We’ve done some things with Breeze AI, and this is what works!
Today: Ask Breeze – Using artificial intelligence in automated workflows.
One feature of HubSpot's Breeze Package is the Ask Breeze functionality in Workflows. It allows workflow creators to formulate AI prompts using variables of the enrolled objects (Contacts, Companies, Tickets etc.). The result is then stored in a response field and can be used for further processing.
It sounds boring and nerdy, but we found some significant value in this little mousy box:
The Problem: While most businesses need a contact page on their website allowing visitors to contact your organisation, it comes at the cost of receiving unsolicited offers from random service providers around the planet. Sometimes, these submissions have some value, but most would be considered "Spam". So, before sending these contacts and their messages to the internal teams, qualifying them for their intent could make sense. For us, it does.
Approach: So, we have formulated the below prompt in an “Ask Breeze” workflow element by passing on the message received through the HubSpot form on the website. It’s pretty brief and instructs Breeze to qualify the message’s intent. Is the Author of this message interested in our services or promoting some sort of service/offer to us? The prompt includes instructions on how “Ask Breeze” should reply.
In our case, we only gave two choices: “inquiry” or “sell”.
We set it up to execute this prompt on every form submission and store the return value (“inquiry”/”sell”) in a response field. We then branch on this response field in our form submission workflow. We have a third “unsure” branch if the AI is unsure about the message intent.
We will route each form submission to different teams based on the Ask Breeze response.
Why do we think this is valuable?
It allows Companies to filter out Spam submissions and only forward inquiries to the sales team. The function enables the team to react immediately to inbound requests without screening all submissions.
Although the sample size is still relatively small, we have seen an accuracy of 100% so far.
What else could you do with that?
While we, as software engineers, generally question and try to avoid unstructured data, it is still a reality for many businesses. If you receive unstructured data, be it a message field in a form submission or an email sent to a generic inbox, you can use the Ask Breeze function to qualify the message. This could be checking for complaints, service incidents, inquiries, partnership requests, or simply sorting out spam.
Imagine you are a small enterprise business headquartered in Europe with subsidiaries in Italy, France, the United States, and China. Traditionally, you have a sales@example.com email address that existing customers frequently use to request a quote.
Someone at HQ reception monitors that inbox, reading all inbound messages and forwarding them to the respective subsidiaries or departments – during EMEA business hours.
With Ask Breeze, you can automatically route these inbound requests based on detected topic, intent, or language - 24 by 7.
We’ve tried this prompt to determine the 2-letter ISO Code for an incoming message text:
And - who would have thought - It just works perfectly fine!
The value of Ask Breeze in such a scenario is at least twofold:
- Free up employees’ time. Employees will only need to screen and forward messages the AI cannot distribute, significantly reducing workload.
- Increase internal SLAs. Due to time differences, holidays, or other staff shortages, it is not uncommon for a message from Asia or America to reach the local subsidiary with significant delays in case of a sales inquiry that might cost the regional sales team considerable time.
Ask Breeze is available at no extra cost in HubSpot Professional and Enterprise Version for Sales, Marketing, Service, and Operations Hub.
If you are unsure how to set this up, contact us. We can help.