Improving the Conversational Experience – What Controls Are Available?

Long-response behavior in Slack: when Moveworks generates a lengthy response in Slack, the bot appears to take the user to the end of the response rather than the beginning, requiring the user to scroll back up before they can start reading. Is this behavior configurable by Moveworks, controlled by Slack, or something that can be influenced through response design/configuration?

Thanks for writing this up so clearly!

You're asking a question that comes up more as teams add use cases: beyond getting the answer right, who controls how the assistant sounds, how much it says, and how it behaves in each channel? Your acronym example is a good one, because it shows the dividing line well.

Where Things Stand Today

The short version: most of how the assistant talks is set in its reasoning layer, and today that layer is managed by Moveworks.

Tenant-level instructions for tone, length, terminology, and similar behavior are added by our team at your request, through your Moveworks account team or support. That's how your acronym change was made. What you control directly is the content the assistant draws on, the options and labels around its responses, the web experience, and any agents you build yourself.

Here is how that applies to each item on your list:

1. Response length and level of detail

• Managed by Moveworks: Assistant-wide changes go through your account team as tenant-level instructions.

• What you control: The knowledge content the assistant summarizes (how articles are structured and worded affects answers more than most teams expect), and the instructions on any agents you build in Agent Studio.

• Docs: Writing AI-Ready KB Articles | LLM Actions

GPT-5 Note: If your assistant runs on GPT-5, you can also choose which user attributes (location, department, role, plus custom attributes you import) the assistant uses to tailor knowledge answers to each employee. This changes what an answer covers for a given person, not its tone or length.

• Where: Moveworks Setup > AI Assistant > Advanced Settings > User Attribute Personalization

• Docs: Assistant Personalization

2. Tone and conversational style

• Managed by Moveworks: Handled through tenant-level instructions requested via your account team or support.

• What you control: For agents you build in Agent Studio, you write the instructions, so you set the tone of those agents yourself.

• Docs: LLM Action Best Practices

3. Use or expansion of acronyms

This one has two sides:

• How the assistant writes: Managed by Moveworks. Rules like "expand acronyms on first use" are tenant-level instructions (which is how your change was made).

• How the assistant understands what employees type: Partly yours. In Entity Catalog, you can add internal software and tool names, including acronyms and alternate names as aliases, so the assistant recognizes them (currently covers software applications, not general business terms).

• Where: Moveworks Setup > Core Platform > Entity Catalog

• Docs: Entity Catalog Overview | Configuring Custom Entities

4. Formatting and structure of responses

• Managed by Moveworks: The response format is designed for each channel, and tenant-level instructions can influence structure (e.g., preferring short lists over long paragraphs).

• What you control: For agents you build in Agent Studio, you control how their output is structured.

5. Follow-up questions and clarification behavior

• Managed by Moveworks: When the assistant asks a clarifying question versus making a reasonable assumption is part of the core conversation design. Tenant-level instructions can nudge this, but there isn't a direct admin toggle for it.

6. How frequently the bot asks for feedback

There are two kinds of feedback:

• Response feedback: Options on individual responses are part of the standard response design and are not configurable.

• CSAT survey: Runs on a fixed cadence (goes to a random 20% of users active in the last 8 days; capped at once every 180 days per user; sent in Slack or Teams, not on web). If this cadence doesn't work for your tenant, raise it with your account team.

• Docs: Moveworks CSAT Survey

7. How references, actions, or help options are presented

This is where you have the most direct control:

• Get Help and disclaimers: You can rename the Get Help link, add a description on the Get Help tab, and set the disclaimer shown on every response, the intro disclaimer, and the disclaimer on the References tab.

  • Where: Moveworks Setup > Display Configurations > "Moveworks AI Assistant Display Settings & Disclaimers"

  • Docs: Configure Disclaimers

• Handoff options: You control the categories and destinations employees see when they need more help (e.g., filing a ticket, live agent, portal link, chat channel, or email address).

   

  • Where: Moveworks Setup > Handoff > Handoff Settings

  • Docs: Smart Handoff Settings

8. Consistency across Slack, Teams, and Web

The reasoning behind every response is shared across channels, so tenant-level instructions apply the same way everywhere. What differs is how responses are rendered, because each platform supports different formatting and interactive elements. Some settings are also channel-specific (e.g., Moveworks Web App settings apply only to web, and CSAT is only sent in Slack and Teams).

The Practical Process Today

Use the Setup and Agent Studio controls above for what you own, and send assistant-wide behavior changes (tone, length, terminology, clarification style) to your account team or via support, as you did with acronyms. AI Assistant Insights is a great tool to find patterns worth raising.

• Docs: Assistant Insights

Forward Looking

As part of our next generation agentic harness, one of the items we are targeting is putting some of these customizability options into the hands of your administrators. Some of the above items like conversational behavior on a given chat surface are customizations that we are planning alongside this upgrade.

Stay tuned with our product updates, there will be more information in the coming weeks!

Thanks ​@jmcginley. This is helpful and clarifies that tenant-level instructions are currently managed by Moveworks.

A few follow-up questions:

1. Is there a way for customers to view the tenant-level instructions currently configured for their environment? We'd like to review what's already in place before requesting further changes.

2. What's the rationale behind keeping these instructions Moveworks-managed rather than allowing customers to configure them directly, especially since they apply to a dedicated enterprise tenant?

3. Are there plans to expose these instructions through an admin interface so customers can review and fine-tune conversational behavior themselves without going through Moveworks Support for every adjustment?

Having visibility and greater control over these instructions would help us continuously improve the employee experience.