Social media chatbots: Your top FAQs answered

If your social media channels draw customer support inquiries that can easily be solved by self-service and are currently being handled by an agent, then a chatbot could be valuable for you! A chatbot helps handle repetitive questions that don’t need the attention of a human.

Public chatbots should be used to share public information and direct community members to helpful resources. They should be used when community members’ questions can be answered and addressed with a single response or link.

Private chatbots are helpful in conversations where solutions are multi-step or when more information is needed before involving an agent.

For example, a public chatbot could share business hours and locations, a link to a Help Center, or information about a new product release. A private chatbot could provide more specific troubleshooting before directing to an agent.

One thing to always consider, especially when setting up a public chatbot, is your customer’s expectations. For example, if agents are answering some questions on social, customers may get frustrated if a bot answers their question. Additionally, you’ll want to be careful and ensure that your bot doesn’t inadvertently answer a customer question inappropriately. We’ve all seen those bot fails.

It depends on whether your use case is the same. For private messages, you can probably use a similar bot setup to what you use on your website or owned community. You want to ensure that you’re using a tool like Khoros Care to ingest these conversations and escalate them to your agents as appropriate.

To mitigate risks, your agent desktop or omnichannel conversation ingestion and response tool should include robust data masking and redaction capabilities that are configurable to:

• Never display or retain sensitive customer data (e.g., passwords, PII, PCI).

• Selectively display or retain data only for specific agents handling the issue or for a limited time.

You should consider a scaled-back approach for public responses, as customers often don’t want to troubleshoot their issues publicly with a bot.

Though bots can be tailored for specific channels, some functionalities, such as a shared knowledge base, branding guidelines, and governance, can be used across multiple use cases.

Ultimately, any bot is built to alleviate agent time and provide high-level, low-touch support for customers. That’s the same whether the bot resides on your social channels, your website, or your owned community.

From a workflow perspective, it should be easier to manage agent escalations from your social media bot if it is connected to your existing omnichannel system. This enables your team to consolidate efforts across service specialties.

Instead of running two separate NLU models, intent models, or classification models, brands should consider using one core model supplemented by the unique NLP skills or domains needed for specific channels.

You want one core model with as much knowledge and skill as possible, which can be tailored and constrained depending on the environment.

Imagine a Pharmacist Bot. It should have access to the entire library of modern pharmacology and be capable of describing side effects, available generics, drug interactions, and long-term efficacy risks. However, the detailed technical specifics of a drug's FDA clinical trials might not be appropriate for most conversations. In this case, you could equip the bot with a “broad recommendations and awareness” and an “in-depth clinical analysis” skill. Then, customize which knowledge each skill includes and where each skill is used.

Are customer intents different on social media, and how should that affect our automation strategy?

The main thing to note about customer intent on social media is that if someone is coming to social media for help, especially in your comments section, they’re hoping for a quick answer. They may be more willing to undergo multi-step troubleshooting on your website or in your brand-owned community, especially if it results in a resolution.

That said, it’s important to tailor your automation strategy to accommodate public interactions and quick response times on social media.

Are there any other data privacy and security considerations that I should be aware of?

Bots used in public replies on social media should never direct community members to publicly share personal or account information. They should have narrowly scripted functions like “chat in DMs” or “check out our support portal link.” You’ll need to ensure that any conversations needing personal information are directed to a more secure channel.

We have a team ready to help! Khoros Strategic Services is an agency-like team nestled within our larger software organization. We leverage industry expertise and the Khoros platform to uncover customer insights, build data-driven strategies, and provide managed services to help our customers scale. We can help you with Social Care Bot Planning and Analysis to ensure that the strategy behind your bot is in line with the rest of your social brand strategy and care model.