Can AI Handle a Crisis? We Gave AI Service Scenarios to ChatGPT to See How It Responded

Download Now: The State of AI Report
Kasia Kowalska
Kasia Kowalska



PwC says that 69% of business leaders go through three company crises within five years. Not only is this stressful on a personal level. It also takes a toll on the entire organization.

service personnel using AI for crisis management

Wrapping your head around everything that’s happening online and offline is tough and practically impossible without using the right technology.

Luckily, ChatGPT can now help companies deal with crisis management more effectively. Read on to find out how far you can rely on it when it comes to preventing and resolving brand crises.

Download Now: The Annual State of Artificial Intelligence Report

ChatGPT and Customer Service

Today, 56% of companies use AI in customer service. Here are some of the most common ways in which ChatGPT can aid your customer service teams.

Help Customers Find Answers Independently

Not every interaction must be handled by a human agent. Some can be automated using self-service or forwarded to a chatbot. In fact, over 50% of businesses already use chatbot technology. This makes perfect sense since text is already the preferred way of communication for over 85% of customers.

Guarantee 24/7 Availability

ChatGPT is available constantly, unlike human agents. This means that it can handle customer queries round-the-clock.

Manage High Volumes of Questions More Efficiently

There are times when the number of customer queries grows sharply, for example, when a product experiences technical difficulties. Using a solution like ChatGPT can help answer more user questions faster.

According to HubSpot’s State of AI report, generative AI can save customer service teams over two hours a day by responding to customer tickets.

ChatGPT and Crisis Management

The speed at which ChatGPT processes and serves information makes it an ideal assistant for teams tackling an unraveling brand crisis. Here are a few ways it can help.

Immediate Access to Information

Not everyone realizes yet that ChatGPT can be used for more than general market research and accessed beyond OpenAI’s website.

In fact, by using a GPT-powered solution like ChatSpot, you can connect it with your CRM and other systems. As a result, whenever you enter a prompt, it has the ability to reach into your internal data and draw up results relevant to your business.

For example, if you uncover that there’s an issue with your service, you could ask ChatGPT to report how many mentions of crisis-prone terms there were in recent conversations between clients and customer support. Then, you could quickly cross-refer these numbers to a bird’s eye view of your entire customer base and check how it affects sales, customer sentiment ratings, and other factors.


AI in crisis management; ChatSpot lets you access quick CRM summaries and analyses

Image Source

All this allows your team to make data-driven decisions and agree on the right course of action for the crisis at hand.

Monitoring Social Media

Brands can use ChatGPT to continuously monitor their social media channels in case any mentions related to a crisis appear.

It’s done in real-time, and the AI tracks brand-related mentions, hashtags, and keywords. This is very helpful in spotting potential problems and addressing them before they escalate.

ChatGPT can even be programmed to send alerts whenever it comes across keywords that might indicate a crisis or have a negative sentiment. This allows businesses to react quickly.

Drafting Statements and Responses

Stress can be debilitating, even for those who tackle customer dissatisfaction and reputational crises on a daily basis. Sometimes, this translates into your team’s inability to craft a relevant response quickly.

Here’s where ChatGPT steps in. After you’ve agreed on how to handle the situation, you can ask it to write a statement or response, which clearly communicates to clients how you’re going to mitigate the issue.

You can instruct the bot to follow your brand’s tone of voice and avoid any crisis-prone terms.

Testing ChatGPT in a Crisis

It’s test time! Let’s see how ChatGPT would handle different crisis scenarios.

Scenario 1: The server is down.

Prompt: “Create a message informing users that the company server is down. Let them know they won't be able to use the product for 5 hours while the engineering team is working on fixing the problem. Apologize to users for the inconvenience caused.”

AI in crisis management; asking ChatGPT to create a message explaining to customers that the server is down

What I think: The message seems to be unnecessarily long. There are some repetitions — it says twice that the users won’t be able to use the product. This might add to their frustration instead of calming them down.

Starting the message with “Dear Valued Users” sounds robotic. But to be fair, if ChatGPT was integrated with a CRM, it would probably be able to use customers’ names, making the email sound more personal.

On a positive note, it’s well written and includes elements like an apology and a reassurance that the team will keep customers posted about work progress. It also emphasizes how important it is for the company to keep its customers happy and satisfied, which is great.

Scenario 2: Responding to a complaint from an influencer

Prompt: “Help me craft a thoughtful social media response to a client who is an influencer with 500,000 followers on Instagram. The client is unhappy with the shoes they've purchased through my store.

“Please inform them that we're happy to provide a refund. We want to reassure the influencer and their followers that we take customer satisfaction seriously. Use an informal tone of voice.”

AI in crisis management; asking ChatGPT to respond to a data leak

What I think: I like how friendly the message turned out. It clearly shows that the company didn’t take offense to the negative post, which is exactly what I asked for in the prompt.

I do believe that it could be shorter, and some sentences are a bit repetitive. Still, that would be an easy fix for me. Personally, I do have one concern.

Admitting straight off that the brand cares about the influencers’ following could pose the question — would they take the situation as seriously if the unhappy customer wasn’t an internet personality?

I would also disregard the hashtag suggestions, as they don’t relate to the topic.

Generally, though, I am satisfied with the outcome and do believe it would help me out if I were in such a situation.

Scenario 3: Creating an action plan for a customer data leak

Prompt: “Help me create an action plan for a customer data leak. The only information that got leaked was customers' phone numbers and email addresses.”

AI in crisis management; asking ChatGPT to create an action plan for a customer data leak-2

What I think: This plan can be a good starting point for action. It covers all key aspects, such as immediate response, containment, customer notification, regulatory compliance, mitigation, customer support, communication, learning from the incident, and ongoing monitoring.

It’s rather generic, but so was the prompt. While it provides a solid framework, it’s vital to remember that there are no “one-size-fits-all” plans and that they should be adjusted to an organization's needs. This includes making sure that there are enough resources and expertise to execute it effectively.

Scenario 4: Reassuring the public of a brand’s political neutrality

Prompt: “Our company's former CEO has affiliated themselves with a political party. Please draft a concise, step-by-step plan on how we can assure our clients that we are and never will engage in politics.”

AI in crisis management; asking ChatGPT to respond to a comment about politics

AI in crisis management; asking ChatGPT to create a crisis prevention plan

Image Source

What I think: While I’m not part of a PR team myself, I believe that the recommendations are clear, easy to follow, and — mostly — objective.

The reason why I’ve said ‘mostly’ is because I feel that the plan is a bit presumptive. Namely, Step 4 presumes that the CEO must have associated themselves with a controversial organization, which doesn’t necessarily have to be the case.

Still, I do understand that ChatGPT referred to similar situations, many of which must have been crisis mitigation strategies, not crisis prevention plans. Having taken that into account, I think the plan is a good starting point for any brand team discussions.

Scenario 5: Responding to sabotage from a competitor

Prompt: “A competitor wrote a series of negative comments on the brand's social media channel, which caused concern among current customers and prospects.

“Please come up with an action plan to address the issue and provide a statement that the company could share with their client base.”

AI in crisis management; asking ChatGPT to respond to sabotage from a competitor

AI in crisis management; asking ChatGPT to respond to sabotage from a competitor

AI in crisis management; asking ChatGPT to respond to sabotage from a competitor

What I think: The plan provides a good overview of the steps that a company should take to handle the crisis, but there are a few areas for improvement, for example:

  • Engage with Legal and Compliance teams. It’s a critical step, but it lacks information that both teams should be involved in the process as early as possible. Legal experts can help to come up with the right response in compliance with the law.
  • Craft a strategic response. It’s great that ChatGPT suggested handling misinformation without direct confrontation. However, it would be worth mentioning that the response should be in line with the brand’s tone of voice. It wouldn’t hurt to include brand guidelines to ensure the response is consistent and professional.

Regarding the statement, while it mentions that the negative comments were posted by a competitor, it lacks context, which might undermine the brand's credibility. It also doesn’t include instructions on how customers can contact the company in case of any questions.

The response could do with a bit more human touch. While it’s nice to include the name, title, etc., the message could benefit from expressing personal commitment to tackling the issue as quickly as possible.

Scenario 6: Preventing a key client from churning

Prompt: “I‘ve just learned that my company’s key client is in talks with our competition. I heard that they're being offered a better price-per-user.

“Please write an email where I thoughtfully and respectfully mention that I've heard that they might be leaving and want to encourage them to reconsider and renegotiate our existing deal. I know the Account Manager well, so the message can be a bit more informal than the standard business email.”

AI in crisis management; preventing a key client from churning

What I think: I like most of the email content. ChatGPT went a bit over the board with a few terms, for example, “what’s cooking.” Other than that, it did a good job of adding a more ‘personal’ feel to the message.

What I don’t like is the email title — it’s slightly confrontational and too forthcoming for my taste. On the other hand, considering how the client might drop the company at any given moment, it is actionable, so maybe there’s something to it. In conclusion — once again, good job, ChatGPT.

ChatGPT in a Crisis

ChatGPT’s ability to craft messages quickly and to suggest general step-by-step plans makes it a helpful tool for companies trying to tackle or prevent a crisis. That said, unless you double-check its output and grant access to your internal sources like your CRM, the responses will be too generic.

Overall, it’s a good starting point, as it helps resolve or prevent crises faster, but still, human touch is required.

New Call-to-action

Related Articles

New research into how marketers are using AI and key insights into the future of marketing with AI.

    Service Hub provides everything you need to delight and retain customers while supporting the success of your whole front office