Try and keep your guidance under 75 words to ensure it is followed most closely, the longer the guidance the harder it will be for the AI to adhere to.

Write guidance in English

While guidance can be written in any language, it will perform best when written in English.

How to optimize your guidance

When writing guidance there are a number of things you need to consider (see here).

However, to make this easy for you we have added in an "Optimize" feature which you can click to rewrite your guidance in an optimal way that will be most effective for the AI agent.

We highly recommend you use this feature to get the best results from your AI agent!

How can I test my guidance?

To test your guidance, you can use the widget adjacent to where the guidance has been entered. Ask a question that would result in the guidance being applied and see how the AI agent responds.

You can always edit your guidance after if it doesn't give the correct effect.

What should guidance not be used for?

Guidance is a powerful feature to control your AI agent, but there are still things it shouldn't be used for:

❌ Taking actions (apart from human handover)

You shouldn't instruct your AI agent to do things like: routing to a specific person in the team, tagging a conversation, closing a conversation etc.

The only action it can take via guidance is initiating a handover to a team member.

❌ Changing language

Guidance should be used to change the language e.g. "respond in Spanish" or "respond in the language you are spoken to in", for this you should use our language feature.

❌ Adding information or knowledge content

You shouldn't give your AI agent knowledge snippets via guidance, these should be added via the Knowledge section of the dashboard or as Custom Answers.

❌ Instructing the AI agent to "not make things up"

Not hallucinating or making things up is a core ability of My AskAI, you do not need to add further instructions for this as guidance.

❌ Controlling where your AI agent will find answers from

Guidance will not be able to direct your AI agent to certain areas of your knowledge or website in order to answer questions.

How do I know if guidance has been used on a response?

To see which guidance is used for a response by your AI agent, you can inspect the conversation.

To do this:

1. Go to your Dashboard.

2. Go to Improve > Inspect & Debug.

3. Search for your conversation (by email, date, ticket ID, or content).

4. Click on the conversation and scroll to the response, you should see a section called "Guidance followed" which will tell you which guidance was used in that response.

How many pieces of guidance can I add?

You can add up to 30 pieces of guidances under each category of guidance, so 90 in total.

If this is insufficient, please get in touch with us via our live chat and ask to talk to a person.

Is it possible to create multi-step guidance?

This is an advanced form of using guidance.

If you want to create guidance that follows a couple of different steps e.g. ask for more information then handover to an agent, you can, here is an example of how you would set this up:

Create a Context & clarification piece of guidance, for example:

When a user reports a bug, first ask for the following information before continuing troubleshooting: 1) what device are you using? 2) URL where the bug is occurring; 3) when was the bug first noticed.

Followed by a Handover & escalation piece of guidance, for example:

If a user provides you with details about a bug, namely a device, URL and when the issue was noticed.

As you can see here, the answer format from the Context & clarification response is being used in the Handover & escalation guidance, giving a result like this:

How do I update my guidance?

To update your guidance:

1. Log into your Dashboard.

2. Go to Improve > Guidance.

3. Locate the guidance you want to update and click the pencil icon next to it.

4. Edit the guidance and click "Save".

5. Your new guidance will be applied instantly.

How do I delete guidance?

To delete your guidance:

1. Log into your Dashboard.

2. Go to Improve > Guidance.

3. Locate the guidance you want to delete and click the pencil icon next to it.

4. Click the trash can icon then click confirm.

5. Your guidance will be deleted instantly.